article
stringlengths 0
745k
| abstract
stringlengths 0
92.7k
|
|---|---|
in this paper we present a partial wave analysis ( pwa ) of a high - statistics sample of events corresponding to the production of the @xmath13 system produced in @xmath14 collisions in two modes : @xmath15 and @xmath16 .
this sample size exceeds , by at least an order of magnitude , the largest published sample size of @xmath17 events to date .
previous @xmath17 analyses led to the discovery and/or determination of properties of the @xmath4 , @xmath5 , @xmath6 , @xmath18 and the @xmath7 resonances @xcite . in 1998 , the e852 collaboration reported evidence for the @xmath19 , a @xmath8 exotic hybrid meson with a mass of 1.6 gev/@xmath10 decaying into @xmath11 @xcite .
that analysis was based on 250,000 events of the reaction @xmath20 collected in 1994 .
we report on the analysis of additional data collected in 1995 including 3.0 m events of the reaction @xmath21 and 2.6 m events of the reaction @xmath20 .
the identification of exotic mesons requires a pwa to extract signals .
the @xmath17 system provides a particularly attractive venue for such searches since earlier work suggests a rich spectrum of meson resonances .
if an exotic meson is produced with relatively small amplitude , its interference with nearby well - established resonances is a sensitive search tool . in almost all of the published amplitude analyses of the @xmath17 system ,
the _ isobar _ model was employed
a @xmath17 system with a particular @xmath22 is produced and decays into a di - pion resonance with well - defined quantum numbers and a bachelor @xmath23 followed by the decay of the di - pion resonance .
this assumption is successful in describing many features of the @xmath17 system and is motivated by the observation that the di - pion effective mass spectrum shows prominent resonance production , for example , the @xmath24 and the @xmath25 .
the di - pion resonances considered in this analysis include the @xmath26 , @xmath24 , @xmath25 and the @xmath27 .
we also include parametrizations of @xmath28-wave @xmath29 scattering . in this analysis
, we observe the @xmath4 , @xmath5 , @xmath6 and the @xmath7 in appropriate partial wave intensities and in their relative phase differences .
we also searched for , but find no evidence for , the @xmath8 exotic @xmath19 in either the @xmath15 or @xmath16 mode .
in order to extract reliable information from a partial wave analysis it is important to establish a procedure for determining a sufficient set of partial waves .
failure to include important partial waves in the series expansion may lead to inconsistent results and erroneous conclusions . in this paper
we describe our procedure for determining a sufficient wave set .
that procedure includes a comparison of moments as calculated from pwa solutions with those computed directly from data .
we emphasize that this analysis is similar to that of references @xcite in that the same isobar model assumptions are made but the final set of partial waves used is different .
both analyses make the same assumptions about coherence between different partial waves .
it is possible that relaxing these coherence assumptions could lead to different results in both analyses .
the @xmath17 system with non - zero charge has isospin @xmath31 , and , since no flavor exotic mesons have been found , we assume @xmath32 .
since a state with an odd number of pions has negative @xmath33 parity , the relationship @xmath34 implies positive @xmath35 parity for the @xmath12 system .
the simultaneous observation of the two @xmath12 modes , @xmath15 and @xmath16 , in the same experiment provides important cross checks .
consider the production of resonance @xmath36 and its subsequent decay into @xmath12 via an intermediate di - pion resonance .
if the intermediate di - pion resonance is an isoscalar , then the yield into @xmath16 should be twice that into @xmath15 .
similarly , if the di - pion is an isovector , then there should be equal yields into @xmath16 and @xmath15 .
since the @xmath0 and @xmath1 modes rely on different elements of the detector , any misunderstandings in the acceptance would affect the two modes differently and ultimately lead to inconsistent interpretations of the pwa results .
this underscores the importance of having these two different modes to validate the analysis .
quantum chromodynamics ( qcd ) predicts a spectrum of mesons beyond the @xmath37 bound states of the conventional quark model .
the spin ( @xmath38 ) , parity ( @xmath39 ) and charge conjugation ( @xmath35 ) quantum numbers of a @xmath37 system are : @xmath40 where @xmath41 is the angular momentum between the quarks and @xmath42 is the total quark spin , @xmath43 and @xmath44 . the @xmath22 combinations : @xmath45 are not allowed and are called
_ exotic _ .
lattice qcd @xcite and qcd - inspired models predict that the excitation of the gluonic field within a meson leads to _ hybrid _ mesons , where the gluonic degrees of freedom allow hybrid mesons to have exotic quantum numbers .
the observation of hybrid mesons and measurement of their properties provides experimental input necessary for an understanding of quark and gluon confinement in qcd .
this paper is organized as follows .
section [ sec : experiment ] is a discussion of experimental details , including a description of the apparatus ( [ sec : apparatus ] ) , event reconstruction and selections ( [ sec : data ] ) and distributions in effective mass before and after acceptance corrections ( [ sec : masses ] ) .
an overview of the pwa methodology , including selection of the wave set , is given in section [ sec : pwamethod ] .
section [ sec : pwaresults ] is a presentation of the pwa results for individual partial waves using the wave set used in this analysis and the wave set used in the analysis of references @xcite for both the @xmath46 and @xmath47 modes .
section [ sec : exotica ] focuses on the exotic @xmath8 waves and how the failure to include important @xmath48 waves could lead to misleading conclusions regarding the existence of an exotic resonance .
section [ sec : systematics ] discusses studies of systematic uncertainties and the conclusions are presented in section [ sec : conclude ] .
the e852 experiment used the upgraded multiparticle spectrometer ( mps ) at the alternating gradient synchrotron ( ags ) at brookhaven national lab ( bnl ) .
the apparatus is described in detail elsewhere @xcite and is shown schematically in figure [ mps ] .
the mps was upgraded for e852 with the inclusion of a cylindrical drift chamber ( tcyl ) @xcite and a cylindrical cesium iodide ( csi ) detector @xcite , both concentric with a lh@xmath49 target .
downstream of the target were proportional wire chambers ( pwcs ) for triggering and tracking , a lead / scintillator array ( dea ) and a 3000-element lead glass electromagnetic calorimeter ( lgd ) .
the csi calorimeter consisted of 180 segments 10 longitudinal segments and 18 azimuthal segments .
the dea had the geometry of a picture frame with a rectangular central hole . a scintillation counter ( cpv )
was placed just upstream of and matched the geometry of the dea .
the lead glass blocks of the lgd @xcite were each 45 cm long with a transverse area of @xmath50 @xmath51 and were arranged in a @xmath52 @xmath51 stack with a 2 block @xmath53 2 block beam hole to allow for passage of non - interacting beam particles .
the dea and csi detected low - energy photons that miss the lgd .
the lgd was used to measure energies and positions of photons from @xmath54 and @xmath55 mesons .
the mps included an analyzing magnet with a 1 t field and drift chambers for tracking .
the lh@xmath49 target was 30 cm long and 6 cm in diameter .
the beam was an 18.3 gev/@xmath3 negatively charged beam consisting mainly ( 95% ) of @xmath56 and had a momentum bite @xmath57 of 3% and a momentum resolution @xmath58 of 1% . the target to lgd distance was approximately 5 m. the interacting beam trigger required no signal from a small scintillation counter placed in the path of the beam in coincidence with beam defining counters upstream of the mps . event triggers included all - neutral , all charged and a mix of neutral plus charged particle topologies .
the neutral particle requirement depended on energy and/or multi - photon effective mass information from the lgd .
we refer to the @xmath59 system as the _ neutral _ mode and the @xmath60 system as the _ charged _ mode . in tables
[ cutsn ] and [ cutsc ] the selection requirements and resulting event sample sizes are summarized for the neutral and charged modes respectively . in
what follows we discuss details of these requirements @xcite .
.[cutsn]summary of requirements applied to the data set leading to @xmath61 .
[ cols="<,>",options="header " , ] another arbiter of wave set sufficiency is the comparison of moments , @xmath62 , of the @xmath63 functions as calculated directly from the data and as calculated using the results of pwa fits .
a pwa using a sufficient set of partial waves will reproduce all angular distributions of the data as evidenced by the agreement of the observed and calculated moments .
we define @xmath64 where @xmath65 represents the euler angles of the @xmath17 system and the intensity @xmath66 is determined directly from experiment or computed using the results of the pwa fits .
both of these techniques were used in this analysis .
the @xmath29 isobars included in this analysis are the @xmath67 , @xmath26 , @xmath24 , @xmath25 and @xmath27 where @xmath68 is meant to indicate a @xmath28-wave @xmath29 system as described in @xcite .
in addition , a background wave is included .
the background wave is characterized by a uniform distribution in the relevant decay angles and is added incoherently with the other waves .
we restrict the parent wave set to those with @xmath69 , @xmath70 and @xmath71 .
waves were sequentially removed from the parent set and the resulting change in likelihood ( @xmath72 ) examined @xcite .
three passes were used .
first , waves with an l=0 isobar ( @xmath73 @xmath28-wave and @xmath74 ) were examined .
subsequent passes examined @xmath75 ( @xmath76 ) and @xmath77 ( @xmath78 ) isobar partial waves .
a pwa with the wave in question removed was performed and the resulting likelihood compared to pwa including the wave . if the change in @xmath79 was less than 40 the wave was removed .
figure [ sig_nosig ] shows the change in @xmath79 when a significant , but not a dominant , wave ( @xmath80wave ) is removed compared to when an insignificant wave ( @xmath81wave ) is removed .
certain partial waves that would have been removed by this selection criterion were kept because the existence of signals in these waves is under consideration .
the negative reflectivity exotic @xmath82 partial waves were kept even though they would have been chosen for removal by the above criterion .
the positive reflectivity exotic partial wave failed the selection criterion everywhere above 1.4 @xmath83 and exhibited large fluctuations in intensity below this mass .
this partial wave was also kept and its contribution to the likelihood is further examined below using the final set of partial waves .
finally , the @xmath84-wave ( which failed the selection test ) was kept to serve as a possible reference wave for evaluation of phase differences of negative reflectivity partial waves . based on these criteria
the compiled set includes 35 waves and a background wave .
we refer to this as the _ high - wave set_. the analysis reported in reference @xcite used a wave set consisting of 20 waves including the background wave we refer to this as the _ low - wave set_. table [ thewaves ] lists the waves used in low - wave and high - wave sets .
the waves included only in the low - wave set are labeled with [ l ] in table [ thewaves ] while the waves included only in the high - wave set are labeled with [ h ] and those included in both sets are labeled with [ b ] . the pwa solutions obtained with the two wave sets were used to compute moments that were then compared directly with data . in figure [ fig2 ]
we show the @xmath85 and @xmath86 moment comparisons as a function of @xmath17 mass for the @xmath46 mode for low - wave and high - wave set pwa fits .
we also show the difference between data and pwa calculations of the moments as @xmath87 ( summing differences squared divided by errors squared summed over all mass bins and divided by the number of mass bins ) for various moments .
similar plots are shown in figure [ fig3 ] for the @xmath88 mode .
the moments calculated using the pwa results from the high wave set have better agreement with experimental moments , but there are moments ( such as @xmath85 - shown in part ( a ) of figure [ fig3 ] ) for which agreement is not achieved .
this may be due to the inherent inadequacy of the isobar model in describing the underlying production mechanism .
an examination of figures [ fig2](c ) and [ fig3](c ) show that for all the moments , the moments calculated using the high - wave set have better agreement with data than do moments calculated using the low - wave set .
the pwa results presented in this section were for data with @xmath89 ( gev/@xmath3)@xmath90 referred to as @xmath91 in table [ tbins ] .
the analysis was also carried out in 11 other bins in @xmath92 some results of which will also be summarized below . figure [ nc_all_bkg ] shows the sum of all waves in the high - wave set along with the background wave for both the charged and neutral modes . the fitting procedure constrains the sum of all the waves to agree with the acceptance - corrected @xmath17 mass distribution .
the background wave ( uniform in all angles ) is added incoherently with the other waves . as can be seen ,
the contribution from the background wave is small ( of order 1% ) .
there is a systematic disagreement in the relative normalization between the two modes resulting from an overall per track inefficiency of 5% in the trigger pwc s .
the overall inefficiency has no bearing on the features of the acceptances and was thus not incorporated into the monte carlo corrections , but results in the neutral mode having a yield consistently 25% higher than the charged mode .
the neutral mode required only one forward - going charged particle in the trigger while the charged mode required three . in the following we discuss intensities and phases of individual partial waves .
a partial wave intensity corresponds to a single term of equation ( [ int ] ) which is a product of the fitted partial wave amplitude , @xmath93 , and the isobar amplitude @xmath94 .
similar to the procedure followed in references @xcite , the intensities are integrated over acceptance corrected data , and are normalized to the total number of events .
the coherent sum of the @xmath95 @xmath96-waves is shown in figure [ coherent1pp ] for the @xmath47 and @xmath46 modes .
for each the results from the high wave set and low wave set are shown .
the @xmath95 wave is a dominant wave and the results of the pwa for the two wave sets ( high and low ) are in good agreement .
the yield for the two charged modes is also consistent with expectations from isospin for the @xmath96 system . the @xmath95 @xmath96 @xmath97-wave is not included in the high - wave set , since it did not meet criteria for inclusion , but it is included in the low - wave set . the observed ratio of @xmath95 @xmath96 @xmath28-wave to @xmath97-wave in the low - wave set is consistent with that of reference @xcite .
figure [ 2pp ] shows the @xmath98-wave @xmath96 partial wave as a function of @xmath99 for both the neutral and charged modes . a clear @xmath5 signal is observed . for a pure @xmath96 isovector resonance
, we expect an equal number of charged mode and neutral mode events .
this wave is clean and robust and will be used for interferometry to study the phase motion of other waves relative to this wave .
figures [ 2mps ] and [ 2mpd ] show the intensities for the @xmath100-wave @xmath101 and @xmath102-wave @xmath101 waves respectively for both the charged and neutral modes .
note that the relative yields between charged and neutral modes , for these @xmath101 decays are consistent with expectations from isospin , after taking into account the overall 25% reduction in yield for the charged mode .
these intensities are also compared with those measured by daum _ et .
@xcite in a similar amplitude analysis of diffractive @xmath17 production in @xmath14 interactions at 63 and 94 gev/@xmath10 .
the daum results have been arbitrarily scaled by a factor of 5 .
the @xmath28 and @xmath97 @xmath101 modes both show a resonant - like structure but not at the same peak position .
the @xmath28 wave peaks at approximately 1.6 gev/@xmath10 while the @xmath97 wave peaks at approximately 1.8 gev/@xmath10 .
this shift in the peak position has been observed in other analyses @xcite .
this has been interpreted as the @xmath101 decay mode of two different resonances .
the pdg places the @xmath6 @xmath103 resonance at a mass of 1.67 gev/@xmath10 based on fits to the @xmath28 wave mode alone .
figure [ a4 ] shows the @xmath104-wave @xmath96 intensity as a function of @xmath99 for both the neutral and charged modes . comparing figures
[ a4 ] and [ 2pp ] the yield of @xmath7 is about 3% that of the @xmath5 indicating the sensitivity of the pwa within the isobar model for finding states with relatively low production cross - sections .
the @xmath105-wave @xmath96 partial wave and its interference with the @xmath106 wave are shown in parts ( a ) and ( b ) of figures [ exotic_n ] and [ exotic_c ] for the neutral mode ( @xmath46 ) and charged mode ( @xmath47 ) . when the low - wave set is used in the fit ,
an enhancement is observed for the exotic @xmath82 wave in the @xmath17 mass region around 1.6 gev/@xmath10 , consistent with the observation of @xcite .
the enhancement disappears when the high wave set is used .
this effect is observed for both @xmath17 channels .
the phase of the exotic wave relative to the dominant @xmath106 wave ( see figures [ exotic_n](b ) and [ exotic_c](b ) ) is similar for both wave sets and dominated by the phase of the @xmath5 .
the apparent phase motion beginning around 1700 mev/@xmath10 is likely due to the @xmath107 @xcite , but the unreliability of the exotic intensity makes this an inappropriate place for its study .
one source of leakage into the exotic wave for the pwa carried out with the low - wave set is the @xmath6 . in figure
[ leaks ] the coherent sum of the intensities for the @xmath108-wave @xmath96 , @xmath109-wave @xmath96 and @xmath110-wave @xmath96 waves for the charged and neutral modes are shown .
these waves are allowed decay modes of the @xmath6 . a clear peak is seen at the @xmath6 mass , the same mass region where the purported exotic meson was seen using the low - wave set .
the three waves shown in figure [ leaks ] were not included in the low - wave set .
the pwa fit was repeated for the high wave set but with the @xmath108-wave @xmath96 , @xmath109-wave @xmath96 and @xmath110-wave @xmath96 waves removed .
the resulting @xmath111-wave @xmath96 intensities for the neutral mode for this modified high wave set fit , along with the unmodified high wave set and low wave set , are shown in figures [ neut_mod ] .
the corresponding results for the charged mode are shown in figure [ chrgd_mod ] .
the agreement between the modified high wave set and the low wave set shows that the @xmath112-wave @xmath96 decays and the @xmath110-wave @xmath96 decay of the @xmath113 are the source of leakage leading to the observed exotic peak in the low wave set .
a similar phenomenon is observed in the @xmath106 wave in the low wave set where the @xmath104-wave @xmath96 is not included . as discussed above
, the @xmath104-wave @xmath96 in the high wave set clearly shows the @xmath7 .
figure [ leakage_2_4 ] shows a semi - log plot of the the @xmath106 intensity for charged @xmath17 mode for the high wave set ( filled circles ) and low wave set ( open circles ) . the @xmath114 intensity for the high wave set
is also shown ( filled triangles ) .
the @xmath5 is dominant in the @xmath106 intensity in both high and low wave sets .
the @xmath106 intensity for the low wave set shows a resonance - like structure near 2 gev/@xmath10 in low wave set but not the high wave set .
however the @xmath106 low wave set intensity agrees well with the @xmath114 high wave set intensity . in this case , leaving out the @xmath114 in the low wave set forces a enhancement at @xmath115 gev/@xmath10 in the @xmath106 wave .
the contribution to the @xmath116 of the positive reflectivity exotic partial wave was examined by removing this wave from the high wave set and comparing the resulting likelihood with that of the original high wave pwa .
figure [ exot_dlnl ] shows the resulting change in likelihood .
the sign convention of this difference is such that positive ( negative ) values indicate inclusion of this partial wave improved ( degraded ) the fit quality .
as can be seen from figure [ exot_dlnl ] there is no region of @xmath117 effective mass where this partial wave is consistently required by the fit .
this behavior is seen in all t - bins presented here .
we examined the stability of the exotic wave intensity and the intensity of other waves across the 12 @xmath118-bins shown in table [ tbins ] for the high wave fit in the charged @xmath17 .
as noted in table [ tbins ] , the width of the @xmath92 bin for the first five bins is 0.02 ( gev/@xmath3)@xmath90 and 0.05 ( gev/@xmath3)@xmath90 for the other bins
. figures [ exoticw ] , [ pi2 t ] and [ phasex ] show the variation of the @xmath105-wave @xmath96 intensity , the @xmath100-wave @xmath101 intensity and the phase difference between the @xmath82 and @xmath103 amplitudes for the charged @xmath17 mode for the 12 @xmath118-bins .
the @xmath100-wave @xmath101 intensity is stable across the 12 @xmath118 bins .
we note that although the @xmath105-wave @xmath96 intensity may indicate an enhancement in the vicinity of @xmath119 gev/@xmath10 for some of the higher @xmath118 bins in figure [ exoticw ] , the corresponding phase difference shown in figure [ phasex ] is not consistent with resonant behavior .
indeed the @xmath100-wave @xmath101 and @xmath105-wave @xmath96 are phase - locked in the 1.6 gev/@xmath10 region .
one possible explanation is production of an exotic resonance with the same mass , width and phase motion as the @xmath6 .
another is some residual leakage of the @xmath6 in the exotic wave in the high - wave set .
the @xmath98-wave @xmath96 intensity ( shown in figure [ plot_a2 ] ) and the @xmath104-wave @xmath96 intensity ( shown in figure [ plot_a4 ] ) are shown for the 12 @xmath118 bins . for completeness , in figure [ dsdt_a2 ] , we compare the @xmath118 dependence of the @xmath98-wave @xmath96 intensity , dominated by the @xmath5 , in both @xmath17 modes in this experiment , with other measurements of @xmath120 at 18.8 gev/@xmath3@xcite .
studies were carried out to determine the effect on the pwa of varying the parameters that define the selection criteria as described in section [ sec : data ] .
different criteria were found to effect the results in different ways but no significant change in our conclusions was observed .
the effects of these perturbations on the analysis are described individually below .
the beam hole cut is nominally @xmath121 .
values of @xmath122 and @xmath123 systematically increased and decreased the intensity of the @xmath124 @xmath28-wave below 1.2 gev/@xmath10 but left the mass region above 1.2 gev@xmath125 indistinguishable from the unperturbed analysis .
the changes observed were approximately 5% of the intensity .
conclusions regarding the low mass behavior of this partial wave should take into account this systematic effect .
other partial waves were unaffected .
the cut to remove the @xmath126 in the charged sample affected the data only above 2.0 gev/@xmath10 . removing this cut from the analysis increased the intensity of the @xmath127 @xmath28-wave by approximately twice its error between 2.0 and 2.4 gev/@xmath128 .
the systematic effect on other partial waves is insignificant when compared with their statistical errors .
the effect of removing the dea cut was similar to using a less restrictive value for the beam hole cut .
small , but statistically significant , increases in the intensity of the @xmath127 @xmath28-wave were observed between 1.2 and 1.6 gev/@xmath128 .
increasing ( decreasing ) the confidence level cut to 0.3 ( 0.1 ) changed the observed intensities everywhere by a mass independent scale factor .
the same effect was observed for the csi cut . removing this cut increased all intensities by a factor of 1.66 but left their shapes unchanged . in summary
, we have investigated the effect of perturbing the selection criteria summarized in tables [ cutsn ] and [ cutsc ] on the results of the partial wave fit and have shown that our results do not depend on the specific values used in the criteria .
we used the isobar model , following the same methodology employed in references @xcite .
this model uses two - body amplitudes for the @xmath29 system .
the @xmath129 isovector and @xmath130 isoscalar amplitudes were parametrized using breit - wigner amplitudes .
there are various @xmath131 , isoscalar parametrizations available in the literature . in general
the @xmath29 amplitude is dominated by a broad @xmath68 meson , the @xmath26 and several resonances above @xmath132 , depending on the specific parametrization .
the results presented in this paper were all based on the parametrization used in references @xcite .
this is based on the model of @xcite with the possibility of direct @xmath26 production ; the @xmath26 was subtracted from the elastic @xmath29 amplitude and added as a breit - wigner resonance .
we have also studied other parametrizations , _
e.g. _ based on ( 1 ) @xmath133 dispersion relations , ( 2 ) explicit inclusions of the @xmath134 channel and/or ( 3 ) including direct production of scalar resonances above @xmath135 @xcite .
all give qualitatively identical results .
a partial wave analysis of the @xmath136 and @xmath137 systems has been presented using a data set whose statistics exceeds published analyses of the @xmath17 systems .
the @xmath4 , @xmath5 , @xmath6 and @xmath7 mesons were observed .
the data provide no evidence for an exotic @xmath8 meson as had been reported earlier using a smaller sample of the @xmath136 mode alone @xcite .
this analysis included the study of criteria for defining a set of partial waves sufficient to describe the data including a study of the effect on the fit quality by removing individual waves and also a comparison of observed moments of the data with moments computed from pwa solutions .
the wave set thus established contained more waves than were used in the analysis reporting evidence for the exotic meson .
our studies indicate that leaving out partial waves corresponding to decay modes of the @xmath6 lead to a false enhancement in the exotic @xmath82 wave .
this analysis and the analysis of references @xcite were based on the isobar model .
the isobar model is known to suffer some limitations , as has been pointed by others @xcite . the so - called _ deck mechanism _
@xcite is particularly relevant to the production of the @xmath17 system and is being applied to these data .
results will be presented in a forthcoming paper .
the authors thank jozef dudek , geoffrey fox and matt shepherd for enlightening discussions and for their careful read and critique of this paper .
this work was supported by grants from the united states department of energy ( de - fg-91er40661/task d and de - fg0287er40365 ) , the national science foundation ( eia-0116050 ) for avid , the russian ministry of science and technology and russian ministry of science and education .
the authors of this paper have made available technical notes and pwa results online @xcite .
the notes include ( 1 ) details of the data selection ; ( 2 ) a comparison of the software used in this analysis with that used in the analysis of reference @xcite ; ( 3 ) the monte carlo details for this analysis ; ( 4 ) a full description of the pwa formalism ; ( 5 ) the moments method technique and definition of angles ; ( 6 ) a study of background contributions to partial wave intensities ; ( 7 ) stability of pwa results against variations in data selection criteria ; ( 8) acceptance and resolution ; ( 9 ) the determination of the minimal partial wave set and ( 10 ) @xmath28-wave parametrization .
in addition , pwa results ( amplitudes and phases ) are available for 12 bins in @xmath118 and for various wave sets .
a. r. dzierba _ et .
the authors of this paper have posted notes on the web regarding the 3@xmath23 data processing and selection , pwa formalism , software and other details .
url : dustbunny.physics.indiana.edu/3pi_paper ( 2005 ) . | a partial wave analysis ( pwa ) of the @xmath0 and @xmath1 systems produced in the reaction @xmath2 at 18 gev/@xmath3 was carried out using an _ isobar _ model assumption .
this analysis is based on 3.0 m @xmath1 events and 2.6 m @xmath0 events and shows production of the @xmath4 , @xmath5 , @xmath6 and @xmath7 resonances .
results of detailed studies of the stability of partial wave fits are presented .
an earlier analysis of 250k @xmath0 events from the same experiment showed possible evidence for a @xmath8 exotic meson with a mass of @xmath91.6 gev/@xmath10 decaying into @xmath11 . in this analysis of a higher statistics sample of the @xmath12 system in two charged modes
we find no evidence of an exotic meson . |
the experimental success in creating ultracold @xmath0k@xmath1rb molecular gas near quantum degeneracy @xcite has drawn considerable attention to in studying the fundamental properties of degenerate dipolar fermi gases . within the framework of the semi - classical theory ,
the ground state properties , the collective excitation , and the free expansion dynamics of a normal state dipolar fermi gas were studied theoretically @xcite .
a recent theoretical work based on a variational approach reveals that , due to the fock exchange interaction , the momentum distribution is stretched along the direction of dipole moment such that the fermi surface becomes an ellipsoid @xcite .
this result was confirmed numerically for both homogeneous @xcite and trapped @xcite systems .
taking into account the effect of the exchange interaction , further theoretical work regarding the normal state of the zero temperature dipolar fermi gas includes studying the free expansion @xcite , collective excitation @xcite , zero sound @xcite , and the fermi liquid properties @xcite .
another interesting feature of the dipole - dipole interaction is that the partially attractive dipolar force is responsible for the formation of anisotropic bcs pairing @xcite . with the recent experimental development on control the hyperfine states of @xmath0k@xmath1rb molecules @xcite , the bcs pairings in a mixture of fermionic polar molecules with two different hyperfine states
are also studied theoretically @xcite . in particular , the effects of the fock exchange interaction to pairing was considered in ref .
@xcite . in this paper
, we extend our previous work on the ground state properties of the dipolar fermi gases to finite temperature case . employing the semi - classical theory
, we numerically obtain the phase - space distribution function through a self - consistent procedure .
we show that the deformations of the distribution function in both momentum and real space become smaller and smaller as the temperature is increased . for homogeneous system
, we also calculate the thermodynamic quantities such as pressure , entropy , and heat capacity .
in particular , at low temperature limit and in weak interaction regime , we derive an analytic expression for the entropy , which agrees qualitatively with our numerical result . for the trapped gases ,
we also explore the temperature dependence of the stability .
the remainder of this paper is organized as follows . in sec .
[ theory ] , we introduce our model and briefly outline the semi - classical theory for a dipolar fermi gas at finite temperature .
the numerical and analytical results are presented in sec .
finally , we conclude in sec .
[ concl ] .
here we consider a system of @xmath2 spin polarized dipolar fermions interacting via dipole - dipole interaction @xmath3 where @xmath4 with @xmath5 being the electric dipole moment . for simplicity
, we have assumed that the dipole moments of all fermions are orientated along @xmath6-axis .
additionally , we shall consider both homogeneous and trapped systems . in the former case
, @xmath7 can be regarded as a box of volume @xmath8 ; while for the latter , the trap is assumed to be a harmonic potential with axial symmetry , i.e. , @xmath9 where @xmath10 is the geometric average of the trap frequencies and @xmath11 is trap aspect ratio . within the framework of the semiclassical theory ,
the thermodynamic properties are completely characterized by the phase - space distribution function @xmath12 , which , at finite temperature @xmath13 , satisfies the fermi - dirac statistics @xmath14 where @xmath15 is the chemical potential introduced to fix the total number of particles such that @xmath16 the dispersion relation of the quasi - particle takes the form @xmath17 where @xmath18 is the fourier transform of @xmath19 and the last two terms represent the mean - field potentials originating from hartree direct and fock exchange interactions , respectively .
we remark that the local density approximation has been employed to obtain eq .
( [ dist ] ) for the trapped system . and @xmath20 for @xmath21 .
the corresponding temperatures are @xmath22 ( solid lines ) , @xmath23 ( dashed lines ) , @xmath24 ( dash - dotted lines ) , and @xmath25 ( dotted line ) .
( c ) the temperature dependence of @xmath26 for various dipolar interaction strengths .
the horizontal lines denote the values of @xmath26 at zero temperature.,width=307 ] equations ( [ dist])-([disp ] ) form a closed system of equations which can be solved numerically through an iterative procedure to obtain the phase - space distribution function @xmath12 . within semi - classical approximation
, the total energy at the finite temperature takes the same form as that at zero temperature , i.e. , @xmath27f(\mathbf{r},\mathbf{k})\nonumber\\ & & \!\!+\frac{1}{2(2\pi)^6}\int d{\mathbf r}d{\mathbf k } d{\mathbf r}'d{\mathbf k}'f(\mathbf{r},\mathbf{k } ) f(\mathbf{r}',\mathbf{k}')v_{d}(\mathbf{r}-\mathbf{r}')\nonumber\\ & & \!\!-\frac{1}{2(2\pi)^6}\int d{\mathbf r}d{\mathbf k } d{\mathbf k } ' f(\mathbf{r},\mathbf{k})f(\mathbf{r},\mathbf{k}')\widetilde v_{d}({\mathbf k}-{\mathbf k}'),\nonumber\end{aligned}\ ] ] where the first line represents the kinetic ( @xmath28 ) and potential ( @xmath29 ) energies , and last two lines are , respectively , the direct ( @xmath30 ) and exchange ( @xmath31 ) interaction energies .
finally , we point out that , for axially symmetric system , the phase - space distribution function reduces to @xmath32 , where @xmath33 and @xmath34 .
this fact can be used to simplify the numerical integration @xcite .
in this section , we present our results on the thermodynamic properties of a dipolar fermi gas . for a homogeneous gas ,
we first obtain the phase - space distribution function numerically , which allows us to calculate other thermodynamic quantities , such as pressure , entropy , and heat capacity . at low temperature limit and in weak interaction regime
, we also derive an analytic expression for the entropy , which qualitatively agrees with our numerical result . in the second part of this section
, we present the numerical results on the thermodynamics of the trapped systems .
( a ) , @xmath31 ( b ) , and @xmath15 ( c ) .
the dimensionless dipolar interaction strength are @xmath35 ( dash - dotted lines ) , @xmath36 ( dashed lines ) , and @xmath37 ( solid lines).,width=259 ] for a homogeneous gas , the phase - space distribution function reduces to a function of momentum , i.e. , @xmath38 . to present our results , it is convenient to introduce a dimensionless dipolar interaction strength @xmath39 where @xmath40 is the number density of the system and @xmath41 with @xmath42 .
we note that @xmath43 and @xmath44 are , respectively , the fermi wavevector and fermi energy of an ideal fermi gas .
we plot the phase - space distribution function @xmath45 in fig .
[ hdist ] ( a ) and ( b ) for various temperatures and @xmath46 . at low temperature ,
the momentum distribution is clearly stretched along @xmath6-axis . however , as one increases the temperature , the momentum distribution becomes less anisotropic .
such behavior can be most easily visualized by calculating the aspect ratio of the cloud in momentum space @xmath47 .
figure [ hdist ] ( c ) shows the temperature dependence of @xmath26 corresponding to various dipolar interaction strengths . for small @xmath13 , @xmath26 approaches to the value at zero temperature ; as one increases the temperature , @xmath26 goes to unit asymptotically .
( dash - dotted lines ) , @xmath36 ( dashed lines ) , and @xmath37 ( solid lines).,width=259 ] due to the spatial homogeneity of the system , the direct dipolar interaction energy vanishes .
therefore , the total energy only contains the contributions from kinetic energy and fock exchange interaction . in fig .
[ henergy ] ( a ) and ( b ) , we plot , respectively , @xmath28 and @xmath31 as functions of the temperature .
the kinetic energy corresponding to stronger dipolar interaction is larger than that with weaker dipolar interaction , since the stronger dipolar interaction results in larger momentum space deformation . as to the exchange interaction energy
, we see that @xmath31 vanishes as @xmath26 approaches @xmath48 at high temperature limit .
we also plot the temperature dependence of the chemical potential in fig .
[ henergy ] ( c ) .
clearly , @xmath15 decreases as one increases @xmath13 , in analogy to the ideal fermi gas .
in addition , the chemical potential is also a decreasing function of @xmath49 .
this can be most easily understood at zero temperature limit , where the chemical potential becomes @xcite @xmath50 in weak interaction regime ( @xmath51 ) .
equation ( [ chem ] ) indicates that , for given density @xmath52 , the chemical potential decreases as the dipolar interaction strength grows . to calculate other thermodynamic quantities , we work with the thermodynamic potential @xmath53 where @xmath54\ln[1-f({\mathbf k})]\big\}.\nonumber\end{aligned}\ ] ] is the entropy of the system .
the pressure directly relates to the thermodynamics potential as @xmath55 and the heat capacity can also be evaluated using the definition @xmath56 . in fig .
[ hpcv ] , we present the temperature dependence of @xmath57 , @xmath58 , and @xmath59 corresponding to different dipolar interaction strengths . as a comparison , we also plot the results for an ideal fermi gas
. in general , the results for dipolar fermi gases approach to those corresponding to the ideal gas at high temperature limit .
figure [ hpcv ] ( a ) also indicates that , at a given temperature , @xmath57 is a decreasing function of @xmath49 , in agreement with the zero temperature result @xcite .
this can be understood since the overall dipolar interaction is attractive for homogeneous gas .
in addition , even though @xmath58 vanishes at zero temperature , the entropy for given @xmath60 also exhibits a similar dependence on @xmath49 as the pressure [ fig .
[ hpcv ] ( b ) ] .
furthermore , the temperature and interaction dependence of the heat capacity , shown in fig . [ hpcv ] ( c ) , can be naturally deduced from the behavior of th entropy . to gain more insight into the entropy
, we shall derive an analytical expression for @xmath58 at low temperature limit . to this end
, we note that when @xmath61 , the derivative of @xmath58 with respect to @xmath13 at constant @xmath62 and @xmath15 can be approximately expressed as @xcite @xmath63 where we have explicitly expressed @xmath64 as a function of the temperature @xmath13 and the zero temperature dispersion relation for the quasi - particle is @xcite @xmath65 with @xmath66 is the second order legendre function , @xmath67 is the polar angle of @xmath68 , and @xmath69 following the same procedure as that in ref .
@xcite , we find @xmath70 where @xmath71_{\varepsilon_0(k,\theta ) = \mu}\sin\theta d\theta.\nonumber\end{aligned}\ ] ] we note that eq .
( [ entropy ] ) is not the entropy in usual sense as it is a function of @xmath15 . to proceed further , we change variable from @xmath15 to @xmath2 by using eq .
( [ chem ] ) , which yields , at low temperature limit and in weak interaction regime , @xmath72 this expression clearly indicates that the entropy is decreasing function of @xmath49 , in qualitative agreement with our numerical results .
now we turn to study the trapped case , for which we shall adopt a set of dimensionless units based on the harmonic oscillator length @xmath73 : @xmath74 for length , @xmath75 for wave vector , and @xmath76 for energy . under these choices
, we may define a dimensionless dipolar interaction strength @xmath77 in addition , the normalization condition for phase - space distribution function becomes @xmath78 .
of a trapped gas in spherical potential for @xmath79 ( a ) , @xmath23 ( b ) , and @xmath48 ( c ) .
the dimensionless dipolar interaction strength is @xmath80.,width=326 ] ( a ) and @xmath81 ( b ) for @xmath82 ( solid lines ) , @xmath48 ( dashed lines ) , and @xmath83 ( dash - dotted lines ) .
the dimensionless dipolar interaction strength are @xmath80.,width=259 ] in fig .
[ tdist ] , we present the typical results for phase - space distribution function @xmath84 for a gas trapped in a spherical potential . similar to the momentum space distribution for a homogeneous gas ( fig .
[ hdist ] ) , the distribution function in real space becomes elongated along @xmath6-axis at low temperature .
however , the deformation becomes smaller and smaller as the temperature is increased . to characterize the real space deformation
, we define the deformation parameter in real space as @xmath85 , where the inclusion of the factor @xmath86 is to eliminate the deformation caused by trapping potential such that @xmath87 for a noninteracting gas .
figure [ tdefor ] ( a ) and ( b ) plot , respectively , the temperature dependence of the deformation parameters in momentum and real spaces .
as expected , stronger dipolar interaction induces larger deformations in momentum and real spaces .
however , both @xmath26 and @xmath81 approach unit at high temperature limit .
finally , we explore the stability of the trapped gases at finite temperature .
similar to the zero temperature case @xcite , due to the partially attractive feature of the dipolar interaction , there also exists a critical dipolar interaction strength @xmath88 for a given temperature @xmath13 , beyond which the system becomes unstable . in fig .
[ tcri ] , we plot the temperature dependence of the critical dipolar interaction strength for various trap geometries . it can be seen that , independent of the trap aspect ratio @xmath11 , @xmath88 is an increasing function of @xmath13 .
this can be understood as higher temperature corresponds to smaller momentum and real space deformations , resulting in smaller dipolar interaction .
in addition , higher temperature also yields higher kinetic energy which stabilizes the system .
to conclude , based on the semi - classical theory , we have studied the thermodynamics of a dipolar fermi gas by numerically finding the phase - space distribution function .
we show that the deformations in both momentum and real space becomes smaller and smaller as one increases the temperature . for homogeneous gases ,
we also calculate pressure , entropy , and heat capacity .
in particular , at low temperature limit and in weak interaction regime , we obtain an analytic expression for the entropy , which agrees qualitatively with our numerical result . the stability diagram for a trapped gas at finite temperature
is also obtained . | based on the semiclassical theory , we investigate the thermodynamic properties of a dipolar fermi gas . through a self - consistent procedure ,
we numerically obtain the phase - space distribution function at finite temperature .
we show that the deformations in both momentum and real space become smaller and smaller as the temperature is increased . for the homogeneous case
, we also calculate pressure , entropy , and heat capacity .
in particular , at the low - temperature limit and in the weak interaction regime , we obtain an analytic expression for the entropy which agrees qualitatively with our numerical result .
the stability of a trapped gas at finite temperature is also explored . |
the origin of extra solar cosmic rays ( cr ) is one of the important unresolved astrophysical questions .
galactic shell type supernova remnants ( snr ) have been proposed as plausible acceleration sites for cosmic rays up to energies of several pev ( @xcite ) and - for very massive sn progenitors - to even higher energies ( @xcite ) .
recently direct experimental evidence for electron acceleration in these objects has been found in the x - ray ( @xcite ; @xcite ; @xcite ) and tev @xmath8-ray range ( @xcite ) . somewhat surprisingly , similar searches for evidence of hadron acceleration have only yielded upper limits on the expected @xmath8-ray emission from the interaction of the hadrons with interstellar matter up to now ( @xcite ; @xcite ; @xcite ) . +
an indirect approach to distinguish between different theoretical models aiming to describe the acceleration of charged cosmic rays ( cr ) is to measure the energy spectrum and composition of cr and compare the results with model predictions . here
the energy regime around the so called `` knee '' between 1 and 10pev is especially interesting(@xcite ) . in this energy range the all - particle cr energy spectral slope - that is constant within measurement errors for lower energies - suddenly increases .
the riddle of the origin of the knee and of the cosmic radiation with energies exceeding it , is not yet finally resolved .
the following general solutions have been discussed : + 1 .
the change in index is due to some propagation effect in an `` original '' cosmic - ray population that displays an unbroken power law from low energies up to energies above the knee .
the most popular idea is that the energy dependence of the diffusion constant of cosmic rays in the galaxy could change in the knee region ( @xcite ; @xcite ) . because of the dependence of the diffusion constant on the nuclear charge z a modest decrease in the fraction of `` light '' elements ( hydrogen and helium )
would be expected . in the simplified chemical model we use below ( heavy elements modelled by 65 @xmath9 oxygen and 35 @xmath9 iron , light elements by 40 @xmath9 hydrogen and 60 @xmath9 helium ) the fraction of light elements
would be expected to decrease from an assumed value of 60 % below the knee to 43 % above the knee .
in such a model ( barring a special cancellation of effects ) the knee is expected to be a relatively smooth feature , extending over about a decade in energy .
a principal problem with this approach is that no plausible galactic source of cosmic rays has been identified which is quantitatively capable of producing the `` original '' cosmic - ray population .
the knee signals in some way the maximum energy for the sources responsible for low energy cosmic rays .
the cosmic rays at higher energies could be `` re - accelerated '' low energy cosmic - rays , e.g. at the shock front of a galactic wind ( jokipii @xmath10 morfill 1987 ) or an ensemble of shock fronts in clusters of massive stars ( @xcite ) . in this case
a phenomenology similar to the diffusion model in the previous paragraph would be expected .
alternatively , above the knee a completely new population of cosmic rays dominates . in this case typically
dramatic changes in chemical composition are expected , e.g. to pure hydrogen in the extragalactical model of protheroe ( @xcite ) and nearly pure heavy elements ( fraction of light elements @xmath11 0.3 far above the knee ) in a model with special snrs by stanev et al .
( @xcite ) .
the special properties of these new sources could in principle allow to understand a knee relatively `` sharp '' in energy .
+ to definitely discriminate between a composition changing as expected in models with an energy dependent diffusion constant ( discussed above under 1 . ) and an unchanging composition , it is necessary to achieve an error of smaller than @xmath4 10 @xmath9 in the experimental determination of the fraction of light elements in the total cosmic radiation above the knee . + while the cosmic - ray composition and energy spectrum are well known from direct balloon and space - borne observations up to energies of about 100tev , no general agreement has been reached at higher energies ( @xcite ) .
the results obtained for cr around the `` knee '' suffer seriously from the fact that due to the low flux of cr above 1pev , only large ground based installations observing the extensive air showers ( eas ) induced by cosmic rays in the atmosphere can provide experimental data
. however the sensitivity of eas observables to the mass of the primary cr is weak .
the analyses are rendered even more difficult due to theoretical uncertainties concerning the high energy interactions in the atmosphere ( @xcite ; @xcite ) .
+ here we present an analysis of eas between 300tev and 10pev which restricts to observables related to the electromagnetic shower component . in the following sections the experimental setup ( section [ sec_exps ] ) , the monte - carlo simulations ( section [ sec_mc ] ) , the event reconstruction ( section [ sec_reco ] ) and analysis methods ( section [ sec_anal ] ) are described .
sections6 presents the results concerning the cr energy spectrum and the coarse mass composition .
a more detailed study of the systematic errors and a discussion of methods to analyse the composition without relying on the absolute penetration depth are discussed in section 7 .
the paper ends with conclusions in section8 .
the air - shower detector complex hegra covers an area of 180@xmath12180m@xmath13 at a height of 2200 m a.s.l .
( 790g/@xmath5 ) ( @xcite ; @xcite ; @xcite ; @xcite ) . in the present analysis
only data of the scintillator array and part of the airobicc array were used .
the former consists of 243 huts with plastic scintillators of an area of 0.96 m@xmath13 , covered with 5 mm of lead on a grid with 15 m spacing ( with a denser part with 10 m spacing in the centre of the array ) .
the part of the latter used in this analysis is formed by 49 open photo - multipliers fitted with winston cones , restricting the viewed solid angle to 0.835 sr and measuring the air cherenkov light of eas on a grid with 30 m spacing .
the gain nonlinearity of all components in the cherenkov - light measurement was carefully checked , both with a led light source with variable light intensity and a direct charge source .
while the used cherenkov light photomultiplier tubes were found to be linear , the used amplifier showed an antilinearity ( gain rises with input amplitude ) which was corrected in the data analysis . above 10 pev
the amplifier begin to show signs of saturation and therefore no data above 10 pev are included in the present analysis .
the trigger conditions used for the data analysed here demand a signal from at least 14 scintillator or 6 airobicc stations within 150ns .
this corresponds to an energy threshold for primary protons and iron nuclei of 25 and 80tev respectively .
eas events were simulated using the corsika code in its versions 5.20 with the qgsjet / gheisha options ( @xcite ; @xcite ) ) .
this generator is based on the quark - gluon string model ( qgs ) with an allowance for semihard processes ( jet ) ( @xcite ) .
complex nuclei were treated with the `` complete fragmentation '' ansatz . the energy cutoff for particles of the electromagnetic cascade
was set to 3mev .
proton , helium , oxygen and iron induced showers were produced at zenith angles of 0@xmath14,6@xmath14 , 12@xmath14 and 18@xmath14 at discrete energies between 300tev and 10pev ( 4400 independent showers of which 1000 are above 1 pev ) as well as an independent sample with a continuous energy distribution at zenith angles of 6@xmath14 and 12@xmath14 between 50 tev and 1pev ( 4330 independent showers ) and following a power law of e@xmath15 between 2.5 pev and 6.5 pev ( 240 showers ) .
the events continuously distributed in energy were spectrally weighted and used in the fits to infer the chemical composition ( section [ chempos ] ) , while the samples with discrete energies were employed to develop the reconstruction methods and to correct the results obtained with biased estimators of the primary energy ( section [ erecmethods ] ) .
note that the simulation of an eas induced by a 1pev primary proton including the cherenkov light production requires about 3h cpu time on a 300mhz pentium - ii pc ( during the same time the hegra experiment registers more than 350 showers with energies larger than 1pev ) .
+ the detector performance was modelled with two independent detector simulations : a full detector simulation ( @xcite ) , and an empirical simulation using measured response functions ( @xcite ; @xcite ) .
each independent generated eas was used 20 times with core positions inside and outside the hegra area to take into account the detector related fluctuations of observables and to check the event selection criteria . with the standard cuts described below , each shower was used on the average two times in the discrete and once in the continuous mc sample .
+ special care has been taken to simulate the density profile and absorption features of the atmosphere above lapalma correctly .
weather balloon measurements as well as comparisons between tev photon data registered by the hegra system of imaging air cherenkov telescopes and simulations were employed for this purpose ( @xcite;@xcite ) .
the shower development and light emission were modelled with the u.s .
standard atmosphere , and the light propagation was then simulated with a special program elba ( @xcite ) , assuming a tropical maritime atmosphere for the summer .
this atmosphere is a good approximation for the conditions at tenerife , an island neighbouring the experimental site ( @xcite ) .
the core position of an eas is reconstructed independently from the data of the scintillator matrix and from airobicc where the latter data allow to tag core positions beyond the hegra boundary .
if the core position lies inside the area covered with detector elements the scintillator derived core coordinates have a resolution of @xmath16 for protons ( iron ) at energies above 300tev ( a little more accurate compared to airobicc mainly due to the smaller grid distances of the scintillator huts ) .
= 8.8 cm the direction of the primary particle is reconstructed nearly independently from the scintillator and airobicc arrival time data ( where the scintillator derived core position is used here ) .
+ the particle density measured by the scintillator array is fitted by the nkg formula ( @xcite ) with a molire radius of 106 m , yielding the shower size @xmath17 and an _ age _ value .
@xmath17 is a factor @xmath181.6 larger than @xmath19 ( the `` true '' shower size at detector level , denoting the number of charged particles above a kinetic energy of 3mev ) due to coverage of the detector huts with a lead layer and the fact that the nkg function does not correctly parametrise the electromagnetic part of hadronic showers .
+ the dependence of the cherenkov photon density @xmath20 , as measured by airobicc , on the distance r from the shower axis can be well described by an exponential in the region 20 m @xmath21r@xmath11100 m ( @xcite ) : @xmath22 the parameter _ slope _ ( in units of [ 1/m ] ) is the most important one in our analysis methods . as an illustration fig.[showerpic ] shows the lateral charged - particle and cherenkov - light distributions for a single shower .
+ the amplitude calibration of the scintillator array is done for samples of 50000 events by comparing the adc spectra of the individual huts - which display a single peak essentially corresponding to the energy deposited by minimum ionising electrons and muons - with the result of mc simulations for identical conditions .
the absolute amount of the air cherenkov light registered by airobicc was calibrated by comparing the energy inferred from the lateral cherenkov light density in the spectral range from 300 nm to 500 nm registered at a shower core distance of 90 m ( referred to as @xmath23 in the following ) and the energy derived from @xmath17 and _ slope _ in the interval @xmath24 ( refer to section [ erecmethods ] below for energy reconstruction methods ) .
the absolute cherenkov - light calibration thus depends on the cr mass composition , because we do not apply a primary - mass independent energy reconstruction here .
we used the low - energy composition at 100 tev as specified by wiebel - sooth et al.(@xcite ) ( 60 @xmath9 light elements , see below section [ chempos ] for details ) for this calibration .
if a pure proton ( iron ) composition is assumed the energy reconstructed from cherenkov light alone is shifted by 3 ( 13)% to higher ( lower ) energies .
+ to select well - reconstructed events the eas core positions and directions as reconstructed with airobicc and the scintillator matrix are demanded to be consistent .
additional cuts ensure the quality of the directional as well as the fits to the lateral particle and cherenkov light density distributions . events with the true shower - core position within the hegra array boundaries for the detector components used in this analysis ( distance to edge of array @xmath25 10 m ) and a zenith angle below 15@xmath26 are used for the further analysis .
the efficiency to select eas events with true core positions in the regarded 160@xmath12160m@xmath13 area is about 98% for primary energies above 300tev ( independent of the primary mass ) . the contamination of the sample with eas , where the true shower cores lay beyond the hegra boundary but which were erroneously reconstructed to fulfil the cuts is less than 1% from our simulations .
+ nights with perfect weather conditions are selected by data of the carlsberg meridian cycle ( @xcite ) and by comparing the cherenkov light measurements with data from the scintillators for samples of 50000 events ( accumulated in about one hour with the used setup and trigger conditions ) .
the data set solely contains nights without any technical problems of the used detector stations . in total
it comprises ( dead - time corrected ) an on - time of 208h .
this corresponds to about 150000 events after all cuts with an energy above about 100tev and a zenith angle below 15@xmath26 .
monte carlo simulations revealed that the distance @xmath27 between detector and shower maximum ( defined as the point in the shower development with the maximal number of charged particles ) can be reconstructed independently of the primary mass with the shape parameter _
slope _ of the lateral cherenkov light density distribution : @xmath28 { \rm \,g / cm^2}. \label{dimax}\ ] ] from this relation the distance to the shower maximum is determined with a resolution ( i.e. root mean square ( rms ) of a ( d@xmath29(true)-d@xmath30(reconstructed ) distribution ) ranging from 40g/@xmath5 at 300tev to 20g/@xmath5 at 10pev ( including all detector effects but no systematic error in the mean d@xmath29 ) .
the most important technical improvement in the data presented here to previous experiments is that these values are distinctly smaller than the width of natural shower fluctuations of proton induced showers in the atmosphere ( see below table [ table1 ] ) .
this makes the shape of the penetration - depth distribution a sensitive parameter for the chemical composition ( see section [ sec_reli ] ) .
simple geometrical relations permit to infer @xmath31 , the depth of the shower maximum in the atmosphere , from @xmath32 .
relation [ dimax ] is only weakly energy dependent ( @xcite ) .
this dependence is neglected here .
methods have been developed to reconstruct the primary cr energy from the scintillator and airobicc data _ independently _ of the primary mass with an accuracy better than 35% ( @xcite ; @xcite ; @xcite ) .
however , these methods lead to a relatively strong correlation between reconstructed energy and @xmath31 ( showers with a maximum position that fluctuated to smaller values compared to the mean @xmath31 , are reconstructed with higher energies ) . in order to infer the chemical composition , a careful modelation of the response function between the variables @xmath17 ( or @xmath23 ) and _ slope _ on the one hand and energy and composition ( or penetration depth ) on the other hand ( e.g. via two - dimensional regularised unfolding ) is then necessary .
such procedures have been employed in some analyses of hegra data ( @xcite ; @xcite ) .
two reasons lead us to prefer to circumvent the mentioned problem with the use of two simpler energy estimators here .
+ one reason is that the methods described below are based on physically transparent properties of air - showers inferred from the monte - carlo simulations . whether these properties really hold , is tested to some degree using different energy estimators with different biases and comparing the obtained results .
these consistency checks are an important advantage over more refined and complete methods when it is doubtful how well the monte - carlo simulation describes the data .
the other reason is that the monte carlo statistics at the highest energy is still rather limited and mean shower properties are inferred with higher certainty than a complete response matrix .
the mass independent energy reconstruction methods will be applied to the data in a forthcoming publication together with a discussion of the influence of different eas simulations .
the energy estimators used in this paper and described below are based on @xmath33 and _ slope _ , or @xmath34 .
both estimators are used under the assumption that all primary cr are either protons or iron nuclei .
these extreme assumptions lead to a bias which then has to be corrected for .
+ using @xmath33 and _ slope _ the energy of the primary cosmic - ray nucleus is reconstructed in two basic steps here ( @xcite ; @xcite ; @xcite ) : first _ slope _ ( a measure of the distance to the shower maximum ) is combined with @xmath33 to estimate the number of particles in the shower maximum which is proportional to the energy contained in the electromagnetic component of the eas . in the second step a specific primary mass
is assumed ; with the assumption of primary proton ( iron ) we denote the methods as 1 ( 2 ) .
this allows to calculate the primary energy from the energy deposited in the electromagnetic component .
the following relation was used in our analysis : @xmath35 ) = a \cdot log(n_s({\rm max } ) ) + b.\ ] ] here @xmath36 were obtained from the discrete monte carlo data as 0.965,@xmath372.545 ( 0.890,@xmath372.010 ) for protons ( iron ) .
@xmath38 is the shower size at the maximum of shower development and is inferred from @xmath17 as : @xmath39 with @xmath40 given as ( 0.57833,-85.146,6181.8,-71054 ) for all primary nuclei .
this procedure is valid because the shape of the shower development is only weakly dependent on the mass of the cr nucleus a , especially after the shower maximum ( @xcite ) .
only the fraction of the total energy fed into the electromagnetic cascade depends on a for a given energy per nucleus .
the comparison of the results assuming initially proton and iron primaries is a consistency check for the dependence of shower size at the maximum of shower development on the energy per nucleon .
+ alternatively the energy is reconstructed from the airobicc data alone ( method 3 ( 4 ) with the assumption of proton ( iron ) primaries ) . here
it turns out that @xmath23 is a good estimator of the energy contained in the electromagnetic eas cascades . from simulations
the relation @xmath41 ) = a \cdot log(l_{\rm 90}[{\rm photons / m^2 } ] ) + b\ ] ] was derived , where the coefficients @xmath42 and @xmath43 are given as 0.958,@xmath371.810 ( 0.840,@xmath371.061 ) for primary protons ( iron ) .
_ slope _ ( the parameter used to estimate the primary mass composition , see next section ) is not involved in this energy reconstruction .
for ease of reference the four energy - reconstruction methods are summarised in table [ taberec ] .
.summary of the energy reconstruction methods 1 - 4 as discussed in the text . [ cols="<,^,^,^",options="header " , ] in the last two of these publications , we tried to lessen the dependence of our composition result on the correct absolute @xmath31 values by allowing a free `` shift parameter '' ; the fit to the @xmath31 was then performed with two free parameters : the ratio of light to all nuclei and an overall shift in penetration depth of all mc distributions . in this way the result is mainly determined by the _
shape _ of the @xmath31 distributions ( in first order its width , i.e. rms value ) .
this width depends only weakly on a systematic uncertainty in the determination of _ slope _ , relative to the expected difference of a purely light or heavy composition .
+ fig.[d2chi ] displays the result of such an attempt in a two dimensional plot showing the reduced @xmath44 for various `` fraction of light nuclei '' - `` shift parameter '' combinations for the data lowest and highest in energy .
the shift is varied in an interval @xmath4 30 g/@xmath5 , estimated from the likely systematic uncertainty of our detector and the monte carlo code .
while in the low energy bin small ( p+@xmath45)/all ratios lead to unsatisfactory @xmath44 values for all shift values , in the highest energy bin - well above the knee - practically all fractions give acceptable @xmath44 values for appropriate shifts as seen in fig.[d2chi ] .
the reason for this behaviour is that - given the small number of events in the high - energy bin - the @xmath31 distribution can be fitted both with the relatively broad predominantly light composition shifted to larger depths in the atmosphere and a mixed heavy/ light composition ( where the difference in mean penetration depth of the heavy and light component contributes to the total width ) shifted to small penetration depth .
we have to conclude that it is not reliably possible to determine the composition based mainly on the width of the @xmath31 distribution .
we found in numerical experiments that with this method , and assuming a monte - carlo simulation describing the experimental data well , together with a statistics increased by about a factor of 100 ( which is difficult but not impossible to reach in future experiments ) it will just be possible to reach the desired precision of 10 % mentioned in the introduction on a 2@xmath46 level beyond the knee .
the results of this paper demonstrate that our data seem to be well described by the qgsjet hadron generator within the corsika program : the energy spectra derived with different air - shower observables and assumptions about the chemical composition of cr differ only within the estimated systematic errors .
this is an argument in favour of the contention that the `` knee '' is a feature in the primary flux of cosmic rays , rather than some new effect in the interaction of cosmic rays at very high energies .
such a new effect would probably lead to inconsistencies in the data analysed by a generator that does not take them into account .
+ up to the knee we find a mixed composition and an energy - spectrum power law index consistent with the results of direct experiments at energies around 100 tev .
the data favour a gradual enrichment in heavy elements at energies above the knee but the systematic errors of our experiment do not allow to rule out a constant composition .
an abrupt change of composition or a substantial enrichment in light elements at the knee is ruled out .
we find an elongation rate of about 78 g/@xmath5 and a mean penetration depth consistent with other experimental data on cr with energies higher than studied here ( @xcite ) .
we thank the instituto de astrofisica de canarias ( iac ) and the orm observatory for support and excellent working conditions .
we are grateful to the authors of corsika and to the kascade collaboration for many stimulating discussions .
+ the experiment hegra is supported by the german bmbf and dfg and the spanish cicyt . | measurements of the lateral distribution of cherenkov photons with the wide - angle atmospheric cherenkov light detector array airobicc and of the charged particle lateral distribution with the scintillator matrix of the hegra air - shower detector complex in air showers are reported .
they are used in conjunction to determine the energy spectrum and coarse chemical composition of charged cosmic rays in the energy interval from 0.3pev to 10pev . with the atmospheric shower - front sampling technique these detectors measure the electromagnetic component of an extensive air shower via the lateral density distribution of the shower particles and of the cherenkov photons .
the data are compared with events generated with the corsika program package with the qgsjet hadronic - event generator .
consistency checks performed with primary energy - reconstruction methods based on different shower observables indicate satisfactory agreement between these extensive air shower simulations and the experimental data .
this permits to derive results concerning the energy spectrum and composition of charged cosmic rays .
+ the energy spectrum features a so called `` knee '' at an energy of @xmath0=@xmath1pev .
power law fits to the differential energy spectrum yield indices of @xmath2 below and @xmath3 above the knee .
+ the best - fit elongation rate for the whole energy range is determined to 78.3 @xmath4 1.0 ( stat ) @xmath4 6.2 ( syst ) g/@xmath5 . at the highest energies it seems to decrease slightly .
the best - fit fraction of light nuclei decreases from @xmath6 ( combined statistical and systematic ) to @xmath7 ( combined statistical and systematic ) in the energy range discussed here .
a detailed study of the systematic errors reveals that a non - changing composition can not be excluded . |
in the past fifteen years , our understanding of gamma - ray bursts ( grbs ) had been revolutionized . as usual ,
some aspects are understood better than others .
for example , the detection of a bright supernova component in the afterglow of some nearby long grbs establishes their collapsar origin and the late ( @xmath8 s after the trigger of the burst ) afterglow data support the external forward shock model ( piran 2004 ; zhang & mszros 2004 ) . yet
the physical origin of the prompt emission of grbs is still not clear .
the leading " internal shock model is found hard to explain some observational facts , motivating people to develop the internal magnetic energy dissipation models and the photosphere models ( see * ? ? ? * ; * ? ? ? * for reviews ) .
it is rather hard to distinguish among these models reliably .
it is widely speculated that the polarimetry of the prompt emission , for example , by polarimeters for energetic transients ( poet , hill et al .
2008 ) and by polar ( orsi 2011 ) , may play key roles in the future . in this letter
we show that some empirical correlations of the prompt emission properties may have shed valuable light on the underlying physics and the photospheric model is favored .
the tight correlation @xmath9 was discovered by wei & gao ( 2003 , see fig.6 therein ) and then has been confirmed by many researches ( e.g. , * ? ? ? * ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
recently , a tight correlation @xmath10 was identified by @xcite and a correlation @xmath11 was suggested by @xcite .
very recently , @xcite and @xcite discovered a tight correlation @xmath12 , where @xmath13 is the isotropic - equivalent energy of the prompt emission and @xmath14 is the total energy of the afterglow emission in x - ray band . in the forward shock afterglow model , @xmath14 is proportional to @xmath15 , the kinetic energy of the outflow @xcite .
therefore @xmath16 is proportional to the grb efficiency @xmath17 as long as @xmath18 is considerably smaller than @xmath15 .
hence one has @xmath19 .
some possible interpretations of the @xmath0 correlation can be found in the literature ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) . in this letter
we aim to interpret all the above four correlations together .
correlation ( amati et al .
2002 ) as well as the @xmath20 correlation ( liang et al .
2010 ) , where @xmath21 is the isotropic energy of the prompt @xmath22-rays . both of them are interpretable if one takes the duration of the bursts to be roughly a constant . ]
the starting point is the extensively discussed speculation that the prompt emission of gamma - ray bursts is mainly from the photosphere which suffers significant modification and its spectrum is normally not thermal - like any longer ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
firstly , we discuss _ the simplest scenario _ , in which the ( luminosity , spectral peak energy , efficiency ) of the emission roughly resemble ( @xmath23 , @xmath24 , @xmath25 ) , where @xmath26 are the ( luminosity , temperature , efficiency ) of the photospheric radiation , and @xmath25 and @xmath23 are related to the total luminosity @xmath27 as @xmath28 . in such a scenario , if there are some valid correlations among @xmath23 , @xmath24 , @xmath6 , and @xmath25 , so are @xmath4 , @xmath5 , @xmath6 and @xmath7 . for a relativistic baryonic fireball ,
the acceleration and the subsequent photospheric radiation have been initially investigated by @xcite and by @xcite . following these approaches ,
@xcite have recently derived the expressions of the initial radius of the accelerated outflow ( i.e. , @xmath29 ) and the final lorentz factor of the outflow ( i.e. , @xmath6 ) @xmath30 @xmath31 respectively . for @xmath32 ( actually even for @xmath33 ,
the difference between @xmath34 and @xmath35 is only by a factor of @xmath36 ) , eq.([eq : gamma_ph ] ) reduces to the form obtained by @xcite , i.e. , @xmath37 as shown in @xcite , for the outflow launched via the annihilation of neutrino pairs emitting from a hyper - accreting disk , the dimensionless entropy of the initial outflow is related to the total luminosity as @xmath38 ( a @xmath39 is derived if the poorly understood collimation process is ignored @xcite . in the following derivation
we regard k as a free parameter " ) .
the final lorentz factor of the accelerated outflow is related to the initial dimensionless entropy as @xmath40 . as long as the thermal radiation is not extremely efficient ( say , @xmath41 ) s ) better - understood afterglow data suggests a typical grb efficiency @xmath42 ( e.g. , fan & piran 2006 ) . ] ,
approximately we have @xmath43 combining eq.(1 ) with eq.(3 ) , we have @xmath44 substituting this relation into eq.(4 ) we have @xmath45 hence eq.(4 ) and eq.(1 ) give @xmath46 and @xmath47 respectively .
finally we have @xmath48 so far we have shown that some correlations should be present . in the current scenario ,
@xmath49 largely resembles @xmath50 , respectively .
so if we take @xmath51 , the expected relations are @xmath52 respectively , which are nicely in agreement with the four correlations summarized in the first paragraph of this section and the only requirement is that @xmath29 depends on @xmath4 insensitively .
_ interestingly _ , the required @xmath51 is close to that ( @xmath39 ) found in a simple analytical approach @xcite .
actually when adopting eq.(18 ) and eq.(16 ) of fan & wei ( 2011 ) , we have @xmath53 and @xmath54 , the coefficients are consistent with those reported in the literature , as long as @xmath29 is in order of @xmath55 cm . these facts together with the plots in fig.1 illustrate that the correlations found in the literature ( including the normalization ) are indeed interpretable within the photosphere model .
secondly , we adopt the so - called _ generic " dissipative photospheric model _ developed by @xcite , in which it is shown that at the radius @xmath56 ( see eq.(5 ) therein ) , where radiation and electrons drop out of equilibrium , the spectral peak of the prompt emission forms ( the optical depth is about tens , see eq.(6 ) therein ) rather than that at the photospheric radius . ] and the lorentz factor can be expressed as ( see eq.(9 ) therein ) @xmath57 where @xmath58 is the number of electron+positron pairs per proton and is expected to be moderate . the acceleration calculation yields @xmath59 ( e.g. , * ? ? ?
* ; * ? ? ?
* ) , with which we have and then @xmath60 .
these coefficients are comparable with that of the observed correlations as long as @xmath61 cm . ]
@xmath62 with the relation @xmath63 , eq.([eq : gen-1 ] ) and eq.([eq : gen-2 ] ) give @xmath64 respectively .
substituting eq.([eq : gen-4 ] ) into eq.([eq : gen-2 ] ) and eq.([eq : gen-3 ] ) , we have @xmath65 @xmath66 respectively .
as long as the radiation efficiency is not very efficient ( say @xmath67 ) , one can take @xmath68 @xcite .
for @xmath51 we have @xmath69which are roughly consistent with the correlations summarized at the beginning of this section .
both long and short grbs follow the @xmath0 correlation @xcite and the @xmath70 correlation @xcite .
when taking the peak time of the gev emission of the short grb 090510 as the deceleration time of the forward shock , we found that the inferred bulk lorentz factor also follows the @xmath1 correlation .
such facts suggest that the photospheric origin of the prompt emission may also apply to some short bursts .
diagram for the bursts discussed in ( * ? ? ?
* excluding those with a @xmath6 in dispute , for example grb 090510 and grb 090328a ) and for the time - resolved thermal radiation of grb 090902b .
the solid line is the best fit @xmath71 obtained in @xcite .
( b ) the @xmath0 diagram for the bursts investigated in @xcite and for the time - resolved thermal radiation of grb 090902b ( please note that we have taken @xmath72 , where @xmath73 is the observed temperature ) . the solid line is the best fit @xmath74 found in @xcite.,width=321 ]
prominent thermal radiation components have been identified in grb 090902b , a very bright burst at a redshift @xmath75 @xcite .
for example , @xcite divided the whole data set of grb 090902b into several time bins and showed that the spectrum in each bin can be nicely fitted by a thermal component plus a power - law spectral component . by applying the same technique ,
we redo the analysis using _ _ f__ermi / gbm data and the newest _ _
f__ermi / lat pass7 data .
the thermal ( blackbody ) and non - thermal ( power - law ) spectral parameters and fluxes are derived in each time bin .
following @xcite and @xcite and assuming a constant thermal radiation efficiency @xmath76 , the bulk lorentz factors of the outflow shells can be straightforwardly evaluated .
we plot the inferred @xmath6 together with the simultaneous luminosity in the @xmath1 diagram presented by @xcite .
as shown in fig.1(a ) these two sets of data are in agreement with each other . for most bursts discussed in @xcite
the measurement of @xmath6 was based on the modeling of the afterglow light curve(s ) .
the physics involved in such a kind of estimation is completely different from that for grb 090902b .
the agreement between these two sets of data thus not only supports our speculation of the photospheric origin of the prompt emission but also validates the robustness of both methods of evaluating @xmath6 . in fig.1(b ) we plot the time - resolved spectral peak energy versus the simultaneous luminosity of grb 090902b in the @xmath0 diagram presented by @xcite .
again , a nice agreement between these two sets of data is present , in support of the photospheric origin of the prompt emission of some gamma - ray bursts .
finally , we d like to point out that all these correlations have not been reasonably interpreted in either the internal shock models or the internal magnetic energy dissipation models ( the outflow is magnetic ) . in the standard internal shock model
, one has @xmath77 ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) then we expect no evident positive correlation between @xmath5 and the luminosity after taking into account the correlation @xmath78 , at odds with the data .
it is also straightforward to show that the correlation @xmath78 predicts an extremely low internal shock efficiency unless the slow material shell has a width much widely than that of the fast shell ( i.e. , the duration of ejecting the slow shell is needed to be a factor of @xmath79 that of ejecting the fast shell , where @xmath80 and @xmath81 are the bulk lorentz factor of the fast and slow shells , respectively ) . for a magnetic outflow ,
it is recognized in @xcite that an interpretation of @xmath1 correlation is not available yet , let alone interpret the others .
all these facts strongly favor the suggestion that the dominant component of the prompt emission of some grbs may be tightly relevant to the photospheric radiation process , though much work on getting a spectrum nicely matching the data is still needed ( veres , zhang & meszaros 2012 , in preparation ) .
we thank b. zhang , d. giannios , r. margutti and m. g. bernardini for helpful communications and the referee for insightful comments .
this work was supported in part by national natural science of china under grants 10973041 , 10921063 , 11073057 and 11163003 , and by national basic research program of china under grant 2009cb824800 .
yzf is also supported by the 100 talents program of chinese academy of sciences .
bbz acknowledges the support from nasa sao sv4 - 74018 .
abdo , a. a. , et al .
2009 , , 706 , l138 amati , l. , et al .
2002 , , 390 , 81 barniol duran , r. , & kumar , p. 2011
, , 417 , 1584 beloborodov , a. m. , 2010 , mnras , 407 , 1033 bernardini , m. g. , margutti , r. , zaninoni , e. , & chincarini , g. 2012 , mnras submitted ( arxiv:1203.1060 ) dai , z. g. & lu , t. 2002 , apj , 580 , 1013 fan , y. z. & piran , t. 2006 , mnras , 369 , 197 fan , y. z. & wei , d. m. 2005 , mnras , 364 , l42 fan , y. z. , & wei , d. m. 2011 , apj , 739 , 47 ghirlanda , g. , nava , l. , ghisellini , g. , celotti , a. , & firmani , c. 2009 , a&a , 496 , 585 ghirlanda , g. et al .
2012 , mnras , 420 , 483 giannios , d. 2012 , , 422 , 3092 hill , j. e. , et al .
2008 , aip conf .
1065 , 331 ioka , k. , murase , k. , toma , k. , nagataki , s. , & nakamura , t. , 2007 , apj , 670 , l77 lazzati , d. , morsony , b. j. , & begelman , m. c. 2011 , apj , 732 , 34 liang , e. w. , dai , z. g. , & wu , x. f. , 2004 , apj , 606 , l29 liang , e .- w .
, et al .
2010 , , 725 , 2209 liu , r .- y . , & wang , x .- y . 2011 , , 730 , 1 l , j. , et al . , 2012 ,
apj , 751 , 49 margutti , r. , et al .
, 2012 , mnras submitted ( arxiv:1203.1059 ) mszros , p. , laguna , p. , & rees , m. j. 1993 , apj , 415 , 181 orsi , s. 2011 , astrophys .
space sci .
trans . 7 , 43 pandey , s. b. , et al . 2010 , , 714 , 799 peer , a. , ryde , f. , wijers , r. a. m. j. , mszros , p. , & rees , m. j. 2007 , apj , 664 , l1 peer , a. , et al .
2012 , , 420 , 468 piran , t. , 2004 , rev .
, 76 , 1143 piran , t. , shemi , a. , narayan , r. 1993 , mnras , 263 , 861 rees , m. j. & mszros , p. 2005
, apj , 628 , 847 ryde .
f , et al . , 2010 apjl , 709 , l172 yonetoku , d. , et al . 2004 , , 609 , 935 wei , d. m. & gao , w. h. 2003 , mnras , 345 , 743 zhang , b. & mszros , p. 2002 , apj , 581 , 1236 zhang , b. , & mszros , p. 2004 , int . j. mod
a. , 19 , 2385 zhang , b. b. , et al .
2011 , apj , 730 , 141 zhang , f. w. , shao , l. , yan , j. z. , & wei , d. m. , 2012 , apj , 750 , 88 | we show that the experiential @xmath0 , @xmath1 , @xmath2 and @xmath3 correlations ( where @xmath4 is the time - averaged luminosity of the prompt emission , @xmath5 is the spectral peak energy , @xmath6 is the bulk lorentz factor and @xmath7 is the emission efficiency of gamma - ray bursts ) are well consistent with the relations between the resembling parameters predicted in the photospheric radiation model of the prompt emission of gamma - ray bursts . the time - resolved thermal radiation of grb 090902b does follow the @xmath0 and @xmath1 correlations
. a reliable interpretation of the four correlations in alternative models is still lacking .
these facts may point towards a photospheric origin of prompt emission of some gamma - ray bursts . |
at very high energies @xmath3 in the small-@xmath4-region the multi - regge - kinematics ( mrk ) respectively the quasi - mrk ( qmrk ) is an especially suited approximation for the calculation of lepto- and hadroproduction cross sections .
we consider the central heavy quark - antiquark - hadroproduction process in figs .
[ fig1],[fig2 ] and choose the frame where the proton momenta are given by @xmath5 in the qmrk we have then an almost transverse momentum transfer @xmath6 and for the t - channel momenta the relations @xmath7 but in general no strong ordering of the momenta of the produced heavy quark and antiquark @xmath8 .
this kinematics corresponds to the production of a @xmath9-cluster with finite invariant mass and therefore makes the production of a @xmath10-bound state possible .
the associated factorization scheme is the @xmath11-factorization and the particles in the t - channel are allowed to have finite transverse momenta .
the cross section for the inclusive hadroproduction of a @xmath9-pair at very high energies where gluon exchange dominates is then given by @xmath12 the most important parts of this equation are the hard scattering amplitude @xmath13 and the unintegrated gluon distribution function @xmath14 , where the longitudinal momentum fractions are defined as usual , @xmath15 and @xmath16 . due to the qmrk the gluons are off - shell and longitudinal , which corresponds to the replacement @xmath17 e.g. for the numerator of the upper gluon propagator .
the use of the qmrk and off - shell gluons together with the requirement of gauge invariance determines the @xmath9-production amplitude @xmath13 in terms of the nlla bfkl production vertex @xmath18 , which is explicitely given by fadin and lipatov in @xcite @xmath19 where the function @xmath20 is defined through @xmath21 and a very similar equation holds for the crossed contribution @xmath22 .
equation ( [ b ] ) includes in the second part on the right hand side the effective gluon - production - vertex @xmath23 the first term of the right hand side of this equation is the standard three - gluon - coupling , while the second term in curly brackets is an additional part which is essential for gauge invariance . the function @xmath24 satisfies the ward - slavnov - indentity @xmath25 , and gauge invariance leads to the vanishing of the amplitude @xmath26 in the limit of vanishing transverse momentum . in order to be able to perform a reasonable comparison to experimental data we use the unintegrated gluon distribution function @xmath27 which has been calculated in @xcite .
the authors extracted @xmath28 from a fit to @xmath29 data and obtained an excellent @xmath30ndof with the use of the following integrated gluon distribution at an input scale of @xmath31gev@xmath32@xmath33 which enters the calculation through the relation @xmath34 for a correct treatment of the region of small transverse gluon momenta @xmath35 we have to modify the cross section formular ( [ crosssection2 ] ) in a way which is compatible with equation ( [ integrated ] ) and the vanishing of the amplitude in the limit @xmath36 , for details see @xcite .
our result for the production of a single bottom quark is shown in fig .
[ singleb ] in comparison to data from the tevatron collider @xcite and nlo calculations in the collinear approach .
the error band is due to a simultaneous variation of the bottom quark mass @xmath37 gev @xmath38 gev and the fundamental scale @xmath39 mev@xmath40 mev , while the scale dependence of the coupling constant @xmath41 is chosen to be @xmath42 .
we find in contrast to the nlo collinear calculation a good agreement between the @xmath43-factorization prediction and the experimental values over the whole range of @xmath44 .
likewise important is the comparison to measurements of @xmath45-correlations @xcite in fig .
[ bbbar65 ] .
allthough the cross section ( [ crosssection2 ] ) is only of lo in @xmath46 , the @xmath11-factorization together with the effective production vertex and off - shell gluons gives a very good despription of the data which lie about a factor of 3 above the nlo collinear result .
motivated by the results for heavy quark production we now pass over to the hadroproduction of charmonia which is a far more controversial subject @xcite .
we consider the production of the @xmath47 , which is a @xmath48-state and the @xmath49 which is a @xmath50 state in the framework of the non - relativistic - qcd ( nrqcd ) approach and have therefore to include the color singlet and color octet contributions to the final color singlet bound states .
the calculation is again based on the cross section formula ( [ crosssection2 ] ) and the unintegrated gluon distribution from @xcite .
the details of the calculations and the binding mechanisms @xmath51 are explained in .
because the color octet contributions depend linearly on the so called color octet matrix elements ( come ) which at the moment can not be calculated but have to be fitted to the data @xcite , the following results should not to be seen as predictions of the corresponding differential cross sections but are required in order to extract the sizes of the come .
the results for the production of the @xmath52 from radiative @xmath47-decay in the @xmath11-factorization are shown on the left side of fig .
[ chi ] together with experimental data from the tevatron and a nlo collinear calculation .
it is clear from the figure that the @xmath53-contribution to @xmath47 is generally suppressed due to the flat slope of the curve in comparison to the data .
the flat curve originates from the additional part of the effective vertex in equation ( [ vertex ] ) which contains factors @xmath6 and leads to a strong suppression of the corresponding come with respect to the results from collinear calculations .
furthermore the landau - yang - theorem , forbidding the production of the @xmath54-state in lo collinear factorization , is not valid in our approach because the gluons are off - shell .
this becomes obvious in a rather impressive way on the right hand side of fig .
[ chi ] where the individual @xmath54- and @xmath55-contributions are plotted .
[ jpsi ] shows our results for direct @xmath56-production in the @xmath11-factorization framework in comparison to data and the nlo color singlet collinear prediction @xcite .
although in this case the color octet contributions @xmath57 , @xmath58 play an important role , the @xmath53-part is suppressed following the same reasons as in the @xmath47-production .
this leads again to a strong reduction of the corresponding come @xmath59 with respect to the collinear result .
this in turn has a definite influence on the production of polarized @xmath56 @xcite and could clarify the so far unsolved puzzle concerning the disagreement between recent experimental results and the nlo collinear predictions of transversally polarized charmonia @xcite .
this work was supported by the dfg and the bmbf .
99 s. catani , m. ciafaloni and f. hautmann , phys .
b242(1990 ) 97 ; j.c .
collins and r.k .
ellis , nucl .
b360(1991 ) 3 ph .
hgler , r. kirschner , a. schfer , l. szymanowski , o.v .
teryaev , phys.rev.d 62 ( 2000 ) 071502 v.s .
fadin and l.n .
lipatov , nucl .
b477(1996 ) 767 j. kwiecinski , a.d .
martin , a.m. stasto , phys.rev .
d56(1997 ) 3991 b. abbott et al . , phys.lett b487(2000 ) 264 f. abe et al . , phys.rev .
d55(1997 ) 2546 abe et al . ,
phys.rev.lett.79(1997 ) 578 abe et al . ,
cdf collaboration , fermilab - conf-95/226-e ph .
hgler , r. kirschner , a. schfer , l. szymanowski , o.v .
teryaev , phys.rev.lett .
86 ( 2001 ) 1446 ph .
hgler , r. kirschner , a. schfer , l. szymanowski , o.v .
teryaev , phys.rev.d63 ( 2001 ) 077501 p. cho , a.k .
leibovich , phys.rev .
d53(1996 ) 150 p. cho , a.k .
leibovich , phys.rev .
d53(1996 ) 6203 f. yuan , k .-
chao , phys.rev.lett .
86(2001 ) 022002 m. beneke , m. kramer , phys.rev .
d55(1997 ) 5269 e. braaten , b.a .
kniehl , j. lee , phys.rev .
d62(2000 ) 094005 t. affolder et al . , phys.rev.lett .
85(2000 ) 2886 b.a .
kniehl , j. lee , phys.rev .
d62(2000 ) 114027 | calculations of heavy quark and charmonium hadroproduction are presented in the @xmath0-factorization approach .
the resulting differential cross sections are compared to experimental values and nlo collinear predictions .
the application of the full effective fadin - lipatov-@xmath1-production - vertex has an important influence on the results and in particular leads to a strong suppression of certain color octet contributions which are related to the outstanding problem of the production of polarized @xmath2 . |
in the past decades there has been an enormous effort , both theoretically and experimentally , in trying to understand the origin of electroweak symmetry breaking . within the standard model ( sm ) ,
the higgs mechanism provides a solution that demands the existence of a fundamental self - interacting scalar field , the higgs boson . after the analysis of significant amounts of data , the lhc and tevatron experiments seem to observe , for the first time , an excess of events that could be associated with a higgs - boson signal .
if this excess is confirmed , a new set of questions will need to be addressed .
in particular , it will be crucial to understand how the electroweak scale @xmath6 is related , if at all , to other scales in nature such as the planck scale @xmath7 .
searching for a framework that provides a natural connection between these scales has been one of the main reasons to propose theories beyond the sm .
some of these theories , such as supersymmetry , promote a perturbative extension of the sm up to scales close to @xmath7 .
others feature new strong interactions and provide a solution valid only up to energies a few orders of magnitude above the weak scale , without necessarily specifying the ultraviolet ( uv ) completion of the theory . in all cases ,
new particles are expected to appear at or slightly above the tev scale to solve the naturalness problem of the higgs mass parameter .
these new particles are expected to have non - negligible effects on the production and decay rates of the higgs boson . in the event of a higgs discovery ,
any departure from sm expectations will then pave the way to probing models of new physics . from this perspective
, precision higgs - boson physics much like rare weak decays and flavor - changing neutral current ( fcnc ) processes is an exquisite tool for probing the structure of electroweak interactions at the quantum level .
the couplings of the higgs boson to photons and gluons vanish at tree level in the sm , but they are non - zero at one - loop order and beyond
. these couplings are therefore particularly sensitive to new heavy particles , which can propagate in the loops . in this article , we concentrate on analyzing higgs - boson properties in models with a warped extra dimension , so - called randall - sundrum ( rs ) models @xcite , which represent a very appealing alternative to more traditional extensions of the sm , such as supersymmetry .
these models provide not only a natural solution to the hierarchy problem , but also a very compelling theory of flavor .
they feature a compact extra dimension with a non - factorizable anti - de sitter ( ads ) metric and two four - dimensional ( 4d ) branes as the boundaries of the warped extra dimension .
the ads background generates an exponential hierarchy of energy scales , so that the fundamental mass scale near the `` uv brane '' is of order @xmath7 , while that near the `` infrared ( ir ) brane '' is suppressed by an exponential warp factor @xmath8 , with @xmath9 being the ads curvature and @xmath10 the radius of the extra dimension .
while all fundamental parameters are of planck size , @xmath11 , due to the curvature of the extra dimension the fundamental scale of the theory near the ir brane can be in the range of a few tev .
hence , as long as the higgs sector is localized near the ir brane , quantum corrections to the higgs potential are cut off at the few - tev scale and the hierarchy problem is solved .
in addition , the large hierarchies observed in the spectrum of fermion masses and in the quark mixing matrix arise in a natural way from the localization of bulk fermions in the extra - dimensional space @xcite .
an important side effect of this mechanism is that flavor - changing couplings between sm fermions , which arise due to the tree - level exchange of kaluza - klein ( kk ) gauge bosons , are suppressed by the same small overlap integrals that generate the fermion masses @xcite , a scheme referred to as the rs - gim mechanism @xcite . in order to fulfill the constraints coming from electroweak precision data and fcnc transitions , the low - lying kk excitations of the sm particles must have masses in the range of a few tev , and hence may be out of the reach for direct production at the lhc . in recent years
, many models have been proposed to alleviate these constraints and address the little hierarchy problem of the basic rs model with bulk fermions .
the model building becomes rather involved and ranges from considering the effect of brane - localized kinetic terms for the gauge fields @xcite to enlarging the bulk gauge symmetries @xcite , thereby adding a whole new layer of additional matter fields .
another class of models takes the higgs sector off the ir brane .
this is particularly well motivated if the higgs boson is identified with the @xmath12 component of a gauge field ( `` gauge - higgs unification '' ) @xcite . in this paper
, we will concentrate on studying the effects of kk excitations on the production of a sm - like higgs boson in the context of the simplest rs model containing bulk fermions and a minimal scalar sector localized on the ir brane .
in particular , we will scrutinize the gluon - gluon fusion process , for which the presence of kk modes in loop corrections can have important effects
. the type of considerations discussed in this work will also be relevant for the higgs - boson coupling to two photons , which is the other loop - induced coupling of the higgs field to gauge bosons . at present , the literature contains two sets of results for the computation of higgs production and decays in rs - type models that are at variance with each other . while in @xcite the authors find a significant suppression of the gluon - gluon fusion cross section relative to the sm , the authors of @xcite find an enhancement .
indeed , when applied to the same model , the two groups obtain effects on the @xmath13 amplitude of approximately equal strength but opposite sign . in this article
we revisit in detail both calculations and present alternative ways to derive them .
for the first time , we obtain closed analytical expressions for both results in terms of the fundamental parameters of the model , valid to all orders in the ratio of the higgs vacuum expectation value @xmath14 and the typical mass scale @xmath15 of kk modes .
we find that the calculations of @xcite are both correct from a mathematical point of view .
the origin of their difference is related to a subtlety connected to the fact that , in order to compute its overlap with the fermion wave functions , it is necessary to regularize the higgs profile in an intermediate step by taking it ( slightly ) off the ir boundary .
the difference between the results obtained by the two groups arises from the different orders in which the higgs regulator is removed and the sum over the kk tower is performed .
we then show that the enhancement in the gluon - gluon fusion cross section obtained in @xcite arises from very heavy kk modes . when the model is defined with a ( warped ) planck - scale cutoff so as to solve the hierarchy problem , the cross section is suppressed , in accordance with the result obtained in @xcite .
the paper is organized as follows . in section [ sec : setup ]
we define our notation and present the setup of the problem . in section [ sec : leff ] we present the low - energy lagrangian for the effective @xmath13 couplings induced by the virtual effects of kk fermions , which is valid for energies below the kk mass scale .
we also describe the setup of the two calculations , one summing first over the infinite kk tower and then taking the higgs regulator to zero , and the other considering the limits in the reversed order . in section [ sec : azatov ]
we perform the first calculation and reproduce the results of @xcite in a different way by using the five - dimensional ( 5d ) fermion propagator of the theory . in this way
, we generalize the result to all orders in @xmath16 . in section
[ sec : finitesum ] we perform the second calculation , considering first a simplified case , in which it is straightforward to obtain a closed answer for the sum over kk modes . based on symmetry considerations
, we then make a conjecture for how to extend this result to the general case , which reproduces the numerical results obtained in @xcite . in section [ sec : interp ]
we explore in more detail the origin of the difference between the two approaches , which in rs models with an ir - localized higgs sector arises from unphysical contributions of kk modes with `` trans - planckian '' masses .
we then argue in section [ sec : cutoff ] that the use of a consistent uv regularization scheme , which for instance is provided by the physical uv cutoff inherent in rs models , eliminates these contributions , thus favoring the result corresponding to a suppression of the gluon - gluon fusion cross section .
phenomenological implications of our findings for higgs - boson searches at the lhc are studied in section [ sec : pheno ] , while section [ sec : concl ] contains our conclusions .
we work with the non - factorizable rs geometry @xcite @xmath17 where @xmath18 denote the coordinates on the 4d hyper - surfaces of constant @xmath19 with metric @xmath20 .
the @xmath12 dimension is an @xmath21 orbifold of size @xmath10 , labeled by @xmath22 $ ] .
two 3-branes are located at the orbifold fixed points at @xmath23 ( uv brane ) and @xmath24 ( ir brane ) .
the curvature parameter @xmath9 and the radius @xmath10 of the extra dimension are assumed to be of planck size .
the warped planck scale @xmath25 sets the effective , position - dependent fundamental scale at a given point along the extra dimension .
it serves as a natural uv cutoff , since quantum gravity would become relevant above this scale . in order to solve the ( big ) hierarchy problem
, one assumes that the effective cutoff scale on the ir brane , @xmath26 , is in the range of 20tev or so .
this requires that @xmath27 .
the `` little hierarchy problem '' , the fact that the higgs - boson mass is two orders of magnitude smaller than the cutoff scale , is not addressed in the minimal rs framework .
the warped curvature scale , @xmath28 , is assumed to lie somewhat lower , in the range of a few tev .
it sets the mass scale for the low - lying kk excitations of the sm fields .
for instance , the masses of the first kk photon and gluon states are approximately @xmath29 @xcite . it will be convenient to introduce a coordinate @xmath30 , which equals @xmath31 on the uv brane and 1 on the ir brane @xcite .
we consider the minimal rs model in which all sm matter and gauge fields propagate in the bulk , while the higgs boson lives on the ir brane .
the 5d theory contains three generations of massive fermions @xmath32 and @xmath33 , @xmath34 , which transform as doublets and singlets under the sm gauge group @xmath35 , respectively . the electroweak gauge symmetry @xmath36 is broken by the higgs sector on the ir brane .
the 5d fields are functions of @xmath37 and @xmath19 .
the doublet fields @xmath38 have left - handed zero modes , while the singlet fields @xmath33 , @xmath34 have right - handed ones .
these correspond to the chiral fermions of the sm .
in this paper we are particularly interested in the couplings of the sm higgs boson to fermions . in unitary gauge
, the higgs - boson couplings to sm quarks and their kk excitations read @xmath39 where the yukawa couplings are given in terms of the overlap integrals @xcite
@xmath40 \\ & = \frac{1}{\sqrt2 } \int_\epsilon^1\!dt\,\delta^\eta(t-1)\,\,{\cal u}_l^{\dagger(m)}(t ) \bigg ( \begin{array}{cc } \bm{0 } & \bm{y}_u \\ \bm{y}_u^\dagger & \bm{0 } \end{array } \bigg)\,{\cal u}_r^{(n)}(t ) \ , , \end{aligned}\ ] ] and likewise in the down - type quark sector . here @xmath41 labels the different mass eigenstates @xmath42 in the 4d effective theory , such that @xmath43 refer to the sm up - type quarks @xmath44 , while @xmath45 label the six fermion modes of the first kk level , and so on .
the quantities @xmath46 and @xmath47 with @xmath48 are diagonal @xmath49 matrices in flavor space , which contain the @xmath50-even and odd fermion profiles along the extra dimension , respectively . these can be expressed in terms of combinations of bessel functions , whose rank depends on the bulk mass parameters @xmath51 and @xmath52 of the 5d fermion fields @xcite .
the @xmath35 gauge symmetry in the bulk implies that @xmath53-doublet quark fields have common @xmath54 parameters .
the 3-component vectors @xmath55 , on the other hand , describe the flavor mixings of the 5d interaction eigenstates into the 4d mass eigenstates , which are generated by the yukawa interactions on the ir brane @xcite . because of electroweak symmetry breaking , these vectors are different for @xmath56 .
for simplicity , from now on we use the generic notation @xmath38 for @xmath57 and @xmath58 for @xmath59 .
the @xmath49 matrices @xmath60 contain the dimensionless yukawa matrices of the 5d theory , which contrary to the sm are assumed to have an anarchical structure , i.e. , they are non - hierarchical matrices with @xmath61 complex elements . the hierarchies of the yukawa matrices of the sm quarks in the effective 4d theory are explained in terms of a geometrical realization of the froggatt - nielsen mechanism in rs models @xcite . in the second line of ( [ gdef ] )
we use a compact notation , in which the profile functions of the left - handed ( right - handed ) interaction eigenstates that can mix into the left - handed ( right - handed ) components of the 4d mass eigenstates are collected in 6-component vectors @xmath62 .
these vectors obey the orthonormality conditions ( with @xmath63 ) @xmath64 they are defined in terms of the kk decompositions @xmath65 where @xmath66 denote the left- and right - handed components of the @xmath67 dirac fermion .
note that even the diagonal yukawa couplings @xmath68 derived from ( [ gdef ] ) are in general complex numbers , such that pseudo - scalar currents appear in the lagrangian @xmath69 , \ ] ] which in principle could have interesting implications for phenomenology .
unfortunately , however , for the sm fermions the imaginary parts of the yukawa couplings are very strongly suppressed @xcite .
it has been emphasized in @xcite that in order to properly evaluate the yukawa couplings it is necessary to regularize the higgs - boson profile .
the reason is that with a brane - localized higgs sector the odd fermion profiles @xmath47 are discontinuous at @xmath70 , and hence the overlap integral of a product of two such functions with the naive higgs profile @xmath71 is ill defined . in ( [ gdef ] ) , we have therefore replaced this profile with a regularized delta function @xmath72 .
the precise shape of this function will be irrelevant ; however , it is important that the higgs profile is non - zero only within a small interval of width @xmath73 next to the ir brane , @xmath74 only if @xmath75 , and that it has unit area .
the @xmath76-dependence of the yukawa couplings is implicit in our notation , but it will play an important role in our analysis . in the presence of the regularized higgs profile ,
the bulk equations of motion ( eoms ) for the profile functions read @xcite @xmath77 where @xmath78 are the mass eigenvalues , and @xmath79 is the generalized mass matrix . without loss of generality
, the hermitian bulk mass matrices @xmath80 of the 5d theory can be taken to be diagonal .
the boundary conditions are such that the odd profiles vanish on the two branes , which implies @xmath81 note that these simple dirichlet boundary conditions only hold because the higgs profile has been regularized and is no longer singular on the ir brane .
when @xmath82 is lowered away from 1 , the profile functions change rapidly over the small interval in which the higgs profile is non - zero .
a careful analysis of the solutions in the vicinity of the ir brane shows that at @xmath83 the solutions obey the equation ( assuming @xmath73 ) @xcite @xmath84 where @xmath85 is a positive definite , hermitian matrix given in terms of the 5d anarchic yukawa matrices of the model .
note that the mixed boundary conditions ( [ naivebcs ] ) hold irrespective of the shape of the regularized higgs profile , as long as it is confined to a box of width @xmath76 .
it is thus possible to take the limit @xmath86 corresponding to a brane - localized higgs boson , drop the yukawa couplings in the bulk eoms ( [ eoms ] ) and ( [ mqdef ] ) , and impose the mixed boundary conditions on the ir brane , i.e. , at @xmath87 , where @xmath88 denotes a point infinitesimally to the left of the ir brane .
the profile functions are discontinuous on the ir brane , so it is necessary to stay an infinitesimal amount away from the boundary . ]
note , however , that to derive the result ( [ naivebcs ] ) it was important to regularize the higgs profile in an intermediate step @xcite .
the mixed boundary conditions ( [ naivebcs ] ) evaluated at @xmath87 are sufficient to determine the masses and profiles of the fermionic kk modes required , e.g. , for the analysis of many tree - level flavor - changing processes @xcite . however
, when computing overlap integrals of fermion profiles with the wave function of the higgs boson the correct procedure is to start from a regularized higgs profile , compute the relevant overlap integrals , and then take @xmath76 to zero @xcite . using the explicit expressions for the profile functions near the ir brane derived in @xcite , which were obtained from the solution of the system of equations ( [ eoms ] ) to ( [ irbcs ] ) assuming @xmath73 , it is straightforward to derive from ( [ gdef ] ) that @xmath89\ , \bm{y}_q\,\bm{c}_n^{(q)}(1-\eta)\,a_n^{(q ) } \ , , \end{aligned}\ ] ] where the cosh and sinh functions in brackets arise from the @xmath50-even and odd fermion profiles in ( [ gdef ] ) , respectively .
we have introduced the notation @xmath90 for the integral of the regularized higgs profile , such that @xmath91 and @xmath92 for @xmath93 .
we observe that the products of the @xmath82-dependent fermion profiles add up to a constant , so we are left with the normalization integral over the higgs - boson wave function , which equals 1 .
we thus obtain @xmath94 interestingly ( and somewhat unexpectedly ) , the same result can be obtained in a more naive way . by evaluating relation ( [ gdef ] ) with a brane - localized higgs boson
, we obtain @xmath95 \ , , \\ & = \frac{\sqrt 2\pi}{l\epsilon}\,a_m^{(q)\dagger}\,\bm{c}_m^{(q)}(1 ^ - ) \left ( 1 - \frac{v^2}{2m_{\rm kk}^2}\,\bm{\tilde y}_q\,\bm{\tilde y}_q^\dagger \right ) \bm{y}_q\ ,
\bm{c}_n^{(q)}(1 ^ -)\,a_n^{(q ) } \ , , \end{aligned}\ ] ] where in the second step we have used the mixed boundary conditions ( [ naivebcs ] ) at @xmath87 .
this result agrees with ( [ g1mi ] ) by virtue of the definition of @xmath96 in ( [ xqdef ] ) .
the yukawa couplings derived in ( [ g1mi ] ) will be used in the analysis in section [ sec : finitesum ] . for the discussion in other parts of our paper
, it will be necessary to keep the regulator @xmath76 non - zero until the sum over the tower of kk modes has been performed .
we are now ready to derive the effective low - energy lagrangian for the higgs - boson couplings to a pair of gluons , which are induced by the exchange of kk quarks .
this lagrangian is valid at energies below the scale @xmath15 , at which these states can be integrated out .
the relevant feynman diagram arising at one - loop order is shown in figure [ fig : graph ] .
since the gluon couplings to fermions are diagonal in the mass basis , a single quark state @xmath98 runs in the loop . summing over the kk tower , we obtain @xmath99 where @xmath100 is the scale at which the effective operators are renormalized , and @xmath101 ( with @xmath102 ) is the dual field - strength tensor . throughout this paper
, @xmath14 denotes the value of the higgs vacuum expectation value in the rs model , which differs from the sm value @xmath103gev by a small amount @xcite .
to first order in @xmath104 , we obtain from the shift of the @xmath105-boson mass predicted by the rs model @xcite @xmath106 where @xmath107 . at the matching scale @xmath108
, we find for the wilson coefficients entering ( [ lhgg ] ) the one - loop results @xmath109 where @xmath110 .
the sums extend over all kk fermion states but exclude the sm fermions ( @xmath111 ) .
this is indicated by the prime on the sum symbol . using the results from @xcite
, we arrive at @xmath112 \
, , \qquad b(\tau ) = \tau \arctan^2\frac{1}{\sqrt{\tau-1 } } \,.\ ] ] since the kk quarks are much heavier than the higgs boson , it is sufficient to use the asymptotic values @xmath113 of these functions .
it then follows that @xmath114 the real part of the sum determines @xmath115 , while the imaginary part gives @xmath116 .
we recall at this stage that each term in the sum ( both the yukawa couplings and the mass eigenvalues ) depends on the regulator @xmath76 used to smear out the higgs profile .
in addition , one needs to worry about the convergence of the infinite sum over kk modes .
indeed , since the yukawa couplings @xmath68 are of @xmath61 and the masses of the kk modes are approximately evenly spaced multiples of the kk scale , naive dimensional analysis would suggest that the sum diverges logarithmically . in order to define it properly
, one should therefore regularize the sum , for instance by introducing a cutoff on the highest kk level that is included .
we thus define @xmath117 where the sum now includes the sm quarks ( @xmath43 ) plus the first @xmath118 levels of kk modes .
each kk level contains six modes with @xmath119 mass splittings , while the different levels are split by an amount of @xmath120 .
it will turn out that the sum over modes is finite despite naive expectation , because non - trivial cancellations happen among the six modes contained in each kk level .
nevertheless , the notion of a cutoff on the kk level will play an important role in our analysis .
the wilson coefficients in ( [ c1c5 ] ) are now obtained as @xmath121 \,,\ ] ] where the last term subtracts the contributions from the sm quarks , each of which equals 1 up to higher - order corrections in @xmath16 .
these contributions can be derived using the explicit expressions for the profile functions of sm quark fields near the ir brane given in @xcite as well as the relations for the @xmath55 vectors presented in @xcite .
we obtain @xmath122 where the quantity @xmath123 contains some small corrections to the leading term .
it receives its main contributions from the third - generation terms , because the quantities @xmath124 defined in @xcite are chirally suppressed for all light quarks . to a very good approximation @xmath125 , \ ] ] and @xmath126 is given by the same formula with @xmath127 and @xmath128 .
analogous expressions hold in the down - type quark sector .
the function @xmath129 determines the values of the fermion profiles on the ir brane @xcite .
note that in the approximation used here both @xmath130 and the zero - mode sum in ( [ smsums ] ) are real and do not contribute to the wilson coefficient @xmath116 in ( [ doublelimit ] ) .
the cross section for higgs - boson production in gluon - gluon fusion can now be written as @xmath131 where @xmath132 has been defined in ( [ kappav ] ) , while @xmath133 here @xmath134 and @xmath135 encode the modifications of the higgs - boson couplings to top and bottom quarks with respect to their sm values .
the former quantity is given to high accuracy by @xcite @xmath136 and an analogous expression with @xmath137 and @xmath138 holds for @xmath139 .
the wilson coefficients @xmath140 must be evolved from the high kk scale to the higgs - boson mass scale .
their scale dependence is governed by evolution equations .
if the kk contributions are dominated by the corrections due to the lowest - lying modes , it is practical to integrate out all kk modes at a common scale of order @xmath15 , as already done in ( [ matching ] ) above .
the evolution from this scale down to the weak scale is described by the renormalization - group equations of the sm . for the qcd evolution of the wilson coefficient @xmath141 ,
one obtains the exact result @xcite @xmath142 in practice , the evolution from the kk scale of several tev down to @xmath143 has only a small effect of about 1% on the value of @xmath115 , since no leading logarithms appear in this result . since the operator multiplying the coefficient @xmath116 in ( [ lhgg ] )
is connected by the adler - bell - jackiw anomaly to current operators with vanishing qcd anomalous dimension , it follows that this coefficient is scale independent in qcd , i.e. , @xmath144 .
much of the analysis in the present paper is concerned with the fact that the order in which the limits @xmath86 and @xmath145 in ( [ doublelimit ] ) are taken is not irrelevant , because these two limits do not commute . in our previous work @xcite
, we have evaluated the yukawa couplings in ( [ gdef ] ) mode by mode , taking the regulator @xmath76 to zero after computing the relevant overlap integrals .
we have then numerically evaluated the contributions of the first few kk levels to the higgs - boson production cross section , observing that the sum over modes converges and approaches a limiting value after the summation over several kk levels .
in essence , this approach corresponds to taking the limit @xmath86 first , and hence @xmath146 = \lim_{n\to\infty}\,\sum_{n=1}^{3 + 6n}\,\frac{v g_{nn}^q}{m_{q_n } } \bigg|_{\eta\to 0 } \,.\ ] ] yet we did not attempt to derive an analytical expression for the infinite sum . in the approach taken in @xcite ,
on the other hand , one considers the infinite sum over modes from the very beginning .
this is accomplished by means of completeness relations for the fermion profiles .
the regularization of the higgs profile is taken to zero at the end of the calculation .
it follows that in this approach @xmath147 = \lim_{\eta\to 0}\,\sum_{n=1}^\infty\,\frac{v g_{nn}^q}{m_{q_n } } \,.\ ] ] in the following two sections , we will derive closed analytical expressions for the limiting values in ( [ ours ] ) and ( [ theirs ] ) . in section [ sec : interp ] , we will then discuss the physical interpretation of these results .
we begin with the study of the sum in ( [ theirs ] ) , in which the regulator on the higgs profile is removed after the infinite sum over kk modes has been performed .
the authors of @xcite succeeded to evaluate this sum using the completeness relations ( @xmath63 ) @xmath148 for the fermion profiles .
the analysis relied on a perturbative treatment of the yukawa couplings , which yields the first non - trivial contribution in the expansion in powers of @xmath16 .
a subtlety in this calculation , worth recalling , is that in the perturbative approach the odd fermion profiles @xmath47 vanish on the ir brane . as a result
, the contribution of each individual mode , and hence of any truncated kk sum , vanishes in the limit where the width of the higgs - boson profile is taken to zero . a non - zero result for the @xmath149 correction to ( [ theirs ] )
is obtained only after summing over the infinite tower of kk states . in our notation , the main result of @xcite reads @xmath150 an elegant way to derive a closed expression for the sum ( [ theirs ] ) , valid to all orders in @xmath16 , is to relate it to the 5d fermion propagator of the theory .
we study propagator functions built using the 6-component spinors @xmath151 where the profile functions @xmath62 have been defined in ( [ notat ] ) .
the grand , @xmath152 propagator in the mixed momentum / position representation @xcite is @xmath153 \frac{i}{\rlap{\hspace{0.3mm}/}{p}-m_{q_n } } \left [ \,{\cal q}_l^{(n)\dagger}(t')\,\frac{1+\gamma_5}{2 } + { \cal q}_r^{(n)\dagger}(t')\,\frac{1-\gamma_5}{2 } \right ] , \hspace{4 mm } \end{aligned}\end{aligned}\ ] ] where @xmath154 denotes time ordering . in our notation
, the dirac operator takes the form @xmath155 with the generalized mass matrix defined in ( [ mqdef ] ) . using the bulk eoms ( [ eoms ] ) and the completeness relations ( [ completeness ] ) , it is straightforward to show that @xmath156 since with our exact treatment of yukawa interactions there are no massless zero modes of the fermion fields , the propagator does not exhibit a singularity at @xmath157 .
we can therefore study the special limit @xmath158 without complications .
we obtain @xmath159 , \ ] ] where @xmath160 it now follows from the second relation in ( [ gdef ] ) that the infinite sum in ( [ theirs ] ) , for fixed non - zero @xmath76 , can be expressed as @xcite @xmath161 where @xmath162 .\ ] ] notice that only the off - diagonal blocks of the propagator enter in this result .
the dirac equation ( [ dirac ] ) for the 5d propagator implies the differential equation @xmath163\,\bm{\delta}_{rl}^q(t , t ' ) = \frac{1}{m_{\rm kk}}\,\delta(t - t ' ) \,.\ ] ] we will construct solutions to this equation assuming first that @xmath164 , distinguishing the cases where @xmath165 and @xmath166 . later , these solutions are patched together by means of the jump condition @xmath167 = \frac{1}{m_{\rm kk } } \,.\ ] ] since we have regularized the higgs profile , the boundary conditions on the uv and ir branes take the simple form @xmath168 which follows since the @xmath50-odd profile functions @xmath47 , which sit in the upper components of the right - handed spinors in ( [ notat ] ) , vanish on the branes .
the three equations above determine the matrix function @xmath169 completely . for our purposes , it suffices to construct the solution in the region close to the ir brane , where @xmath170 . for @xmath165 , equation ( [ delta_eom ] )
is solved by the @xmath82-ordered exponential @xmath171\,\bm{\delta}_{rl}^q(1,t ' ) \,,\ ] ] where ( [ bcs ] ) implies that the boundary function @xmath172 has only lower components .
the ordering symbol is required since the matrices @xmath173 with @xmath174 as defined in ( [ thetadef ] ) , do not commute at different @xmath82 values . for coordinates @xmath82 near the ir brane , however , the first term in ( [ intm ] ) can be neglected , and we can approximate @xmath175 \bm{\delta}_{rl}^q(1,t ' ) \,,\ ] ] where the harmless @xmath176 terms can be dropped .
the resulting matrix exponential can then be evaluated in terms of hyperbolic trigonometric functions .
we then lower the value of @xmath82 , implement the jump condition ( [ jump ] ) when @xmath82 crosses @xmath177 , and finally derive an expression for @xmath178 , which still contains two unknown coefficients that are matrices in generation space and functions of @xmath177 .
they can be determined using the boundary condition on the uv brane . for @xmath179
the generalized mass matrix @xmath180 in ( [ mqdef ] ) is diagonal , since the contribution from the higgs profile vanishes .
the first boundary condition in ( [ bcs ] ) therefore implies that @xmath181 as long as @xmath182 , and this equation yields the desired constraints on the propagator function . in ( [ trdef ] )
we only need the off - diagonal blocks of the function @xmath169 for @xmath183 .
we find @xmath184 , \ ] ] where @xmath185 and @xmath186 .
notice that , for the special case where @xmath187 , the result in expression ( [ beauty ] ) is independent of both @xmath82 and @xmath76 .
this means that in ( [ gsum ] ) only the normalization of the higgs profile enters , which is independent of its shape and equal to 1 .
we conclude that the sum in ( [ gsum ] ) is in fact independent of @xmath76 , and hence @xmath188 the first non - trivial term in the taylor expansion agrees with ( [ eq28 ] ) .
note that this result is real , and hence it only contributes to the wilson coefficient @xmath115 .
remarkably , the infinite sum of kk states converges despite of the fact that it is superficially divergent .
this hints to the existence of intricate cancellations between different contributions to the sum . from ( [ doublelimit ] ) and ( [ smsums ] ) , we now obtain @xmath189 \\ & = \sum_{q = u , d}\,\left [ \mbox{tr}\left ( \bm{x}_q^2 - \frac{\bm{x}_q^4}{3 } + \dots \right ) + \mbox{re}\,\varepsilon_q \right ] , \end{aligned}\ ] ] and @xmath190 recall that the correction terms @xmath130 are very small , and to an excellent approximation they are real , see ( [ vareps ] ) .
the wilson coefficient @xmath116 is therefore predicted to be tiny , of order @xmath191 at most .
the contribution to the coefficient @xmath115 is positive and hence yields to an enhancement of the @xmath0 amplitude compared with the sm .
we now consider the sum in ( [ ours ] ) , for which compared to ( [ theirs ] ) the limits @xmath145 and @xmath86 are taken in the opposite order . in this case one
first considers a finite sum over kk modes and derives the relevant yukawa coupling for each mode by regularizing the higgs profile , computing the overlap integral in ( [ gdef ] ) , and then taking the limit @xmath86 .
the relevant yukawa couplings in that limit have been given in ( [ g1mi ] ) .
because the sum in ( [ ours ] ) extends over a finite number of kk levels , it is not possible to use the elegant method of 5d propagators described in the previous section . in order to obtain a closed expression
nevertheless , we adopt the strategy of first finding a solution in a special case , where the bulk eoms and the eigenvalue equation determining the masses of the kk modes can be solved analytically .
we will then argue that , as in the case ( [ satz ] ) , the solution is independent of the bulk mass parameters .
the special case we consider is that of one generation of fermion fields , whose bulk mass parameters vanish : @xmath192 . the eoms for the profile functions are then solved in terms of simple trigonometric functions , which can be evaluated in the limit @xmath193 ( recall that @xmath31 is tiny in the rs model ) .
we find that the kk masses in units of @xmath15 are given by the solutions to the eigenvalue equation @xmath194 where without loss of generality we assume that @xmath195 is real and positive .
this equation can be solved to give @xmath196 \displaystyle\frac{n}{2}\,\pi - x_1 \ , ; & \quad n=2,4,6,\dots , \end{array } \right.\ ] ] where @xmath197 denotes the mass of the zero mode ( the `` sm quark '' ) in units of the kk scale .
the corresponding even profile functions are @xmath198 where the upper ( lower ) signs hold for odd ( even ) values of @xmath41 . inserting these results into ( [ g1mi ] ) and using ( [ xn ] )
, it follows that @xmath199 we can now readily compute the sum over kk modes required in ( [ ours ] ) , with the result that @xmath200 .\ ] ] in the one - generation case , @xmath201 refers to the zero mode , while @xmath202 labels the kk modes .
the first term in the bracket of the final expression arises from the `` sm quark '' , while the sum is over pairs of kk modes belonging to the @xmath203 kk level .
note that for large @xmath9 the individual terms in the sum fall off only like @xmath204 , but each pair combines to a contribution decreasing like @xmath205 .
hence , as long as we sum over complete levels of kk states first , the sum over @xmath9 is convergent , and it is possible to take the limit @xmath145 .
we thus obtain for ( [ ours ] ) in the case of a single fermion generation with vanishing bulk masses @xmath206 which differs from the finite sum in ( [ toysum ] ) by terms of @xmath207 . for one generation
, it is not difficult to solve the eoms numerically also for general @xmath208 . the profile functions and eigenvalue equation in this case involve bessel functions .
we have confirmed in this way that the simple answer ( [ magic ] ) still holds in this more general case . given the similarity of the above result with equation ( [ satz ] ) derived in the previous section
, we conjecture that in the case of three generations @xmath209 once again the answer is real , such that only @xmath115 receives a contribution . combining ( [ doublelimit ] ) , ( [ smsums ] ) , and ( [ conjecture ] ) , we then find @xmath210 \\ & = \sum_{q = u , d } \left [ \mbox{tr}\left ( - \bm{x}_q^2 + \frac{7\bm{x}_q^4}{3 } + \dots \right ) + \mbox{re}\,\varepsilon_q \right ] , \end{aligned}\ ] ] while @xmath116 is still given by ( [ c5res ] ) .
this time the contribution to @xmath115 is negative and thus yields a suppression of the @xmath0 amplitude compared with the sm .
we have checked that this formula indeed reproduces our numerical results obtained in @xcite . incidentally , the first term in the expansion in powers of @xmath211 has the opposite sign from the result ( [ funny ] ) , which was first obtained in @xcite .
the conjecture that equation ( [ magic ] ) can be generalized to the case of three generations will be proved elsewhere .
we note , however , that symmetry arguments help us to constrain the form of the answer .
it is plausible to assume that , after the limit @xmath145 has been taken , the result for the sum ( [ ours ] ) should depend only on the fundamental parameters of the underlying 5d theory , i.e. , the bulk mass parameters and yukawa couplings .
all reference to the properties of individual states , such as their masses or profile functions , should disappear .
we express this assertion by writing @xmath212 .
the dependence of this quantity on the fundamental parameters can be constrained using symmetry arguments .
the eoms ( [ eoms ] ) and boundary conditions ( [ irbcs ] ) are valid in an arbitrary basis , in which the bulk mass matrices @xmath80 are not necessarily diagonal .
these relations are invariant under a set of three global symmetries .
the first one is a symmetry under the exchange of @xmath53 doublets and singlets along with @xmath213 .
in addition , there are two symmetries related to unitary transformations of the yukawa and bulk mass matrices . when combined with the fact that in the one - generation case the result is found to be independent of the bulk mass parameters ,
these symmetries imply that the quadratic term in @xmath214 must indeed be of the simple form shown in ( [ conjecture ] ) .
using symmetry considerations alone , we can however not exclude a dependence on the parameters @xmath80 starting at @xmath215 , provided that it cancels in the case of one generation .
the discussion of the previous two sections shows that the calculations presented in @xcite and @xcite are free of mistakes .
the results for the effective @xmath13 couplings derived in these two papers differ , in magnitude and sign , because both groups considered different orders in which the limits @xmath86 and @xmath145 in ( [ doublelimit ] ) were taken , and the two limits do not commute . hence , the question poses itself which of the orders of limits is the more reasonable one from a physical point of view .
we will address this question in section [ sec : cutoff ] .
a second , equally puzzling question arises once we realize that the non - commutativity of the limits @xmath86 and @xmath145 implies that the origin of the discrepancy between the results ( [ satz ] ) and ( [ conjecture ] ) must be due to contributions from kk modes with large @xmath41 and hence very heavy masses , and related to the behavior of the profile functions for such modes very close to the ir brane .
the fact that in both approaches one finds that the infinite sum over kk modes converges , signaling that heavy modes decouple , raises the question : how is it possible that modes with very heavy masses give an @xmath61 contribution to the sum , which is large enough to change the sign of the answer ? in order to understand the physics behind this effect , it is useful to study a toy model that can be solved exactly also for finite @xmath76 . to this end
, we consider again the case of a single fermion generation and vanishing bulk mass parameters , which proved so useful for our analysis in section [ sec : finitesum ] . to obtain analytic solutions to the bulk eoms for the fermion profiles we adopt a particularly simple form for the regularized higgs profile , namely a box of width @xmath76 : @xmath216 0 \,\ , ; & \quad \mbox{for~~ } \epsilon\le t < 1-\eta \ , . \end{array } \right.\ ] ] moreover , since for @xmath217 the wave functions are non - singular near the uv brane , it is possible to set @xmath218 . in appendix
[ sec : toy ] we derive the explicit forms of the bulk profiles for this toy model and the eigenvalue equation that determines the masses of the kk modes .
we then present a formula for the sum @xmath219 entering ( [ doublelimit ] ) .
it turns out that the nature of the solution differs depending on whether @xmath220 or @xmath221 , where we have introduced the abbreviation @xmath222 .
the latter condition corresponds to masses @xmath223 , where @xmath224 is of order the weak scale , since we as usual assume @xmath225 in the anarchic rs model ( recall that we can choose @xmath195 real and positive ) .
the appearance of the scale @xmath226 , which for very small @xmath76 lies far above the tev scale , will be of crucial importance to the resolution of the puzzle .
we find that the mass eigenvalues in our toy model are determined by the conditions @xmath227 = \left\ { \begin{array}{ll } \displaystyle \pm\sqrt{\frac{z\mp x_n}{z\pm x_n } } \tanh\big[\sqrt{z^2-x_n^2}\,\eta\big ] \ , ; & \quad \mbox{for~~ } x_n < z \ , , \\[5 mm ] \displaystyle - \sqrt{\frac{x_n\mp z}{x_n\pm z } } \tan\big[\sqrt{x_n^2-z^2}\,\eta\big ] \ , ; & \quad \mbox{for~~ } x_n > z \ , , \end{array } \right.\ ] ] where as before the upper ( lower ) signs hold for odd ( even ) values of @xmath41 .
these equations can easily be solved numerically .
the sum over the kk contributions to the effective @xmath13 couplings is then obtained by evaluating the sum @xmath228
\right ) \right ] , \\
n_n^{-2 } & = 1 + \frac{z}{z^2-x_n^2 } \left [ \pm\frac12 \sin\big[2x_n(1-\eta)\big ] - \eta\left ( z\mp x_n\cos\big[2x_n(1-\eta)\big ] \right ) \right ] .
\end{aligned}\ ] ] in the left plot in figure [ fig : surprise ] , we show numerical results for the sum ( [ toysum2 ] ) as a function of the number of kk states included , for four different values of @xmath76 . for the purpose of illustration
we take @xmath229 .
we observe that for low values of @xmath118 the sum quickly converges toward a value close to the result ( [ ours ] ) , which equals 0.701 in the present case .
but then there is an intermediate region , roughly given by the range @xmath230 , in which the value of the sum changes by an @xmath61 amount .
after this transition region , the sum converges to the value corresponding to the result ( [ satz ] ) , which equals 1.082 for our choice of @xmath231 . in order to understand the origin of the three regions in particular the appearance of the intermediate region , in which the sum grows by an @xmath61 amount despite of the fact that the corresponding kk masses are extremely heavy we show in the right plot the values of the yukawa couplings @xmath68 in our toy model ( for the case where @xmath232 ) . for low values of @xmath41 ,
the couplings for each pair of kk states in the same kk level have equal values and opposite signs . in this case , like in ( [ toysum ] ) , the contribution from the @xmath203 kk level is @xmath233 in this region the sum over @xmath9 is convergent and dominated by the contributions from the first few kk levels .
the situation changes drastically in the intermediate region , where the average coupling @xmath234 in each kk level , shown by the dark points in the right plot , no longer vanishes .
there is thus a range of logarithmic growth of the sum , over which its contribution can be estimated as @xmath235 where @xmath236 denotes the mean value of @xmath234 in the interval @xmath237 $ ] , in which the average coupling in each kk level departs from zero . for even larger values ,
the average coupling decreases quickly , giving rise to very small contributions to the sum .
physically , the intermediate region arises because , for fixed @xmath76 , there exists a minimum kk mass @xmath238 beyond which the kk profile functions begin to penetrate the box modeling the higgs - boson profile .
when this happens , the cancellation of the yukawa couplings of kk modes within one kk level is no longer operative . only for yet much higher kk levels , when the profiles exhibit a large number of oscillations within the box , the couplings average out to zero and hence decrease with increasing @xmath9 . while this behavior might seem strange at first sight , one should remember that by naive dimensional analysis the sum @xmath219 is logarithmically divergent for large @xmath118 .
it is only due to subtle cancellations that it converges ; however , when the profiles penetrate the box these cancellations no longer occur , and the generic logarithmic growth arises .
in order to decide which of the two calculations presented in @xcite and @xcite is physically more meaningful , we recall the importance of using a consistent uv regularization scheme when evaluating the gluon - gluon fusion amplitude . this is true in the sm , and even more so in its 5d extensions . here
we will study two different regularization schemes : dimensional regularization and the use of a hard momentum cutoff . while a dimensional regulator is particular convenient in that it preserves gauge and lorentz invariance ,
the second option is also a natural choice in the present case .
this is because the rs model must be considered as an effective theory below the planck scale , which requires a uv completion incorporating the effects of quantum gravity .
a peculiar feature of warped extra - dimension models is that the effective uv cutoff depends on where the theory is probed along the extra dimension @xcite ( see also @xcite for related works ) .
the physical reason is that due to warping the fundamental length and energy scales change along the extra dimension .
more specifically , the effective cutoff scale at the position @xmath82 in the extra dimension is of the order of the warped planck scale @xmath239 the cutoff should be sufficiently high that at least a small number of kk modes have masses below @xmath240 , and hence @xmath241 or so .
otherwise incalculable `` threshold corrections '' of order @xmath242 become important and take away the predictive power of the model .
we stress that imposing a uv cutoff is crucial in order for the rs model to provide a viable solution to the hierarchy problem .
quantum corrections to the higgs potential in the rs model exhibit an even stronger divergence than in the sm @xcite . however , as long as the higgs sector is localized near the ir brane ( i.e. , in the vicinity of @xmath70 ) , the effective cutoff @xmath243 is of the order of several tev , and hence the ( big ) hierarchy problem is solved .
the question of how to introduce such a cutoff in practical one - loop calculations is far from trivial .
however , at large loop momentum ( of order several times @xmath15 ) external momenta can be neglected , and hence there is a single 4d ( euclidean ) loop momentum @xmath244 on which the cutoff should be imposed .
we propose to associate a @xmath82-dependent cutoff with every vertex of a feynman diagram .
this can be thought of as modeling the effect of a form factor , which accounts for the impact of quantum gravity on energy scales above the effective planck scale at that point . in general
, the @xmath245 coordinates of the vertices are integrated over the entire bulk ( @xmath246 ) , and hence the cutoff values vary between the tev scale and the fundamental planck scale .
this is indeed an important effect , which makes gauge - coupling unification possible in warped extra - dimension models @xcite .
the situation simplifies considerably for one - loop diagrams containing vertices with higgs bosons , such as the one in figure [ fig : graph ] . denoting the coordinate of the two gluons by @xmath247 and @xmath248 and that of the higgs boson by @xmath249 , the fact that @xmath250 ensures that the momentum cutoff on the 4d loop integral is @xmath251 the same mechanism guarantees that the hierarchy problem is solved in rs models .
it is well known that evaluating the triangle diagram in figure [ fig : graph ] in the 4d theory requires a consistent , gauge - invariant regulator such as dimensional regularization , the reason being that the integral is superficially uv divergent .
if one were to evaluate it in @xmath252 dimensions , then a momentum - independent term proportional to @xmath253 would appear , violating gauge invariance . in order to regularize the loop integral
we thus evaluate it in @xmath254 dimensions , with @xmath255 , so as to regularize uv divergences .
we then find that the sum in ( [ sigdef ] ) gets modified to @xmath256 where @xmath257 is the regularization scale .
for very large masses @xmath258 , the dimensional regulator gives rise to a suppression , which renders the sum over kk modes convergent even for arbitrary @xmath61 yukawa couplings .
the limits @xmath145 and @xmath86 can therefore be taken without encountering any ambiguities .
the contribution from the dangerous intermediate region from super - massive kk modes with @xmath238 in figure [ fig : surprise ] , which previously gave rise to an unsuppressed contribution of the form ( [ loggrowth ] ) , now receives an extra suppression factor @xmath259 , and vanishes when one takes the limit @xmath86 ( at fixed @xmath260 ) .
the infinite sum then coincides with the result ( [ ours ] ) up to harmless @xmath261 corrections . in the dimensional regularization scheme ,
gauge invariance is manifest in the 4d theory . since the dimensional regulator also regularizes the infinite kk sum , we are guaranteed that the 5d theory remains gauge invariant , too . on the other hand ,
dimensional regularization is perhaps not the most intuitive way in which to introduce a uv cutoff . as an alternative
, we will therefore rephrase the discussion in a regularization scheme based on using the hard momentum cutoff given in ( [ eq : cutoff ] ) . in order to ensure 4d gauge invariance in this case
, we first extract two powers of the external gluon momenta by taking appropriate derivatives , after which the remaining loop integral is superficially convergent . introducing the uv cutoff on this integral , and neglecting the higgs - boson mass compared with @xmath262
, we obtain @xmath263 for small masses @xmath264 this reduces to the simple expression @xmath265 used in the sum ( [ sigdef ] ) .
for very large masses @xmath266 , on the other hand , the uv cutoff gives rise to a strong suppression proportional to @xmath267 , implying that such heavy kk modes decouple rapidly .
it follows that , due to physical reasons , the sum over kk modes in ( [ sigdef ] ) is effectively truncated once the kk masses exceed the scale @xmath240 . in the rs model with a brane - localized higgs sector ,
the scale @xmath226 at which the high - mass kk modes start to contribute a positive contribution to @xmath219 with logarithmic growth is parametrically much larger than the effective cutoff scale @xmath240 .
it is then appropriate to truncate the sum at a value @xmath268 corresponding to kk masses much smaller than @xmath226 .
it follows that @xmath269 where the truncation error has the form of a threshold correction , which is always present in effective - theory calculations . we can summarize the above discussion by emphasizing the subtle fact that , in order to obtain the correct answer for the gluon - gluon fusion cross section in the rs model , it is essential to employ a consistent uv regularization scheme when evaluating the loop integral , despite of the fact that this integral is convergent .
when this is done , the convergence of the infinite sum appearing in ( [ doublelimit ] ) is improved in such a way that the order in which the two limits @xmath145 and @xmath86 are taken becomes irrelevant .
specifically , in the two regularization schemes we have considered in our analysis , we find @xmath270 \displaystyle \sum_{n=1}^\infty\ , \frac{v g_{nn}^q}{m_{q_n } } \left ( \frac{\lambda_{\rm tev}^2}{\lambda_{\rm tev}^2+m_{q_n}^2 } \right)^2 \bigg|_{\eta=0 } \!= \sigma_q^{\rm ( cghnp ) } + { \cal o}\bigg ( \frac{n_{\rm kk}^{\rm max}\,v^2}{\lambda_{\rm tev}^2 } \bigg ) \ , .
\end{array } \right.\ ] ] in both schemes the infinite sums are superficially convergent .
the first line refers to the dimensional regularization scheme , while the second line corresponds to employing a hard momentum cutoff .
it would be instructive to rederive these results by means of a properly uv - regularized 5d loop calculation , rather than by regularizing the sum over kk modes , as we have done above .
in such a computation , in which there would be no reference to individual kk states , one should find that the correction to the @xmath13 amplitude is of the form ( [ twoschemes ] ) once the regulator on the higgs - boson profile is removed ( @xmath86 ) .
for the simplified case of one generation , we have convinced ourselves that this is indeed the case .
details of this rather cumbersome analysis , as well as its extension to three generations , will be presented elsewhere .
the terms suppressed by a power of the uv cutoff , which appear on the right - hand side of ( [ eq : threshold ] ) , parameterize the difference between the asymptotic value @xmath271 of the infinite sum and the sum over a finite number of kk modes .
more generally , such threshold terms may also arise from the yet unknown effects of the uv completion of the rs model . from a low - energy perspective , the only requirement on such a completion that is relevant to us is that it must cure the hierarchy problem , by taming loop momenta exceeding the fundamental scale of quantum gravity . as long as this is the case
, the gluon - gluon fusion amplitude will also be regularized in the way discussed above . in the context of a specific uv completion
, the threshold effects could be modeled at low energies by means of a brane - localized effective @xmath272 operator , suppressed by @xmath273 .
the additional factor @xmath274 reflects the high multiplicity of degrees of freedom in the low - energy effective theory . for quite generic reasons , the coefficient of this operator must contain the loop factor @xmath275 factored out in ( [ lhgg ] ) , even in cases where the uv completion of the rs model is strongly coupled .
the reason is that the on - shell external gluons couple proportional to their qcd charges , and that any new heavy state that couples to the higgs boson must be color neutral , so it can not have a tree - level coupling to gluons .
hence , a generic uv completion will indeed give rise to a threshold correction of the form shown in ( [ eq : threshold ] ) . the difference between the two sums ( [ ours ] ) and ( [ theirs ] ) , or the corresponding expressions ( [ funny ] ) and ( [ central_result ] ) , can however _ not _ be attributed to such a brane - localized threshold term .
rather , it is related to the question whether a physical cutoff and a sensible uv completion are present at all .
we now study the implications of our results for higgs - boson production in gluon - gluon fusion at the lhc .
the master formula for the cross section has been given in ( [ sigmars ] ) , where the expressions for @xmath132 and @xmath276 can be found in ( [ kappav ] ) and ( [ kappas ] ) , respectively , while to an excellent approximation we can set @xmath277 . for the calculation of the wilson coefficient @xmath115 we use our central result ( [ central_result ] ) and assume that the parametrically - suppressed threshold effects appearing in ( [ eq : threshold ] ) can be neglected .
note that the trace over a function @xmath278 of the matrix defined in ( [ xqdef ] ) is determined by the non - negative , real square roots @xmath279 of the eigenvalues of the hermitian matrices @xmath280 , i.e.@xmath281 since in the rs model with anarchic 5d yukawa couplings these matrices are structureless , the above result is proportional to the rank of the yukawa matrices , which in the case of the minimal rs scenario is equal to the number of fermion generations .
in other words , the kk towers of all six quarks give comparable contributions to the effective @xmath13 vertex , irrespective of the mass of the corresponding sm fermion .
we emphasize that this feature is not present in many other extra - dimensional extensions of the sm .
for instance , in models based on universal extra dimensions the 5d yukawa couplings are hierarchical , like in the sm , and hence the higgs - boson couplings to light fermions and their kk excitations are strongly suppressed @xcite .
more interestingly , in several gauge - higgs unification models , in which the higgs appears as a pseudo - goldstone boson , one finds that the contribution to the @xmath13 amplitude from the kk excitations of the sm quarks exactly cancels the dominant effect due to the corrections to the yukawa couplings of the sm quarks , leaving only chirally - suppressed corrections , which are very small for all quarks other than the top quark @xcite .
hence , in these new - physics scenarios only the top quark and its heavy partners contribute to the effective @xmath13 couplings , while higgs - boson production is independent of the masses and couplings of the kk excitations of light sm quarks . by randomly generating a large set of 5d yukawa matrices , which are required to satisfy @xmath282 and to correctly reproduce the wolfenstein parameters @xmath283 and @xmath76 of the unitarity triangle
, we have found that to a good approximation the average result for the trace appearing in ( [ trace ] ) can be parametrized as defined in ( [ xqdef ] ) instead of the original 5d yukawa matrices @xmath60 . in practice , it turns out to be immaterial whether one considers @xmath60 or @xmath284 as random complex matrices . ]
@xmath285 and @xmath286 .
likewise , the parameter @xmath287 entering ( [ kappas ] ) can be approximated as @xmath288 where to an excellent approximation the contribution proportional to @xmath289 cancels in the sum @xmath290 appearing in the numerator of the expression for @xmath276 in ( [ kappas ] ) .
furthermore , the very small effects due to @xmath139 and @xmath291 can be neglected for all practical purposes . in the left plot in figure [ fig : rh ]
, we show our results for the higgs - boson production cross section in gluon - gluon fusion relative to the sm cross section , @xmath292 as a function of the mass of the lightest kk gluon state , @xmath293 .
we use the lightest kk gluon mass as a reference , because its value is a model - independent prediction of the rs scenario .
the masses of the lightest kk fermions have very similar values but depend to some extent on the bulk mass parameters . here and below we employ a higgs - boson mass of 125gev .
the solid red , green , and light blue lines show the approximate results obtained from ( [ param1 ] ) and ( [ param2 ] ) employing the values @xmath294 , 1.5 , and 0.5 , respectively .
the underlaid density plots indicate the distribution of the predictions for a large number of anarchic yukawa matrices @xmath60 , subject only to the constraint that their elements are smaller in magnitude than a given value @xmath295 .
the fact that the regions of highest density in the scatter plots nicely reproduce the results obtained directly from ( [ param1 ] ) and ( [ param2 ] ) shows that the requirement to reproduce the correct values of @xmath283 and @xmath76 does not play an important role numerically .
the contour plot displayed on the right in the figure gives a two - dimensional representation for the cross section as a function of @xmath296 and @xmath295 , obtained by employing again ( [ param1 ] ) and ( [ param2 ] ) .
we observe from figure [ fig : rh ] that @xmath297 is strictly below 1 and decreases ( increases ) with increasing @xmath295 ( kk scale ) . in other words , the minimal rs model with a brane - localized higgs sector
predicts a depletion of @xmath298 relative to the sm . in the region where @xmath299 is a suitable expansion parameter
, we obtain the approximate result @xmath300 where the constant term in parenthesis is due to the effect of @xmath132 . for not too small yukawa couplings
this observable is dominated by the effects of kk quark loops . given the strong dependence of the ratio @xmath301 on @xmath295
, we find that for @xmath302tev and yukawa couplings close to the perturbativity bound @xmath303 @xcite , the new - physics contributions to the higgs - boson production cross section in gluon - gluon fusion can become so large that they completely cancel the sm contribution .
in fact , the sensitivity of @xmath301 to the overall size of the 5d yukawa couplings is even more pronounced than the one arising in the case of dipole - operator transitions such as @xmath304 @xcite . while the latter contributions also scale with @xmath305 , unlike @xmath301 they are ( at the one - loop level ) insensitive to the multiplicity of states in the fermionic sector of the rs model under consideration .
this feature underscores our assertion ( made in the introduction ) that precision measurements of the higgs - boson properties furnish a superb tool for illuminating the quantum structure of electroweak interactions in rs scenarios . in the minimal rs model considered here
, constraints from electroweak precision observables @xcite and flavor physics @xcite require that the lightest kk excitations of sm particles must have masses in the 10tev range , which puts them outside of the reach for production at the lhc .
figure [ fig : rh ] shows that even in this case there can be significant virtual effects of kk particles on the higgs - boson production cross section , provided that the 5d yukawa couplings are not too small .
the bounds from electroweak precision measurements , in particular , can be relaxed in several ways .
for instance , `` little rs models '' , in which the size @xmath306 of the extra dimension is reduced @xcite , would only have a minor impact on our analysis .
other extensions , such as models with a custodial @xmath307 gauge symmetry in the bulk @xcite , might however give rise to a rather different higgs - boson phenomenology @xcite .
the announcements of the first direct hints for a higgs - boson signal by the lhc and tevatron experiments open up a new chapter in particle physics . although the significance of the various measurements is not yet sufficient to preclude the possibility of statistical fluctuations accounting for the observed effects , hopes are high that , with atlas and cms accumulating more data , a higgs boson will be discovered ( or excluded ) by the end of this year .
this discovery would not only mark the birth of the hierarchy problem , but it will also reshape some of the fundamental questions of our field . in particular , the focus of large parts of the lhc physics program will shift towards determining the higgs - boson properties as accurately as possible , with the ultimate goal of probing possible deviations from the sm expectations . in close analogy to flavor physics ,
precision higgs - boson physics represents a powerful way to investigate the dynamics of electroweak symmetry breaking at the quantum level .
in fact , the couplings of the higgs boson to a pair of gluons and photons vanish at tree level in the sm , but are induced by the exchange of virtual top quarks and @xmath105 bosons at one - loop order .
the effective @xmath13 and @xmath308 couplings thus offer a distinctive window to physics beyond the sm , where new heavy particles can propagate in the loops , thereby potentially affecting both the production cross section and the decay rates of the higgs boson .
the main goal of this article was to perform an analytic calculation of the impact of kk fermions on the production of a sm - like higgs boson in the minimal rs scenario featuring an @xmath36 bulk gauge symmetry and a brane - localized higgs sector .
in particular , we have revisited the gluon - gluon fusion process , for which two independent calculations previously found contradictory results .
a significant suppression of the cross section was reported in @xcite , while in @xcite an effect of similar magnitude but opposite sign was obtained .
as we have shown , the discrepancy is not due to a simple computational mistake . on a technical level , it can be traced to the fact that , in order to make the overlap integrals of the wave functions of the brane - localized higgs - boson with the bulk fermions mathematically well defined , a regularization of the higgs profile is unavoidable in an intermediate step of the calculation .
it is achieved by smearing out the profile over a finite width @xmath76 , thereby moving the higgs boson slightly into the bulk . given that the yukawa couplings depend on @xmath76 , and that the calculations in @xcite both involve the summation over an infinite number @xmath118 of kk levels , the question arises whether the result for the @xmath0 amplitude might depend on the order in which the limits @xmath86 and @xmath145 are taken .
we have demonstrated that the order in which these limits are performed indeed explains the aforementioned discrepancy .
while the results of @xcite are reproduced if one first takes the limit of the higgs regulator to zero and then sums over the infinite tower of kk states , the findings of @xcite correspond to the reversed order of taking the two limits .
we have presented for the first time closed analytical expressions for both results in terms of the fundamental parameters of the rs model , valid to all orders in the ratio @xmath16 of the higgs vacuum expectation value and the kk mass scale .
we have then pointed out that the non - commutativity of the limits @xmath86 and @xmath145 is due to a hidden uv sensitivity of the infinite kk sum , and illustrated this phenomenon by means of a simple toy model , which can be solved exactly .
this computation highlights that for fixed @xmath76 there exists a region , starting at around @xmath309 , beyond which the kk - fermion wave functions resolve the width of the higgs profile . as a result ,
the subtle cancellation of the yukawa couplings of kk modes within each kk level , which is instrumental for the convergence of the kk sum , is spoiled
. thus logarithmic growth of the @xmath0 amplitude kicks in , which continues until the kk fermion profiles exhibit a large number of oscillations within the region where the higgs profile is localized .
excitations with kk numbers higher than @xmath310 quickly decouple , rendering the infinite kk sum ultimately convergent . the question which of the two calculations of the gluon - gluon fusion amplitude gives the correct physics result
is resolved by introducing a proper uv regularization scheme .
specifically , we have studied the problem using dimensional regularization and using a hard momentum cutoff , which may be identified with the inherent uv cutoff of rs models set by the warped planck scale . in both cases ,
the regulator improves the convergence of the kk sum and removes the unphysical , previously unsuppressed contributions from super - heavy kk modes . the regularized sum over kk states
yields the result of @xcite , up to possible threshold corrections suppressed by inverse powers of the effective uv cutoff .
this shows that higgs production is an uv - insensitive process in the rs model and can be calculated unambiguously once a proper regularization is employed .
it would be worthwhile to reproduce our main result ( [ central_result ] ) for the induced @xmath13 vertex by performing an actual 5d loop calculation rather than an infinite kk sum .
this is left for future work .
we have finally studied the numerical impact of our results for higgs - boson production in gluon - gluon fusion , thereby extending our previous , more phenomenologically oriented analyses of higgs physics in rs models @xcite .
we have reemphasized the important point that , regardless of the mass of the corresponding sm fermion , in warped extra - dimensional models the loop - induced couplings of the higgs field to gauge bosons receive similar contributions from the kk towers of each fermion state , and that these contributions scale with the square of the 5d yukawa couplings . assuming a similar kk mass scale , rs scenarios hence predict much stronger effects than many other extra - dimensional extensions of the sm , such as universal extra dimensions and models in which the higgs emerges as a pseudo - goldstone boson . while we believe this fact to be true in general , it is not unlikely that the structure of the corrections depends on the precise realization of the scalar sector .
exploring the associated model dependence in more detail is left for future work . in order to probe the large effects predicted by our calculations
, it would be particularly useful to measure the cross sections for higgs - boson production in both the gluon - gluon and vector - boson fusion channels .
a combination of theses two measurements would allow for a clean extraction of the @xmath0 amplitude , and hence represents a unique way to study the yukawa sector of rs theories .
we are grateful to kaustubh agashe , aleksandr azatov , adrian carmona , lisa randall , manuel perez - victoria , edurado ponton , jose santiago , and fabio zwirner for useful discussions .
the research of s.c .
is supported by the dfg cluster of excellence _ origin and structure of the universe_. the research of f.g .
is supported by the swiss national foundation under contract snf 200020 - 126632 .
the research of m.n . is supported by the advanced grant eft4lhc of the european research council ( erc ) , grant 05h09ume of the german federal ministry for education and research ( bmbf ) , and the rhineland - palatinate research center _ elementary forces and mathematical foundations_. fermilab is operated by fermi research alliance , llc under contract no .
de - ac02 - 07ch11359 with the u.s .
department of energy .
in the case where all bulk mass parameters @xmath311 are set to zero , and where in addition we set @xmath218 ( which in this case is unproblematic ) , the eoms ( [ eoms ] ) can be solved exactly if we work with a sufficiently simple regularized higgs profile .
this provides a nice test case for our general results and conclusions . in order to obtain analytic expressions for the wave functions of the individual kk modes
, we consider the higgs profile ( [ hbox ] ) and furthermore assume that there is a single fermion generation with real and positive yukawa coupling @xmath195 . the generalized mass matrix in ( [ mqdef ] )
is then given by @xmath312 \hspace{3.0 cm } 0 \ , ; & \text{for \ ; } \epsilon \leq t < 1-\eta \ , .
\end{cases}\ ] ] the wave functions @xmath62 are now 2-component objects . in the solution of the eoms
we must distinguish two cases .
for @xmath220 , we find @xmath313 and @xmath314 \\ \mp r_2\,\sinh\big[\sqrt{z^2-x_n^2}\,(1-t)\big ] \\ r_2\,\sinh\big[\sqrt{z^2-x_n^2}\,(1-t)\big ] \\
\pm r_1 \cosh\big[\sqrt{z^2-x_n^2}\,(1-t)\big ] \end{array } \right ) \ , ; \quad t\ge 1-\eta \,,\ ] ] where the coefficients @xmath315 are given by @xmath316}{\cosh\big[\sqrt{z^2-x_n^2}\,\eta\big ] } \ , , \qquad r_2 = \frac{\sin\big[x_n(1-\eta)\big]}{\sinh\big[\sqrt{z^2-x_n^2}\,\eta\big ] } \,.\ ] ] for @xmath221 , we find instead @xmath317
\\ \mp r_2\,\sin\big[\sqrt{x_n^2-z^2}\,(1-t)\big ] \\ r_2\,\sin\big[\sqrt{x_n^2-z^2}\,(1-t)\big ] \\
\pm r_1 \cos\big[\sqrt{x_n^2-z^2}\,(1-t)\big ] \end{array } \right ) \ , ; \quad t\ge 1-\eta \,,\ ] ] where @xmath316}{\cos\big[\sqrt{x_n^2-z^2}\,\eta\big ] } \ , , \qquad r_2 = \frac{\sin\big[x_n(1-\eta)\big]}{\sin\big[\sqrt{x_n^2-z^2}\,\eta\big ] } \,.\ ] ] the eigenvalues @xmath318 are determined by matching the profile functions at @xmath83 .
this leads to the conditions shown in ( [ eigenvals ] ) . evaluating the normalization constraint
, we obtain the expression for @xmath319 given in the second line of ( [ toysum2 ] ) . using these results , it is not difficult to calculate the relevant overlap integrals with the higgs profile . in this way
we arrive at the result for the sum @xmath219 shown in the first line of ( [ toysum2 ] ) .
y. grossman and m. neubert , phys .
b * 474 * , 361 ( 2000 ) [ hep - ph/9912408 ] .
t. gherghetta and a. pomarol , nucl .
b * 586 * , 141 ( 2000 ) [ hep - ph/0003129 ] .
s. j. huber and q. shafi , phys .
b * 498 * , 256 ( 2001 ) [ hep - ph/0010195 ] .
k. agashe , g. perez and a. soni , phys . rev .
lett . * 93 * , 201804 ( 2004 ) [ hep - ph/0406101 ] .
k. agashe , g. perez and a. soni , phys .
d * 71 * , 016002 ( 2005 ) [ hep - ph/0408134 ] .
h. davoudiasl , j. l. hewett and t. g. rizzo , phys .
d * 68 * , 045002 ( 2003 ) [ hep - ph/0212279 ] .
m. s. carena , e. ponton , t. m. p. tait and c. e. m. wagner , phys .
d * 67 * , 096006 ( 2003 ) [ hep - ph/0212307 ] .
k. agashe , a. delgado , m. j. may and r. sundrum , jhep * 0308 * , 050 ( 2003 ) [ hep - ph/0308036 ] .
k. agashe , r. contino , l. da rold and a. pomarol , phys .
b * 641 * , 62 ( 2006 ) [ hep - ph/0605341 ] .
m. bauer , r. malm and m. neubert , phys .
* 108 * , 081603 ( 2012 ) [ arxiv:1110.0471 [ hep - ph ] ] .
r. contino , y. nomura and a. pomarol , nucl .
b * 671 * , 148 ( 2003 ) [ hep - ph/0306259 ] .
k. agashe , r. contino and a. pomarol , nucl .
b * 719 * , 165 ( 2005 ) [ hep - ph/0412089 ] .
s. casagrande , f. goertz , u. haisch , m. neubert and t. pfoh , jhep * 1009 * , 014 ( 2010 ) [ arxiv:1005.4315 [ hep - ph ] ] .
a. azatov , m. toharia and l. zhu , phys .
d * 82 * , 056004 ( 2010 ) [ arxiv:1006.5939 [ hep - ph ] ] .
h. davoudiasl , j. l. hewett and t. g. rizzo , phys .
b * 473 * , 43 ( 2000 ) [ hep - ph/9911262 ] .
s. j. huber , nucl .
b * 666 * , 269 ( 2003 ) [ hep - ph/0303183 ] . c. csaki , a. falkowski and a. weiler , jhep * 0809 * , 008 ( 2008 ) [ arxiv:0804.1954 [ hep - ph ] ] . s. casagrande , f. goertz , u. haisch , m. neubert and t. pfoh , jhep * 0810 * , 094 ( 2008 ) [ arxiv:0807.4937 [ hep - ph ] ] .
m. blanke , a. j. buras , b. duling , s. gori and a. weiler , jhep * 0903 * , 001 ( 2009 ) [ arxiv:0809.1073 [ hep - ph ] ] .
a. azatov , m. toharia and l. zhu , phys .
d * 80 * , 035016 ( 2009 ) [ arxiv:0906.1990 [ hep - ph ] ] . c. bouchart and g. moreau , phys .
d * 80 * , 095022 ( 2009 ) [ arxiv:0909.4812 [ hep - ph ] ] . m. beneke and m. neubert , nucl .
b * 651 * , 225 ( 2003 ) [ hep - ph/0210085 ] .
a. djouadi , phys .
rept . *
459 * , 1 ( 2008 ) [ hep - ph/0503173 ] .
f. goertz , u. haisch and m. neubert , phys .
b * 713 * , 23 ( 2012 ) [ arxiv:1112.5099 [ hep - ph ] ] .
t. inami , t. kubota and y. okada , z. phys .
c * 18 * , 69 ( 1983 ) .
v. ahrens , t. becher , m. neubert and l. l. yang , eur .
j. c * 62 * , 333 ( 2009 ) [ arxiv:0809.4283 [ hep - ph ] ] .
m. puchwein and z. kunszt , annals phys .
* 311 * , 288 ( 2004 ) [ hep - th/0309069 ] .
m. s. carena , a. delgado , e. ponton , t. m. p. tait and c. e. m. wagner , phys .
d * 71 * , 015010 ( 2005 ) [ hep - ph/0410344 ] .
a. falkowski , phys .
d * 77 * , 055018 ( 2008 ) [ arxiv:0711.0828 [ hep - ph ] ] .
l. randall and m. d. schwartz , jhep * 0111 * , 003 ( 2001 ) [ hep - th/0108114 ] .
a. pomarol , phys .
lett . * 85 * , 4004 ( 2000 ) [ hep - ph/0005293 ] . k. -w .
choi , h. d. kim and i. -w .
kim , jhep * 0211 * , 033 ( 2002 ) [ hep - ph/0202257 ] .
w. d. goldberger and i. z. rothstein , phys .
lett . * 89 * , 131601 ( 2002 ) [ hep - th/0204160 ] .
k. agashe , a. delgado and r. sundrum , nucl .
b * 643 * , 172 ( 2002 ) [ hep - ph/0206099 ] .
f. j. petriello , jhep * 0205 * , 003 ( 2002 ) [ hep - ph/0204067 ] .
a. azatov and j. galloway , phys .
d * 85 * , 055013 ( 2012 ) [ arxiv:1110.5646 [ hep - ph ] ] .
k. agashe , a. azatov and l. zhu , phys .
d * 79 * , 056006 ( 2009 ) [ arxiv:0810.1016 [ hep - ph ] ] .
m. s. carena , a. delgado , e. ponton , t. m. p. tait and c. e. m. wagner , phys .
d * 68 * , 035010 ( 2003 ) [ hep - ph/0305188 ] .
h. davoudiasl , g. perez and a. soni , phys .
b * 665 * , 67 ( 2008 ) [ arxiv:0802.0203 [ hep - ph ] ] . | measurements of the higgs - boson production cross section at the lhc are an important tool for studying electroweak symmetry breaking at the quantum level , since the main production mechanism @xmath0 is loop - suppressed in the standard model ( sm ) .
higgs production in extra - dimensional extensions of the sm is sensitive to the kaluza - klein ( kk ) excitations of the quarks , which can be exchanged as virtual particles in the loop . in the context of the minimal randall - sundrum ( rs )
model with bulk fields and a brane - localized higgs sector , we derive closed analytical expressions for the gluon - gluon fusion process , finding that the effect of the infinite tower of virtual kk states can be described in terms of a simple function of the fundamental ( 5d ) yukawa matrices . given a specific rs model , this will allow one to easily constrain the parameter space , once a higgs signal has been established .
we explain that discrepancies between existing calculations of higgs production in rs models are related to the non - commutativity of two limits : taking the number of kk states to infinity and removing the regulator on the higgs - boson profile , which is required in an intermediate step to make the relevant overlap integrals well defined . even though the one - loop @xmath0 amplitude is finite in rs scenarios with a brane - localized higgs sector , it is important to introduce a consistent ultraviolet regulator in order to obtain the correct result .
mz - th/12 - 13 + march 30 , 2012 + revised july 20 , 2012 * higgs production in a warped extra dimension * marcela carena@xmath1 , sandro casagrande@xmath2 , florian goertz@xmath3 , + ulrich haisch@xmath4 and matthias neubert@xmath5 + |
we give here explicitly the set of thirteen ordinary differential equations ( odes ) for the amplitudes of the fourier modes @xmath88 that define the truncated euler equation ( tee ) in the case @xmath89 .
note that @xmath88 are real numbers because of the free - slip boundary conditions ( see text ) .
@xmath90 \end{split}\ ] ] | it is shown that the truncated euler equations , _ i_.e .
a finite set of ordinary differential equations for the amplitude of the large - scale modes , can correctly describe the complex transitional dynamics that occur within the turbulent regime of a confined @xmath0d navier - stokes flow with bottom friction and a spatially periodic forcing .
in particular , the random reversals of the large scale circulation on the turbulent background involve bifurcations of the probability distribution function of the large - scale circulation velocity that are described by the related microcanonical distribution which displays transitions from gaussian to bimodal and broken ergodicity .
a minimal @xmath1-mode model reproduces these results .
the formation of large scale coherent structures is widely observed in atmospheric and oceanic flows and ascribed to the nearly bi - dimensional nature of these flows .
kraichnan showed that in two - dimensional ( 2d ) turbulence , the energy is transferred from the forcing scale to larger scales due to the conservation of both energy and enstrophy by the inviscid dynamics @xcite . in a confined flow domain and without large scale friction ,
the energy accumulates at the largest possible scale , thus generating coherent structures in the form of large scale vortices . it has been observed in laboratory experiments that the large scale circulation generated by forcing a nearly @xmath0d flow at small scale can display random reversals @xcite .
the large scale velocity has a bimodal probability density function ( pdf ) with two symmetric maxima related to the opposite signs of the large scale circulation .
this regime bifurcates from another turbulent regime with a gaussian velocity field with zero mean when the large scale friction is decreased .
when the friction is decreased further , the reversals become less and less frequent and a condensed state with most of its kinetic energy in the large scale circulation is reached @xcite .
a similar sequence of transitions is observed in numerical simulations of the @xmath0d navier - stokes equation ( nse ) with large scale friction and spatially periodic forcing @xcite .
these transitions correspond to bifurcations of a mean flow on a strongly turbulent background for which no theoretical tool exists so far .
we show in this letter that the truncated euler equation ( tee ) , _ i_.e . a finite set of ordinary differential equations ( odes ) for the amplitude of the large - scale modes without forcing and dissipation , can correctly describe these transitions . to wit , we compare the dynamical regimes observed in numerical simulations of the @xmath0d nse with the ones obtained with the tee when the characteristic scale @xmath2 of the initial conditions is changed , with @xmath3 where @xmath4 is the kinetic energy of the flow and @xmath5 is its enstrophy ( integrated squared vorticity )
. the dimensionless 2d nse reads for an incompressible velocity field @xmath6 , @xmath7 where @xmath8 is the stream function and @xmath9 is the usual poisson bracket .
the first term on the right hand side represents the frictional force in the bottom boundary layer and @xmath10 is the spatially periodic forcing , explicitly given by @xmath11 . to model flow confinement
we use free - slip boundary conditions ; therefore , the stream - function can be fourier expanded as @xmath12 the non - dimensional parameters are the reynolds number , @xmath13 and @xmath14 , which represents the ratio of the inertial term to the friction on the bottom boundary .
here , @xmath15 is a characteristic large scale velocity , @xmath16 is the length of the square container , @xmath17 is the kinematic viscosity and @xmath18 is the damping rate related to the friction .
the above equation has been made dimensionless using the length scale @xmath16 and the velocity scale @xmath15 .
( -125,182)*(a ) dns * ( -190,140)*(a1 ) @xmath19 * ( -190,97)*(a2 ) @xmath20 * ( -190,55)*(a3 ) @xmath21 * ( -190,4)*(a4 ) @xmath22 * ( -135,182)*(b ) truncated euler * ( -190,140)*(b1)@xmath23 * ( -190,97)*(b2)@xmath24 * ( -190,55)*(b3)@xmath25 * ( -190,4)*(b4)@xmath26 * + ( -160,160)*(c ) * ( -130,89 ) ( -160,160)*(d ) * we perform direct numerical simulations ( dns ) of eq . , using standard pseudospectral methods @xcite with @xmath27 collocation points and @xmath28 circular dealiasing : @xmath29 .
time stepping is performed with a second - order , exponential time differencing runge - kutta method @xcite .
dnss of the nse ( [ eq : ns_incom ] ) are carried out for @xmath30 and @xmath31 with @xmath32 .
very long time integration is needed to accumulate reliable statistics for the reversals , which become rare with increase in @xmath33 ( see below ) .
direct time recordings of the amplitude of the lowest wave number mode of the stream function , @xmath34 , are displayed in fig .
[ fig:1](a ) for @xmath31 and different values of @xmath33 . for @xmath19 , the amplitude of
the large scale circulation fluctuates around zero and its pdf is gaussian ( not shown ) .
when @xmath33 is increased , a first transition is observed within the turbulent regime and can be characterized by a change of the shape of the pdf that becomes bimodal .
@xmath34 fluctuates around two non zero most probable opposite values and displays random transitions between these two states ( see fig .
[ fig:1](a2 ) ) .
this corresponds to random reversals of the large scale circulation on a turbulent background .
the mean waiting time between successive reversals increases with @xmath33 ( see fig .
[ fig:1](a3 ) ) and finally a large scale circulation with a given sign becomes the dominant flow component ( see fig .
[ fig:1](a4 ) ) .
this is the condensed state described by kraichnan @xcite .
the regime with random reversals of the large scale circulation is therefore located in parameter space between the condensed states and the turbulent regime with gaussian velocity pdfs as observed in experiments @xcite .
we now consider the approach , introduced by lee @xcite and developed by kraichnan @xcite that relies on the 2d tee .
they showed that the euler equation , truncated between a minimum and a maximum wave number , gives a set of odes for the amplitudes of the modes that follow a liouville theorem @xcite . for 2d flows , the kinetic energy @xmath4 and the enstrophy @xmath5 ( integrated squared vorticity )
are conserved ; therefore , the boltzmann - gibbs canonical equilibrium distribution is of the form @xmath35 , where @xmath36 is the partition function and @xmath37 and @xmath38 can be seen as inverse temperatures , determined by the total energy and enstrophy . using this formalism , kraichnan @xcite derived the absolute equilibria of the energy spectrum @xmath39 and showed the existence of different regimes depending on the values of @xmath37 and @xmath38 . on the other hand ,
microcanonical distributions are defined by @xmath40 , where @xmath41 and @xmath42 are respectively the energy and enstrophy of the initial conditions , and should be used to compute the pdfs in the reversal and condensed state ( see below ) .
the tee is obtained by performing a circular galerkin truncation at wave - number @xmath43 of the incompressible , euler equation @xmath44 , which is eq . without forcing or dissipation .
the tee in spectral space reads @xmath45 with @xmath46 the kronecker delta and with fourier modes satisfying @xmath47 if @xmath48 .
note that , because of the free - slip boundary conditions eq . , the fourier modes @xmath49 are real numbers .
this truncated system exactly conserves the quadratic invariants , energy and enstrophy , given in fourier space by @xmath50 and @xmath51 . for tee
we take @xmath43 as a free parameter and the same stream function expansion as that used for the ns eq .
, thus the numerical integration method is the same as the one described above for the nse . in both cases , the minimum wavenumber is @xmath52 .
we use an initial velocity field with an energy spectra @xmath53 , where by varying @xmath37 and @xmath38 we can obtain different flow regimes in accordance with the kraichnan s absolute equilibrium predictions .
we introduce a wave - number @xmath54 given by @xmath55 , which acts as an important control parameter of the system .
we next consider the results obtained using the tee with @xmath56 ( the nse forcing wavenumber ) and initial conditions with different values of @xmath54 .
figure [ fig:1](b ) shows the transitions between different turbulent regimes when @xmath54 is decreased .
the corresponding pdfs of @xmath34 obtained for different values of @xmath54 are displayed in the inset of fig .
[ fig:1](c ) .
we observe the transition from gaussian to bimodal pdf when @xmath54 is decreased and then the transition to the condensed regime with a given sign of the large scale circulation . for nse at large @xmath57 ,
the effect of the large scale friction is to stop the inverse cascade before reaching the scale of the flow domain .
@xmath33 thus determines the largest scale of the flow that we can define using the wave number @xmath55 .
@xmath54 is displayed in fig .
[ fig:1](c ) for two values of @xmath57 .
it weakly depends on @xmath57 and monotonously decreases when @xmath33 is increased .
when @xmath33 is large ( small friction ) , the kinetic energy accumulates at the scale of the flow domain and the condensed state is obtained .
although , we do not have a quantitative agreement between the transition values for @xmath54 for the tee and @xmath33 for the nse when using the relation between @xmath54 and @xmath33 displayed in fig .
[ fig:1](c ) , the same sequence of transitions is observed in both cases .
figure [ fig:1](d ) shows that we keep this qualitative agreement when the truncation is lower , @xmath58 .
this truncation leads to only @xmath1 odes for the amplitudes of the large scale modes ( see supplemental material @xcite ) .
it is remarkable that this set of equations correctly describes the transitions observed between the different turbulent regimes observed in direct numerical simulations and experiments .
the tee model is a finite number of quadratic nonlinear odes for real variables @xmath59 ( see the remark following eq . about the amplitudes of the fourier modes noted @xmath59 hereafter to simplify the notations ) that conserve both the energy @xmath60 and the enstrophy @xmath61 ( see supplemental material @xcite ) . by making use of the identities @xmath62 and @xmath63 we can write the total microcanonical phase space volume @xmath64 as @xmath65 with @xmath66 using the steepest descent method @xcite on the integral eq .
, the expression @xmath67 furnishes an explicit parametric expression for @xmath68 at the saddle - point @xmath69 parametrize real values of @xmath70 ) ] that corresponds to values of energy and enstrophy given by the saddle conditions @xmath71 and thus to @xmath55 .
we can estimate the same way the phase space volume for a fixed value of @xmath72 by retracing the steps from eq . to , but with the product and sums going from @xmath73 to @xmath74 instead of @xmath75 to @xmath74 . by combining these parametric representations
, we obtain an explicit expression for the normalized pdf of @xmath72 that is shown in fig.[fig:1](d ) and displays a good agreement with the numerical results .
note that canonical distributions with quadratic invariants are gaussian .
when there is condensation of energy at large scale , only a few modes are present and then the canonical distribution has no reason to reproduce the microcanonical distribution results @xcite .
indeed , @xmath76 with @xmath77 and @xmath78 implies that @xmath79 .
thus the microcanonical pdf of @xmath72 has to obey @xmath80 for @xmath81 which forbids reversals of the large scale circulation .
this represents ergodicity breakdown .
our above results show that this breakdown is preceded by an ergodicity delay , in the sense that @xmath82 becomes very small @xcite .
+ between successive reversals versus @xmath33 , obtained from our dnss of nse for @xmath31 ( blue circles ) .
inset : plot of the reversal frequency @xmath18 versus @xmath54 from the dnss of tee ; it shows that the reversal frequency decreases linearly with @xmath54 with a critical @xmath83 , below which the reversals are not observed for the integration time .
, title="fig : " ] ( -95,30 ) between successive reversals versus @xmath33 , obtained from our dnss of nse for @xmath31 ( blue circles ) .
inset : plot of the reversal frequency @xmath18 versus @xmath54 from the dnss of tee ; it shows that the reversal frequency decreases linearly with @xmath54 with a critical @xmath83 , below which the reversals are not observed for the integration time .
, title="fig : " ] we now consider in more detail the regime with random reversals of the large scale circulation and its transition to the condensed regime for which the flow no longer explores the whole phase space , keeping a given sign of the large scale circulation . as shown above ( compare fig .
[ fig:1](a ) and ( b ) ) , the mean waiting time @xmath84 between successive reversals increases when @xmath33 is increased in the nse , respectively @xmath54 is decreased in the tee .
however , the divergence of @xmath84 does not follow the same law for the nse and the tee .
figure [ fig:2](a ) shows an exponential increase of @xmath84 with @xmath33 in the nse , whereas a fit of the form @xmath85 with @xmath83 is observed in the tee ( see fig . [ fig:2](b ) ) .
the later result is expected since there exists a critical value of @xmath54 below which reversals are not possible in order to fulfill the conservation of both @xmath4 and @xmath5 .
we thus expect that @xmath84 becomes infinite for a finite value of @xmath54 .
a similar trend is not observed in the nse for @xmath84 versus @xmath33 .
this can not be explained using the relation between @xmath54 and @xmath33 displayed in fig .
[ fig:1](c ) that is roughly linear close to the transition to the condensed regime .
in contrast to the tee , the nse does not involve conserved quantities that prevent reversals , even when @xmath33 is large .
in addition , all the modes above @xmath86 that are suppressed in the tee can act as an additional source of noise in the nse and trigger reversals .
+ although it can be expected that viscous dissipation is negligible for the dynamics of large scales , it is remarkable that taking into account the effect of large scale friction by selecting the value of @xmath87 in the initial conditions of the tee is enough to describe the bifurcations of the large scale flow using a small number of modes governed by the euler equation .
thus , one discards the huge number of degrees of freedom related to small scale turbulent fluctuations .
in addition , equilibrium statistical mechanics , using the microcanonical distribution related to the tee , correctly describes the pdf of the large scale velocity in the different turbulent regimes .
transitions between different mean flows are widely observed in turbulent regimes , the most famous example being the drag crisis for which the wake of a sphere becomes narrower . using the navier - stokes equation with noisy forcing @xcite
is a way to describe this type of transitions .
the tee as presented here , can provide a new method to describe the dynamics of large scales in turbulence and to model a bifurcation of the mean flow on a strongly turbulent background .
+ we thank francois ptrlis for useful discussions .
support of the indo - french centre for the promotion of advanced research ( ifcpar / cefipra ) contract 4904-a is acknowledged .
this work was granted access to the hpc ressources of mesopsl financed by the region ile de france and the project equip@meso ( reference no . anr-10-eqpx-29 - 01 ) of the programme investissements davenir supervised by the agence nationale pour la recherche . vs acknowledges supported from euhit - european high- performance infrastructure in turbulence , which is funded by the european commission framework program 7 ( grant no .
312778 ) .
18ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ]
+ 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty @noop * * , ( ) link:\doibase 10.1017/s0022112086000836 [ * * , ( ) ] \doibase doi:10.1209/0295 - 5075/111/44002 [ * * , ( ) ] link:\doibase 10.1103/physreve.91.053005 [ * * , ( ) ] @noop _ _
( , , ) link:\doibase 10.1006/jcph.2002.6995 [ * * , ( ) ] link:\doibase 10.1017/s0022112075000225 [ * * , ( ) ] @noop * * , ( ) @noop @noop _ _ ( , ) http://opac.inria.fr/record=b1104896[__ ] , international series in pure and applied mathematics ( , , )
\doibase http://dx.doi.org/10.1063/1.862210 [ * * , ( ) ] link:\doibase 10.1103/physreve.90.043010 [ * * , ( ) ] @noop * * , ( ) \doibase http://dx.doi.org/10.1016/s0375-9601(98)00802-0 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.102.094504 [ * * , ( ) ] |
the search for a possible critical point of qcd is one of the driving motivations for the low energy runs at rhic / bnl and the upcoming facilities fair at gsi / darmstadt and nica in dubna .
the assumed existence of a critical point is based on two observations : at vanishing baryochemical potential highly sophisticated lattice qcd calculations revealed an analytic crossover @xcite and at high baryochemical potentials various approaches to effective models in mean - field and beyond find a first order phase transition , which then necessarily terminates in a critical point @xcite . at both ,
the first order phase transition and the critical point , an enhancement of fluctuations is predicted to be seen in heavy - ion collisions . for the first order phase transition these are based on the growth of fluctuations due to the nonequilibrium effect of supercooling leading to nucleation and spinodal decomposition @xcite .
such fluctuations should be visible in studies of single event particle distributions .
the critical point , however , is supposed to imprint its signals in event - by - event fluctuations of conserved quantities , such as net - charge or net - baryon number distributions @xcite . in thermodynamic systems
the correlation length diverges .
systems created in a heavy - ion collision , however , are finite in size , highly dynamic and a temperature and baryochemical potential can at most be defined locally .
the finite size of the system naturally limits the growth of the correlation length .
the finite time , which the dynamic systems spends at a critical point , limits the growth even further because relaxation times also become infinite at the critical point .
this phenomenon is called critical slowing down . assuming a phenomenological time evolution of the correlation length with parameters from the @xmath0d ising universality class it was found that the correlation length does not grow beyond @xmath1 fm @xcite . in all of these
approaches the background medium is assumed to be homogeneous in its thermodynamic quantities , like the temperature .
recently , the model of nonequilibrium chiral fluid dynamics ( n@xmath2fd ) for the explicit propagation of fluctuations coupled to a dynamic expansion of a heavy - ion collision was developed @xcite showing that nonequilibrium effects can have an important influence on the evolution of the fluctuations of the order parameter of chiral symmetry @xcite . in this model
the expansion of the fireball created in a heavy - ion collision is described by ideal fluid dynamics , where the fluid consists of quark and antiquarks with the local thermodynamic properties of the quark - meson model @xcite . in this work
we want to investigate the influence of the inhomogeneity of the fluid dynamic background on the sigma field and the correlation length . for this purpose
we restrict ourself to a local equilibrium treatment of the sigma field , instead of an explicit propagation but work with concepts that have already been studied in n@xmath2fd .
we refer to these references @xcite for more information .
for a fixed baryochemical potential @xmath3 the strength of the phase transition can be tuned via the quark - meson coupling @xmath4 . at @xmath5 and for a realistic coupling of @xmath6
the transition is a crossover , for @xmath7 it is a critical point with @xmath8 mev and a first order phase transition with @xmath9 mev is found for @xmath10 .
the thermodynamic potential of the quark - meson model in mean - field approximation is @xmath11 with the chiral potential @xmath12 and the quark contribution to fermionic one - loop level .
the equilibrium values for the chiral fields , @xmath13 and @xmath14 , are obtained from minimizing the thermodynamic potential ( [ eq : lsm_omegaq0 ] ) with respect to @xmath15 and @xmath16 .
while @xmath13 becomes finite at lower temperatures due to dynamic symmetry breaking , @xmath16 vanishes for all temperatures .
we do not continue to explicitly mention the pion field in the notation .
the minima of the thermodynamic potential @xmath17 define the thermodynamically stable states of the matter . in this case
the pressure and the energy density of the system are @xmath18 the fermions are described fluid dynamically by the relativistic equation of energy - momentum conservation @xmath19 with the energy - momentum tensor of an ideal fluid .
starting from initial conditions from urqmd @xcite the fluid has an inhomogeneous temperature distribution , which is determined by the evolution of the energy density .
therefore , the equilibrium value of the sigma field is defined only locally in each cell of the fluid dynamic grid .
speaking of the temperature of the system thus implies some form of averaging over parts of the system . as a consequence ,
all quantities deduced from thermodynamic expressions are subject to the same averaging .
a prominent feature in a scenario with a first order phase transition is the discontinuity of the sigma field at the transition temperature , which will be studied in section [ sec : sec1 ] . in a critical point scenario
we investigate the correlation of the fluctuations based on the definition via the curvature of the thermodynamic potential in section [ sec : sec2 ] .
in figure [ fig : fo ] we present the temperature and the sigma field averaged over an inner volume of about @xmath20 @xmath21 .
as the system expands it dilutes and the energy density and the temperature become smaller until the coexistence region of the first order phase transition is reached .
here , the local temperature stays constant .
since almost all of the volume has an energy density in the coexistence region , also the averaged temperature stays almost constant .
the averaged sigma field increases during the transition .
if we assumed a homogeneous temperature of the system with the same evolution as the averaged temperature in figure [ fig : fo ] and assumed the corresponding equilibrium value of the sigma field for the entire system , we recover the discontinuity in the evolution of the sigma field .
this discontinuity is washed out when averaging over an inhomogeneous system .
in the right plot of figure [ fig : cp ] the same curves are plotted as in figure [ fig : fo ] but here for a scenario with a critical point .
again we find that in the transition region the evolution of the sigma field is washed out in an inhomogeneous system after averaging compared to an assumed homogeneous temperature evolution . in the left plot
we show the influence of the inhomogeneity on the correlation length , which is locally evaluated from @xmath22 we observe that due to averaging the correlation length in an inhomogeneous system does not diverge at the averaged critical temperature of the system .
it remains finite and grows up to @xmath23 fm .
when we assume again that the temperature evolution was a homogeneous temperature of the system we can calculate a single correlation length , which would correspond to the system s correlation length .
it is shown in figure [ fig : cp ] for comparison .
in finite systems one speaks about a critical region around the critical point , where the correlation length is still significantly enhanced but does not diverge . for an inhomogeneous system
we find the same situation when averaging over parts of the system .
we showed that the inhomogeneity of a system created in a heavy - ion collision smears out the evolution of the order parameter of chiral symmetry , the sigma field , in the transition region .
the effect on the correlation length is more dramatic , as its divergence at the critical temperature in a homogeneous system is reduced to an increase up to @xmath23 fm . in the investigated scenario ,
however , fluctuations of the sigma field are not directly studied .
it needs to be said that the growth of the correlation length is a very dynamic process and should be studied in a fully dynamic setup , what is currently worked on within n@xmath2fd .
2em this work was supported by the hessian excellence inititive loewe through the helmholtz international center for fair .
we are grateful to the center for scientific computing ( csc ) at frankfurt for providing the computing resources . | according to a fluid dynamic expansion of the fireball we investigate how the inhomogeneity of the system influences the chiral phase transition of qcd .
we compare the averaged values of the order parameter in equilibrium with that of a homogeneous system .
if the temperature is averaged over a certain region of the fireball the corresponding correlation length does not diverge in an expansion with a critical point . |
let @xmath27 .
let @xmath2 be an algebraically closed field and @xmath0 a @xmath1-dimensional symplectic vector space over @xmath2 .
the symplectic group @xmath15 acts naturally on @xmath0 from the left hand side , and hence on the @xmath28-tensor space @xmath18 .
let @xmath29 be the brauer algebra over @xmath2 with generators @xmath30 @xmath31 and parameter @xmath32 ( see [ dfn1 ] for their definitions ) .
there is a right action of @xmath17 on @xmath9 which commutes with the left action of @xmath15 .
let @xmath33 be the following natural @xmath2-algebra homomorphisms : @xmath34 for any positive integer @xmath35 , a composition of @xmath35 is a sequence of non - negative integers @xmath36 with @xmath37 .
we use @xmath38 to denote the largest integer @xmath39 such that @xmath40 .
a composition @xmath36 of @xmath35 is said to be a partition if @xmath41 ; in that case , we write @xmath42 .
the following results are referred as brauer schur weyl duality .
[ thm11 ] ( @xcite , @xcite , @xcite ) [ ddh1 ] 1 ) the natural left action of @xmath15 on @xmath18 commutes with the right action of @xmath17 ; \2 ) both @xmath43 and @xmath44 are surjective ; \3 ) if @xmath45 then @xmath43 is injective , and hence an isomorphism ; \4 ) if @xmath46 , then there is a decomposition of @xmath18 as a direct sum of irreducible @xmath47-@xmath17-bimodules : @xmath48}\bigoplus_{\substack{{\lambda}\vdash n-2f\\ \ell({\lambda})\leq m}}\delta({{\lambda}})\otimes d^{(f,{\lambda})},\ ] ] where @xmath49 ( respectively , @xmath50 ) denotes the irreducible @xmath15-module ( respectively , the irreducible @xmath17-module ) corresponding to @xmath24 ( respectively , corresponding to @xmath51 ) .
there is a variant of the above brauer schur
weyl duality as we shall describe .
let @xmath52 be the two - sided ideal of @xmath17 generated by @xmath53 .
we set @xmath54 [ w1n ] we call @xmath55 the subspace of harmonic tensors or traceless tensors .
note that our definition of harmonic tensors looks slightly different from that given in @xcite and @xcite .
the two definitions are reconciled in corollary [ 2dfn ] . if @xmath46 , then we shall write @xmath56 instead of @xmath57 in order to emphasize the base field .
we have that @xmath58 .
the right action of @xmath17 on @xmath18 gives rise to a right action of @xmath59 on @xmath55 .
( ( * ? ? ?
* ( 10.2.7 ) ) , @xcite ) [ gw1 ] the natural left action of @xmath15 on @xmath55 commutes with the right action of @xmath59 .
if @xmath60 , then there is a decomposition of @xmath61 as a direct sum of irreducible @xmath47-@xmath62-bimodules : @xmath63 where @xmath64 ( respectively , @xmath65 ) denotes the irreducible @xmath47-module ( respectively , the irreducible @xmath62-module ) corresponding to @xmath24 . as before , we have two natural @xmath2-algebra homomorphisms : @xmath66 in @xcite , de concini and strickland proved that @xmath67 is independent of the field @xmath2 and @xmath68 is always surjective .
furthermore , they showed that if @xmath45 , then @xmath68 is an isomorphism .
their proof makes use of the previous results in @xcite and @xcite on multilinear invariants of a variety and symplectic standard tableaux which eventually relies on some algebro - geometric arguments . in @xcite , using the theory of rational representations of symplectic group , maliakas proved that @xmath69 has a good filtration whenever @xmath45 and he claimed that it is true for arbitrary @xmath70 .
the starting point of this paper is , on one hand , to generalize the above duality to the case of partially harmonic tensors of arbitrary valence @xmath8 , and on the other hand , to provide a self - contained and purely representation - theorietic approach which makes it possible to work also in the quantized case ) which prevent us from generalizing the main results of this paper to the quantized case .
the main difficulty lies in that there are various choices of tangle which can specialize to the same brauer diagram and it also becomes much harder to describe the action of a tangle on tensor space in a simple combinatorial way ( as in the brauer algebra case ) . ] .
we are mostly interested in the non - semisimple case . to describe our main results , we need some more notations and definitions .
for each integer @xmath6 $ ] , let @xmath3 be the two - sided ideal of @xmath17 generated by @xmath5 . by convention , @xmath71 and @xmath72 + 1)}=\{0\}$ ] .
set @xmath73 following ( * ? ? ?
* ( 10.3.1 ) ) , we call @xmath74 the space of _ partially harmonic tensors of valence @xmath8_. classically , these spaces play an important role in the study of the @xmath75-module structure on the @xmath28-tensor spaces @xmath76 ( cf .
one of our motivation for studying them is to try to use them to construct some natural ( integral ) quotient of the symplectic schur algebras . by (
* ( 10.3.14 ) ) , it is easy to see that if @xmath46 then @xmath77 as a @xmath78-linear space . in particular , @xmath79 as a @xmath78-linear space .
note that the space @xmath0 can be defined over arbitrary field ( and even over @xmath80 ) .
so we can consider the spaces @xmath81 , @xmath82 and @xmath83 over an arbitrary field @xmath2 .
all the results we obtain in this paper are valid over these more general ground fields .
however , we shall always assume that @xmath2 is algebraically closed whenever notions and results from algebraic groups theory ( e.g. , good filtration , weyl filtration ) are needed .
the transition between an arbitrary field and its algebraic closure usually follows from some standard arguments in commutative algebras .
note that @xmath83 is a @xmath15-@xmath84-bimodule .
we use @xmath85 to denote the following natural homomorphism : @xmath86 let @xmath87 denote the value of the usual kronecker delta . for each integer @xmath88 , we set @xmath89 .
we fix an ordered basis @xmath90 of @xmath0 such that @xmath91 we use @xmath92 to denote the free @xmath80-submodule of @xmath93 generated by @xmath94 . for any commutative @xmath80-algebra @xmath95
, we set @xmath96 . the brauer algebra @xmath17 can also be defined over @xmath95 and we denote it by @xmath97 . to simplify notations , the two - sided ideal of @xmath97 generated by @xmath5 will be still denoted by @xmath3 .
the main results in this paper are the following theorems and corollaries .
[ mainthm0 ] for each integer @xmath98 $ ] , \1 ) @xmath99 is a pure @xmath80-submodule of @xmath100 , equivalently , @xmath101 is a free @xmath80-module ; \2 ) both @xmath102 and @xmath83 are stable under base change , i.e. , for any commutative @xmath80-algebra @xmath95 , the canonical maps @xmath103 are isomorphisms . in particular , the character formulae of the left @xmath15-modules @xmath104 are both independent of the field @xmath2 .
[ mainthm1 ] let @xmath2 be an algebraically closed field .
for each integer @xmath8 with @xmath98 $ ] , both @xmath83 and @xmath102 have a good filtration as @xmath15-modules . [ maincor1 ]
let @xmath2 be an algebraically closed field .
let @xmath6 $ ] .
then the dimension of @xmath82 is independent of @xmath2 , and there is a @xmath15-@xmath105-bimodule isomorphism : @xmath106 in particular , the dimension of @xmath74 is independent of @xmath2 too .
[ maincor15 ] let @xmath2 be an algebraically closed field .
for each @xmath6 $ ] , the @xmath15-module @xmath82 always has a good filtration and the @xmath15-module @xmath74 always has a weyl filtration . [ mainthm2 ]
let @xmath2 be an algebraically closed field .
for each integer @xmath8 with @xmath98 $ ] , \1 ) the dimension of the endomorphism algebra @xmath13 is independent of @xmath2 ; \2 ) @xmath107 .
the proof of the above results are _ completely self - contained _ and use purely representation - theorietic knowledge . as a consequence of these theorems and corollaries , we recover and extend the previously mentioned results of de concini and strickland @xcite and the result of maliakas @xcite . in the way of our proof
, we also obtain the following result , which seems of independent interest .
[ mainthm3 ] 1 ) as a @xmath15-@xmath17-bimodule , @xmath18 has a filtration such that each successive quotient is isomorphic to some @xmath108 with @xmath20 , @xmath21 and @xmath22 $ ] , where @xmath23 is the co - weyl module associated to @xmath24 and @xmath25 is a maximal vector of weight @xmath24 ( see [ zdfn ] for its definition ) ; \2 ) for any partition @xmath24 of @xmath109 with @xmath22 $ ] and @xmath21 and any commutative @xmath80-algebra @xmath95 , the canonical map @xmath110 is always an isomorphism .
in particular , the dimension of @xmath26 is independent of @xmath2 .
in fact , if @xmath46 then @xmath111 is a simple right @xmath112-module . therefore , for any field @xmath2 , the dimension of @xmath26 is always equal to the number of @xmath113-permissible up - down tableaux of shape @xmath114 and length @xmath28 ( cf .
* theorem 1.1 , theorem 1.2 ) and @xcite ) , where @xmath114 denotes the conjugate of @xmath24 .
the paper is organized as follows . in section 2
we recall some basic knowledge about brauer algebras and their actions on @xmath28-tensor spaces .
in particular , we show that our definition [ w1n ] of harmonic tensors coincides with that given in @xcite and @xcite . in section 3 , we give the proof of theorem [ mainthm0 ] , [ mainthm1 ] and corollary [ maincor1 ] .
the main idea is to show that @xmath102 can be identified with the image of @xmath18 under a truncation functor @xmath115 associated with a saturated set @xmath116 of dominant weights .
the proof makes use of the main result obtained in @xcite , some results on weyl filtration ( resp .
, good filtration ) and a key lemma [ hu2lem ] . as a consequence , we prove the first part of theorem [ mainthm3 ] .
section 4 is devoted to the proof of lemma [ hu2lem ] .
the proof relies on lusztig s theory of canonical bases and based modules . as a result
, we get the second part of theorem [ mainthm3 ] . in section 5
we prove theorem [ mainthm2 ] , which gives one side of the brauer - schur - weyl duality between @xmath117 and @xmath118 on @xmath119 .
we conjecture that the other side of this duality is also true .
let @xmath120 .
the brauer algebra @xmath17 with parameter @xmath121 and size @xmath28 was first introduced by richard brauer ( see @xcite ) when he studied how the @xmath28-tensor space @xmath122 decomposes into irreducible modules over @xmath123 . in his language , @xmath17 was defined as the @xmath78-linear space with a basis being the set @xmath124 of all the brauer @xmath28-diagrams . by definition ,
a brauer @xmath28-diagram is a diagram with specific @xmath125 vertices arranged in two rows of @xmath28 each , the top and bottom rows , and exactly @xmath28 edges such that every vertex is joined to another vertex ( distinct from itself ) by exactly one edge .
we label the vertices in each row of a brauer @xmath28-diagram by the integers @xmath126 from left to right .
the multiplication of two brauer @xmath28-diagrams is defined as follows .
we compose two diagrams @xmath127 by identifying the bottom row of @xmath128 with the top row of @xmath129 ( such that the vertex @xmath130 in the bottom row of @xmath128 is identified with the vertex @xmath130 in the top row of @xmath129 for each @xmath131 ) .
the result is a graph , with a certain number , @xmath132 , of interior loops . after removing the interior loops and the identified vertices , retaining the edges and remaining vertices
, we obtain a new brauer @xmath28-diagram @xmath133 , the composite diagram .
then we define @xmath134 . in general
, the multiplication of two elements in @xmath17 is given by the linear extension of a product defined on diagrams .
for example , let @xmath135 be the following brauer @xmath136-diagram .
let @xmath137 be the following brauer @xmath136-diagram .
then @xmath138 is equal to [ 0.55 ] alternatively , one can define the brauer algebra using generators and relations .
[ dfn1 ] let @xmath2 be a field .
the brauer algebra @xmath17 over @xmath2 is a unital associative @xmath2-algebra with canonical generators @xmath139 and relations ( see @xcite ) : @xmath140 replacing @xmath2 by any commutative @xmath80-algebra @xmath95 , we can define the brauer algebra @xmath97 over @xmath95 in a similar way .
the algebra @xmath97 is a free @xmath95-module with rank @xmath141 , and the canonical map @xmath142 is an isomorphism for any commutative @xmath80-algebra @xmath95 .
the two definitions of the brauer algebra @xmath17 can be identified as follows .
let @xmath130 be an integer with @xmath143 .
the generator @xmath144 corresponds to the brauer @xmath28-diagram with edges connecting vertices @xmath130 ( respectively , @xmath145 ) on the top row with @xmath145 ( respectively , @xmath130 ) on the bottom row , and all other edges are vertical , connecting vertices @xmath35 on the top and bottom rows for all @xmath146 .
the generator @xmath147 corresponds to the brauer @xmath28-diagram with horizontal edges connecting vertices @xmath148 on the top and bottom rows , and all other edges are vertical , connecting vertices @xmath35 on the top and bottom rows for all @xmath146 .
note that the subalgebra of @xmath149 generated by @xmath150 is isomorphic to the group algebra of the symmetric group @xmath151 over @xmath80 .
the brauer algebra was studied in a number of literatures , see @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite .
we are mainly interested in their actions on @xmath28-tensor space @xmath18 and related schur
weyl dualities involving symplectic groups . from now
on let @xmath2 be an algebraically closed field and @xmath0 a @xmath1-dimensional @xmath2-vector space equipped with a non - degenerate skew - symmetric bilinear form @xmath152 .
let @xmath15 be the corresponding symplectic group . recall that for each integer @xmath88 , @xmath89 .
let @xmath153 be a @xmath2-basis of @xmath0 such that @xmath91 for each integer @xmath88 , we define @xmath154 then @xmath153 and @xmath155 are dual bases for @xmath0 in the sense that @xmath156 for any @xmath157 .
there is a right action of @xmath17 on @xmath18 which is defined on generators by @xmath158 assume @xmath159 . for any pair of integers @xmath160
we define the @xmath161-contraction operator @xmath162 by @xmath163 where @xmath164 , @xmath165 mean that we omit the tensor factors @xmath166 and @xmath167 in the tensor product , and we define the @xmath161-expansion operator @xmath168 by @xmath169 ( @xcite ) for each pair of integers @xmath160 , both @xmath170 and @xmath171 are @xmath15-module homomorphisms . for each pair of integers @xmath160 , we use @xmath172 to denote the unique brauer @xmath28-diagram which satisfies the following two conditions : 1 . for any integer @xmath173 , the vertex labeled by @xmath174 in the top row is connected with the vertex labeled by @xmath174 in the bottom row ; 2 .
the vertex labeled by @xmath175 in either the top row or the bottom row is connected with the vertex labeled by @xmath39 in the same row .
in particular , @xmath176 for any integer @xmath177 .
[ contra ] for any pair of integers @xmath160 , we have @xmath178 . in particular
, @xmath179 by definition , for any integers @xmath180 , @xmath181 it follows that for any @xmath182 , @xmath183 . on the other hand ,
it is clear that @xmath171 maps the basis of simple @xmath184-tensors to a set of @xmath2-linearly independent elements in @xmath18 .
hence it is an injective map from @xmath185 to @xmath18 .
it follows that @xmath178 .
in particular , @xmath186 [ br ] let @xmath187 .
the following three statements are equivalent : 1 .
@xmath188 contains exactly two horizontal edges ( one edge in each of the top and the bottom rows in the diagram ) ; 2 .
@xmath189 , for some @xmath190 and two integers @xmath191 ; 3 .
@xmath192 for some @xmath193 and two integers @xmath194 . by @xcite
, @xmath188 contains exactly two horizontal edges if and only if @xmath195 for some @xmath196 , @xmath197 , where @xmath198 is the set of distinguished right coset representatives of @xmath199 in @xmath200 .
it follows that @xmath201 since both @xmath202 and @xmath203 are of the form @xmath172 for two distinct integers @xmath160 , the lemma follows at once . for each integer
@xmath204 $ ] , we denote by @xmath3 the two - sided ideal of @xmath17 generated by @xmath5 . note that @xmath3 is spanned by all the brauer @xmath28-diagrams which contain at least @xmath205 horizontal edges .
recall the definition of @xmath55 given in definition [ w1n ] .
[ 2dfn ] with the notations as above , we have that @xmath206 this follows directly from lemma [ contra ] and [ br ] . note
that the above corollary ensures that our definition [ w1n ] of harmonic tensors coincides with that given in @xcite and @xcite .
@xmath207 " be the @xmath2-algebra anti - automorphism ( of order two ) of @xmath17 which is defined on generators by @xmath208 for any right @xmath17-module @xmath209 , the dual space @xmath210 is naturally endowed with a right @xmath17-module structure via the anti - involution @xmath207 " .
that is , @xmath211 . for each integer @xmath212
, we set @xmath213 for any @xmath214 , we write @xmath215 .
the bilinear form @xmath216 on @xmath0 naturally induces a non - degenerate bilinear form on @xmath18 such that @xmath217 by definition and an easy check , we see that for any @xmath218 , @xmath219 in other words , the bilinear form @xmath220 induces a @xmath15-@xmath17-bimodule isomorphism @xmath221 by definition and ( [ asso ] ) , we deduce that @xmath222 restricts to an inclusion @xmath223 in the next section , we shall show that @xmath224 is actually an isomorphism , see corollary [ comdim ] .
the purpose of this section is to give a proof of theorem [ mainthm0 ] , [ mainthm1 ] and corollary [ maincor1 ] .
one of the key steps is the use of lemma [ hu2lem ] , whose proof will be given in the next section .
let @xmath225 .
we identify @xmath0 with @xmath226 and @xmath15 with @xmath227 by using the ordered basis @xmath228 .
then @xmath229 where @xmath230 denotes the matrix transpose of @xmath231 , @xmath232 and for each @xmath233 , @xmath234 denotes the matrix unit which is @xmath235 on the @xmath236th position and @xmath237 elsewhere . recall that @xmath227 is a connected semisimple linear algebraic group over @xmath238 .
let @xmath239 .
let @xmath240 be the subgroup consisting of all diagonal matrices in @xmath227 .
that is , @xmath241 then @xmath240 is a maximal torus of @xmath227 .
let @xmath242 be the weight lattice of @xmath227 .
an element in @xmath243 will also be called a @xmath240-weight .
let @xmath244 be the set of dominant @xmath240-weights of @xmath227 .
let @xmath245 be the maximal torus consisting of all the diagonal matrices in @xmath246 . for each @xmath88
, let @xmath247 be the function which sends a matrix in @xmath245 to its @xmath130th element in the diagonal .
then @xmath240 is a subtorus of @xmath245 .
let @xmath248 , which is the set of simple roots of @xmath249 .
let @xmath250 be the corresponding positive borel subgroup .
we refer the reader to ( * ? ? ?
* , page 144145 ) for the explicit description of @xmath250 . for each @xmath251 , we denote by @xmath252 and @xmath253 the simple module , co - weyl module and weyl module for @xmath227 associated to @xmath24 respectively .
we identify the weight @xmath254 with @xmath255 .
thus a weight @xmath256 is dominant if and only if @xmath257 is a partition . for any @xmath258 , we define @xmath259 let @xmath8 be an integer with @xmath6 $ ] .
let @xmath264 be the set of dominant @xmath240-weights of @xmath227 appearing in @xmath265 .
by @xcite , @xmath266$. } \end{matrix}\biggr\}.\ ] ] it is well - known that @xmath116 is saturated in the sense of ( * ? ? ? *
part ii , a.2 ) .
let @xmath267 denote the category of finite dimensional @xmath227-modules @xmath209 such that all composition factors of @xmath209 have the form @xmath268 with @xmath269 .
following ( * ? ? ?
* part ii , chapter a ) , we define a functor @xmath115 from the category of finite dimensional @xmath227-modules to the category @xmath267 as follows : @xmath270 let @xmath209 be a finite dimensional @xmath227-module . recall that ( ( * ? ? ?
* part ii , 4.16 ) ) an ascending chain @xmath271 of @xmath227-submodules is called a good filtration of @xmath209 if each @xmath272 is isomorphic to some @xmath273 with @xmath274 .
we use @xmath275 to denote the number of factors in the filtration isomorphic to @xmath273
. similarly , an ascending chain @xmath276 of @xmath227-submodules is called a weyl filtration of @xmath209 if each @xmath277 is isomorphic to some @xmath278 with @xmath279 .
we use @xmath280 to denote the number of factors in the filtration isomorphic to @xmath278 .
note that the left action of @xmath227 on @xmath18 commutes with the right action of @xmath17 . by definition
, @xmath291 is a sum of some submodules of the form @xmath292 where @xmath293 . applying the result 1 ) that we have proved
, we get that ( as a @xmath227-module ) @xmath294 therefore , it follows from definition that @xmath295 .
we want to show that @xmath296 . before proving this equality , we need some preparation . for simplicity , for any two finite dimensional @xmath227-modules
@xmath297 , we use @xmath298 to denote the sum of all the image subspaces @xmath299 , where @xmath300 runs over all the @xmath227-module homomorphisms from @xmath209 to @xmath301 .
it is clearly a @xmath227-submodule of @xmath301 .
we use @xmath302 to denote the following embedding : @xmath303 recall that there is a natural right action of the brauer algebra @xmath315 on the @xmath316-tensor space @xmath317 .
we use @xmath318 to denote the natural @xmath2-algebra homomorphism from @xmath319 to @xmath320 . by (
* theorem 1.4 ) , we know that @xmath318 is surjective . therefore , @xmath320 is spanned by the elements @xmath321 , where @xmath135 runs over the brauer @xmath316-diagrams in @xmath322 .
let @xmath323 .
we regard @xmath321 as an element in @xmath324 using the isomorphism @xmath325 . by the definition of @xmath318 , it is easy to see that @xmath326 for some @xmath327 . now using the isomorphism
@xmath328 one can check the image of @xmath321 .
it follows that @xmath329 is spanned by all @xmath305 , where @xmath306 .
it remains to prove that @xmath330 .
it suffices to show that the image of @xmath331 is contained in @xmath102 for each @xmath323 .
note that @xmath332 .
in particular , we can write @xmath333 where @xmath334 are @xmath335 independent summation indices . the positions of each pair of @xmath336 in the above sum are uniquely determined by @xmath337 and hence by @xmath135 .
for example , if then @xmath338 each independent summation index @xmath339 was attached with a pair of integers @xmath340 such that @xmath341 appear in the position @xmath342 respectively . using ( * ? ? ?
* lemma 4.2 ) and the definition of @xmath328 , it suffices to show that @xmath343 we fix a bijection between the horizontal edges in the top row of @xmath135 and the horizontal edges in the bottom row of @xmath135 .
we consider a vertex @xmath231 in the top row of @xmath135 which is labeled by an integer @xmath188 with @xmath344 .
there are only three possibilities : 1 .
@xmath231 is connected with a vertex @xmath250 in the top row of @xmath135 which is labeled by an integer @xmath345 with @xmath346 , then the horizontal edge @xmath347 must correspond to a horizontal edge @xmath348 in the bottom row of @xmath135 . applying ( * ? ? ?
* lemma 4.2 ) and the definition of @xmath325 , we see that the edges @xmath349 determine two integers @xmath350 , such that @xmath351 ; 2 .
@xmath231 is connected with a vertex @xmath250 in the top row of @xmath135 which is labeled by an integer @xmath345 with @xmath352 , then the horizontal edge @xmath347 must correspond to a horizontal edge @xmath348 in the bottom row of @xmath135 . applying ( * ? ? ?
* lemma 4.2 ) and the definition of @xmath325 , we see that the edges @xmath348 determines an integer @xmath353 , such that @xmath354 ; 3 . @xmath231 is connected with a vertex @xmath250 in the bottom row of @xmath135 , then applying ( * ? ? ?
* lemma 4.2 ) and the definition of @xmath325 , we see that the vertical edge @xmath347 determines an integer @xmath353 , such that @xmath355 .
[ bp2 ] let @xmath209 be a @xmath227-module and @xmath372 .
suppose @xmath373 is a surjective @xmath227-module homomorphism such that @xmath374 for some highest weight vector @xmath375 of weight @xmath24 , then @xmath325 is an isomorphism .
in particular , @xmath376 .
suppose that @xmath387 .
then there exist some integers @xmath388 such that @xmath389 comparing the coefficients of @xmath247 for each @xmath390 and adding all of them together , we get that @xmath391 which is a contradiction to the assumption that @xmath392 . as a @xmath227-module ,
@xmath393 is a tilting module . by (
* ? ? ? * proposition e.7 ) , we know that @xmath18 is a tilting module too
. in particular , @xmath18 has both a weyl filtration and a good filtration .
we fix an integer @xmath8 with @xmath6 $ ] . in view of lemma [ bp ] and lemma [ compare ] ,
we can find a good filtration of @xmath18 : @xmath394 such that 1 .
@xmath395 for @xmath396 , where @xmath397 , @xmath398 for some integers @xmath399 $ ] and @xmath400 ; 2 .
@xmath401 for any @xmath402 , and @xmath403 only if @xmath404 ; 3 .
there exists an integer @xmath405 , such that @xmath406 if and only if @xmath407 .
\1 ) if @xmath409 is a ( non - zero ) maximal vector of weight @xmath410 in @xmath18 , then @xmath411 is also a ( non - zero ) maximal vector of weight @xmath410 in @xmath412 .
moreover , every maximal vector of weight @xmath410 in @xmath412 arises in this way ; \2 ) since @xmath418 has a good filtration with each section being isomorphic to some @xmath419 with @xmath420 , it follows from lemma [ bp ] that @xmath421 therefore , the natural embedding @xmath422 becomes an equality , from which we deduce that @xmath423 .
\3 ) this follows from the following simple fact : if @xmath209 is a finite dimensional @xmath227-module with a good filtration and @xmath372 is minimal such that @xmath23 appears in a good filtration of @xmath209 , then @xmath424 is a direct sum of copies of @xmath23 and @xmath425 has a good filtration without factor @xmath23 .
in particular , @xmath426 is a right module for @xmath427 and @xmath425 has an @xmath427-module structure .
let @xmath35 be an integer and @xmath24 a composition of @xmath35 .
the conjugate of @xmath24 is the partition @xmath428 where @xmath429 .
the young diagram of @xmath24 is the set @xmath430:=\bigl\{(a , b)\bigm|1\leq b\leq{\lambda}_a\bigr\}.\ ] ] assume that @xmath24 is a partition of @xmath28 .
a @xmath24-tableau is a bijective map @xmath431\rightarrow\{1,2,\dots , n\}$ ] .
a standard @xmath24-tableau is a @xmath24-tableau in which the entries increase along each row and down each column .
let @xmath432 ( resp . ,
@xmath433 ) be the standard @xmath24-tableau such that the numbers @xmath126 appear in order along the rows ( resp .
, along the columns ) .
let @xmath434 be the young subgroup of @xmath200 corresponding to @xmath24 , which is the subgroup fixing the sets @xmath435 .
let @xmath436 be the set of distinguished right coset representatives of @xmath434 in @xmath200 .
let @xmath437 such that @xmath438 .
then @xmath439 . let @xmath440 .
recall the definition of @xmath290 in definition [ alpha ] .
recall that @xmath357 is the unipotent radical of the positive borel subgroup @xmath250 .
for each @xmath227-module @xmath209 , we use @xmath443 to denote the subspace of maximal vectors in @xmath209 of weight @xmath24 .
the following lemma plays a key role in this section . the proof will be given in the next section .
[ hu2lem ] let @xmath441 be an integer with @xmath22 $ ] and @xmath24 be a partition of @xmath109 with @xmath21
. then @xmath444 .
in particular , @xmath25 is a non - zero maximal vector of weight @xmath24 in @xmath18 .
moreover , the dimension of @xmath445 ( and hence of @xmath26 ) is independent of @xmath2 ; by construction , @xmath448 .
we fix a decomposition : @xmath449 where @xmath450 for each @xmath451 .
let @xmath452 be a fixed non - zero maximal vector of weight @xmath410 in @xmath413 . for each @xmath451 ,
let @xmath453 be a fixed non - zero maximal vector of weight @xmath410 in @xmath454 and let @xmath455 be the unique isomorphism which sends @xmath452 to @xmath453 .
then it is easy to see ( by definition and comparing dimensions ) that the following map @xmath456 is a left @xmath417-module isomorphism .
we claim that it is also a right @xmath17-module homomorphism .
it suffices to show that @xmath457 for each @xmath451 , @xmath416 and @xmath458 .
we write @xmath459 , where @xmath460 for each @xmath39 . then by definition , @xmath461 . by the commuting action between @xmath417 and @xmath17 , it is easy to check that @xmath462 is a left @xmath417-homomorphism from @xmath413 to @xmath272 . by direct verification , @xmath463 ,
hence @xmath464 for any @xmath465 .
since @xmath466 and @xmath467 , it follows that any non - zero map in @xmath468 must be injective .
we deduce that @xmath464 for any @xmath458 , as required .
this proves that @xmath469 is a right @xmath17-module homomorphism as well . by lemma [ keylem1 ] ,
it is enough to show that @xmath472 since @xmath265 has a weyl filtration such that each section is isomorphic to some @xmath473 with @xmath269 and @xmath474 has a good filtration such that each section is isomorphic to some @xmath475 with @xmath476 , it follows that @xmath477 hence @xmath478 which implies that @xmath479 as required .
first , by lemma [ jakey ] 2 ) , we have that @xmath484 note that the character @xmath485 of @xmath23 is given by the weyl character formula .
in particular , @xmath486 is independent of @xmath2 .
note also that the character @xmath487 is equal to @xmath488 which is also independent of @xmath2 .
so the filtration multiplicities @xmath489 is nothing but the coefficient of the weyl character @xmath490 in the expansion of @xmath488 into a linear combination of weyl characters .
in particular , @xmath491,\ ] ] which is independent of @xmath2 .
since the set @xmath116 is also independent of @xmath2 , we conclude that @xmath492 is independent of @xmath2 .
moreover , by the construction of @xmath493 , it is clear that @xmath494 let @xmath353 be an integer with @xmath507 . by the properties of the filtration ( [ filtra1 ] ) and the filtration ( [ filtra2 ] ) we have constructed and lemma [ bp ] 4 ) ,
it is clear that @xmath508 .
it follows that the natural embedding @xmath509 becomes an equality . by lemma [ jakey2 ]
, we see that the following map @xmath510 is an embedding .
it follows that @xmath511 for each @xmath507 . in particular
, @xmath512 induces a natural map from @xmath513 to @xmath272 .
recall that for each @xmath408 , @xmath514 .
in particular , @xmath515 for each @xmath407 . for each @xmath407 , we define @xmath516 .
then @xmath517 is a maximal vector of weight @xmath410 in @xmath265 and @xmath518 ( cf .
lemma [ hu2lem ] and proposition [ mainprop2 ] ) .
it is clear that @xmath519 .
let @xmath520 be a @xmath227-submodule of @xmath513 such that @xmath521 .
applying lemma [ bp2 ] , we deduce that @xmath522 .
therefore , combining with our induction hypothesis , we can find a @xmath227-submodule @xmath523 of @xmath18 such that @xmath524 by the proof of proposition [ mainprop2 ] , we can deduce that @xmath525 . as a consequence
, we get that @xmath526 since @xmath527 , it follows that @xmath528 , as required . with theorem [ keystep3 ] in hand , theorem [ mainthm0 ] and [ mainthm1 ]
follow almost trivially from some standard arguments in commutative algebras .
for the reader s convenience , we include the details below . by theorem [ keystep3 ] and lemma [ jakey ] , we know that @xmath102 has a good filtration .
note that @xmath18 has a good filtration too . applying ( * ? ? ?
* part ii , 4.17 ) , we deduce that both @xmath119 and @xmath530 have a good filtration .
this proves theorem [ mainthm1 ] and the first statement of corollary [ maincor15 ] .
it is clear that @xmath99 is a free @xmath80-module of finite rank @xmath531 .
since @xmath532 , it follows that @xmath533 . by theorem [ keystep3 ] , we know that @xmath482 is independent of @xmath2 and hence @xmath534 .
thus the canonical surjection @xmath535 must be an isomorphism .
now let @xmath536 be an arbitrary field and @xmath537 be the algebraic closure of @xmath536 . recall that @xmath538 , @xmath539 .
it is clear that the canonical map @xmath540 is an isomorphism .
it follows that @xmath541 .
in particular , the first statement of corollary [ maincor1 ] also follows . for any commutative @xmath80-algebra @xmath536 which is a field
, the canonical surjection @xmath542 must be an isomorphism .
for any commutative @xmath80-algebra @xmath95 , let @xmath536 be a field such that @xmath543 for some maximal ideal @xmath70 of @xmath95 .
we have the following commutative diagram of maps : @xmath544 where @xmath545 are all canonical maps .
since @xmath546 is an isomorphism , it follows that the canonical surjection @xmath547 must be injective and hence an isomorphism .
this proves the first isomorphism in part 2 ) of theorem [ mainthm0 ] . using the same argument as before , we know that @xmath548 note that @xmath549 . in order to show that @xmath550 is a free @xmath80-module of rank @xmath551 , we consider the following commutative diagram of maps : @xmath552^{\quad\iota\otimes\operatorname{id } } \ar[d]^{\wr } & v_{{\mathbb z}}^{\otimes n}\otimes_{{\mathbb z}}f \ar@{>>}[r]^{\pi\otimes\operatorname{id}\quad\quad\,}\ar[d]^{\wr } & v_{{\mathbb z}}^{\otimes n}/v_{{\mathbb z}}^{\otimes n}{\mathfrak{b}}_n^{(f)}\otimes_{{\mathbb z}}f \ar@{>>}[d]^{\theta}\\ v_{f}^{\otimes n}{\mathfrak{b}}_n^{(f ) } \ar@{^{(}->}[r ] & v_{f}^{\otimes
n } \ar@{>>}[r ] & v_{f}^{\otimes n}/v_{f}^{\otimes n}{\mathfrak{b}}_n^{(f)},}\ ] ] where @xmath553 denotes the natural injection @xmath554 , @xmath555 denotes the natural projection @xmath556 . by diagram chasing , it is easy to see that the natural surjection @xmath557 is an injection and hence an isomorphism . in particular , @xmath558 we claim that @xmath559 must be a free @xmath80-module .
in fact , suppose this is not the case , then @xmath559 must contain a non - zero @xmath560-torsion element for some prime number @xmath560 .
it follows that @xmath561 which is a contradiction .
this proves our claim . as a consequence ,
we conclude that @xmath99 is a pure @xmath80-submodule of @xmath100 . now using a commutative diagram similar to ( [ cd1 ] ) ( replacing @xmath99 with @xmath101 )
, we can argue as before that for any commutative @xmath80-algebra @xmath95 , the canonical map @xmath562 is always an isomorphism .
this completes the proof of theorem [ mainthm0 ] .
it remains to prove the second statement of corollary [ maincor1 ] .
let @xmath2 be an algebraically closed field .
recall the bimodule isomorphism @xmath222 introduced in the paragraph below ( [ asso ] ) .
@xmath222 induces a map @xmath563 from @xmath74 to @xmath564 . the second equality in ( [ asso ] ) and the definition of @xmath74
imply that the image of @xmath563 is contained in @xmath565 .
therefore , @xmath563 is a bimodule homomorphism from @xmath81 to @xmath566 .
we claim that @xmath563 is injective .
let @xmath567 such that @xmath568 .
now for any @xmath569 , we can write ( by applying corollary [ maincor25 ] ) @xmath570 where @xmath571 . using the assumption that @xmath568 and applying the two equalities in ( [ asso ] ) we get that @xmath572 since the bilinear form @xmath573 on @xmath18 is non - degenerate , it follows that @xmath574 .
hence @xmath563 is injective .
this proves our claim . to emphasize the base field
, we use the notations @xmath575 , @xmath576 . note that we have already known that the dimension of @xmath291 is independent of the field @xmath2 .
the space @xmath74 can be identified as the solution space of a homogeneous system of linear equations whose coefficients are all defined over @xmath80 . in particular , @xmath577 by ( * ? ? ?
* ( 10.3.14 ) ) , we know that @xmath578 by the injectivity of @xmath563 , we know that @xmath579 it follows that @xmath580 .
therefore , @xmath563 must be an isomorphism .
this completes the proof of corollary [ maincor1 ] .
note that the space @xmath55 can be identified as the solution space of a homogeneous system of linear equations whose coefficients are all defined over @xmath80 .
this implies that @xmath582 . now applying corollary [ maincor1 ] and ( [ embed ] )
, we prove the corollary .
suppose that @xmath954 . by the proof of lemma
[ jakey2 ] , @xmath102 is a sum of some submodules @xmath955 such that @xmath956 for each @xmath130 .
it follows that @xmath957 since @xmath265 is a tilting module , @xmath265 has a weyl filtration . for each @xmath958 , @xmath959 only if @xmath960 for some integer @xmath961 $ ] . by lemma [ bp ]
, we get that @xmath965\\ 0\leq s < f}}\bigl(v^{\otimes
n-2f}:\delta({\lambda})\bigr ) \bigl(v^{\otimes
n}/v^{\otimes n}{\mathfrak{b}}_n^{(f)}:\nabla({\lambda})\bigr)\\ & = 0 , \end{aligned}\ ] ] which is a contradiction .
we have the following exact sequence of maps : @xmath968 since @xmath18 has a weyl filtration , and by theorem [ mainthm2 ] , @xmath102 has a good filtration , it follows that @xmath969 .
this implies that the canonical map @xmath966 is surjective .
this proves 1 ) .
the above exact sequence implies that @xmath970 since @xmath18 has a weyl filtration as well as a good filtration , and @xmath102 has a good filtration , and both the character formula of @xmath18 and of @xmath291 are independent of @xmath2 , it follows that @xmath971 which is independent of @xmath2 .
note that @xmath972 so 2 ) also follows .
* proof of theorem [ mainthm2 ] : * let @xmath973 denote the natural @xmath2-algebra homomorphism : @xmath974 .
then @xmath975 in view of lemma [ lm42 ] , it suffices to show that @xmath973 is surjective .
we consider the following commutative diagram of maps : @xmath976^{\varphi_k}\ar[d]_{\varphi''_k } & \operatorname{end}_{kg}\bigl(v^{\otimes n}\bigr ) \ar@{>>}[d]^{\theta_1}\\ \operatorname{end}_{kg}\bigl(v^{\otimes n}/v^{\otimes
n}{\mathfrak{b}}_n^{(f)}\bigr ) \ar[r]_{\iota_1\,\,\,\quad}^{\sim\,\,\quad } & \operatorname{hom}_{kg}\bigl(v^{\otimes n},v^{\otimes n}/v^{\otimes n}{\mathfrak{b}}_n^{(f)}\bigr),}\ ] ] by @xcite , the top horizontal map is surjective . by lemma [ lm42 ]
, @xmath977 is surjective . by lemma [ hom0 ]
, @xmath978 is an isomorphism .
it follows that @xmath973 must be surjective , as required .
let @xmath981 be the symplectic schur algebra . by theorem [ thm11 ]
, there is a natural surjection @xmath982 . the action of @xmath983 on @xmath83 factors through an action of @xmath984 .
it follows that @xmath985 let @xmath986 be the algebraic closure of @xmath2 . by @xcite and @xcite , @xmath987 .
it follows that @xmath988 in particular , @xmath989 hence @xmath979 is independent of the infinite field @xmath2 .
note that @xmath990 .
it follows that @xmath991 from which the proposition follows immediately .
let @xmath8 be an integer with @xmath204 $ ] and @xmath2 an infinite field .
proposition [ mainprop ] proves one side of the brauer - schur - weyl duality between @xmath117 and @xmath118 on @xmath119 .
we conjecture that the other side of the brauer - schur - weyl duality is also true .
that is , the dimension of the endomorphism algebra @xmath998 is independent of @xmath2 and the natural @xmath2-algebra homomorphism @xmath999 is also surjective .
the purpose of this section is to give a proof of lemma [ hu2lem ] as well as theorem [ mainthm3 ] 2 ) .
the key step in our proof is to show that ( for any @xmath583 ) @xmath584 and @xmath585 has a weyl filtration as a @xmath417-module .
the proof makes use of lusztig s theory of canonical basis and based modules . to this end
, we have to work in a quantized setting .
let @xmath586 be an indeterminate over @xmath80 .
let @xmath587 $ ] be the ring of laurent polynomials in @xmath586 .
let @xmath588 be the specialized birman - murakami - wenzl algebra ( @xcite , @xcite ) . by definition , it has generators @xmath589 which satisfy the following relations : 1 .
@xmath590 , for @xmath591 , 2 .
@xmath592 , for @xmath591 , 3 .
@xmath593 , for @xmath594 , 4 .
@xmath595 , for @xmath596 , 5 .
@xmath597 , for @xmath594 , 6 .
@xmath598,for @xmath594 , 7 .
@xmath599 , for @xmath591 . 8 .
@xmath600,for @xmath594 .
let @xmath601 be the quantized enveloping algebra of @xmath602 over @xmath603 .
let @xmath604 be the chevalley generators of @xmath605 .
then it is a hopf algebra with coproduct @xmath606 , counit @xmath607 and antipode @xmath608 defined on generators by @xmath609 where @xmath610 we regard @xmath80 as an @xmath611-algebra by specializing @xmath586 to @xmath235 . for each @xmath611-module @xmath301 and @xmath612 , we set @xmath613 and @xmath614 .
we call @xmath615 the specialization of @xmath188 at @xmath616 .
let @xmath617 be lusztig s @xmath611-form in @xmath618 .
let @xmath619 be the free @xmath611-module spanned by @xmath153 .
it is well - known that there are natural commuting actions ( cf .
* section 2,3 ) ) of @xmath617 and @xmath588 on @xmath620 . if we specialize @xmath586 to @xmath235 , then each @xmath621 ( resp . , @xmath622 ) specializes to @xmath623 ( resp .
, to @xmath147 ) , and @xmath624 will become the brauer algebra @xmath4 in this paper .
moreover , the action of @xmath624 on @xmath620 becomes the action of @xmath4 on @xmath100 in this paper .
we recall that the representation @xmath625 of @xmath588 on @xmath620 is defined on generators as follows : @xmath626 for all @xmath627 , where @xmath628 and each @xmath234 is the matrix unit ( i.e. , @xmath629 for each @xmath212 ) .
let @xmath630 be the iwahori
hecke algebra associated to the symmetric group @xmath200 , defined over @xmath611 and with parameter @xmath586 . by definition , @xmath630 has generators @xmath631 which satisfy the following relations : @xmath632 let @xmath633 . for each composition @xmath24 of @xmath28
, we use @xmath634 to denote the hecke algebra over @xmath603 associated to the young subgroup @xmath434 . for each @xmath635
, we use @xmath636 to denote the minimal integer @xmath35 such that @xmath637 ; in that case , we call @xmath638 a reduced expression of @xmath639
. we define @xmath640 if @xmath641 with @xmath642 .
it is well - known that this is well defined , i.e. , independent of the choice of the reduced expression .
let @xmath643 be the free @xmath611-submodule of @xmath619 generated by @xmath644 .
we recall that the representation @xmath645 of @xmath630 on @xmath646 is defined on generators as follows : @xmath647 where @xmath648 let @xmath24 be a partition with @xmath21 .
recall the definition of @xmath649 in definition [ zdfn ] .
we have that @xmath650 for each @xmath651 .
let @xmath668 .
let @xmath669 be the quantized enveloping algebra of the general linear lie algebra @xmath670 over @xmath603 .
let @xmath671 be lusztig s @xmath611-form in @xmath672 .
there is a natural representation of @xmath671 on @xmath643 and hence on @xmath646 ( cf .
@xcite ) . by (
* ( 27.3 ) ) , the @xmath673-module @xmath674 is a based module .
there is a bar involution @xmath675 which is defined on @xmath674 . for any integers @xmath676 ,
there is a unique element @xmath677 , such that 1 . @xmath678 , and 2 .
@xmath679 is equal to @xmath680 plus a linear combination of elements @xmath681 with @xmath682 and with coefficients in @xmath683 $ ] , where @xmath684 is a partial order root system . ]
defined in ( * ? ? ?
* ( 27.3.1 ) ) .
let @xmath685 .
then @xmath686 is the set of lusztig s canonical bases of @xmath674 .
in particular , @xmath686 is an @xmath611-basis of @xmath646 .
similarly , @xmath687 is a based module as a @xmath601-module .
let @xmath250 be the set of canonical bases of @xmath209 .
then @xmath250 is actually an @xmath611-basis of @xmath688 .
recall that @xmath689 is the set of dominant @xmath240-weights appeared in @xmath209 ( cf .
( [ pif ] ) ) . by (
* ( 27.2.1 ) ) , there is a partition @xmath690.\ ] ] for each @xmath691 , we set @xmath692:=\bigsqcup_{{\lambda}\leq\mu\in\pi_0}b[\mu],\quad b[>\!{\lambda}]:=\bigsqcup_{{\lambda}<\mu\in\pi_0}b[\mu].\ ] ] let @xmath693_{{\mathscr a } } , m[>\!{\lambda}]_{{\mathscr a}}$ ] be the @xmath611-submodule of @xmath694 generated by the canonical basis elements in @xmath695 , b[>\!{\lambda}]$ ] respectively .
by @xcite , both @xmath693_{{\mathscr a}}$ ] and @xmath696_{{\mathscr a}}$ ] are @xmath697-stable and we have that @xmath698_{{\mathscr a}}/m[>\!{\lambda}]_{{\mathscr a}}\cong \bigl(\delta({\lambda})_{{\mathscr a}}\bigr)^{\oplus n_{{\lambda}}},\ ] ] for some @xmath699 .
moreover , the canonical image of each @xmath700 $ ] in @xmath693_{{\mathscr a}}/m[>\!{\lambda}]_{{\mathscr a}}$ ] is mapped to a canonical basis element of some direct summand @xmath701 in the right - hand side of ( [ filtra3 ] ) .
recall that a weight of @xmath669 is identified as an element @xmath702 by setting @xmath703 for @xmath390 ; in this case , @xmath383 is said to be dominant if @xmath704 .
for any two weights @xmath705 of @xmath669 , we write @xmath706 if @xmath707 for some @xmath708 .
let @xmath709 be the set of dominant weights of @xmath669 appeared in @xmath710 .
then @xmath711 is nothing but the set of partitions of @xmath28 with no more than @xmath70 parts . for each @xmath712 , let @xmath713^{hi}$ ] be as defined in @xcite . using @xmath711 and the partial order @xmath714 , one can define the subsets @xmath715 , \hat{b}[>\!{\lambda}]$ ] and the @xmath611-submodule @xmath716_{{\mathscr a } } , \hat{m}[>\!{\lambda}]_{{\mathscr a}}$ ] in a similar way as before .
we set @xmath717^{hi},\quad { b}^{hi}:=\bigsqcup_{{\lambda}\in { \pi}_{0}}{b}[{\lambda}]^{hi}.\ ] ] let @xmath718 be a partition of @xmath28 satisfying @xmath719 and @xmath720 .
we define @xmath721 which is ( by definition ) a canonical basis element in @xmath686 . for each integer @xmath723 ,
let @xmath724 denote the kashiwara operators as defined in @xcite .
since we are following the notations and definitions of coproduct in lusztig s book , the actions of kashiwara operators on the tensor product of crystal bases are given by the following formulae ( cf .
@xcite ) @xmath725 where @xmath726 , @xmath727 are the crystal bases of the @xmath669-modules @xmath728 respectively , and @xmath729 let @xmath723 and @xmath730 .
recall that @xmath731 to prove the lemma , it suffices to show that for each @xmath390 , @xmath732 but this follows from a direct verification ( using the above tensor product rule ) and an inductive argument .
we write @xmath741 , where @xmath742 . by (
* theorem 27.3.2 ) , there is a bar involution @xmath743 on @xmath666 , and for each @xmath180 , there is a unique canonical basis element @xmath744 satisfying : we claim that @xmath751 for any @xmath752 .
we use induction on @xmath28 .
suppose that our claim is true when @xmath28 is replaced by @xmath753 .
let @xmath754 , @xmath755 be the bar involutions defined on @xmath756 , @xmath757 respectively .
let @xmath222 be the quasi-@xmath758-matrix of @xmath601 introduced in @xcite .
we shall mainly follow the notations in @xcite without any further explanation , except that we use @xmath586 to denote the quantum parameter instead of @xmath409 in lusztig s book .
we define @xmath759 in a similar way , one can define the subalgebras @xmath760 , @xmath761 of @xmath669 using the chevalley generators of @xmath669 .
it is well - known that both @xmath762 and @xmath760 can be presented by their chevalley generators and quantum serre relations ( @xcite ) . using the pbw bases ( @xcite ) for @xmath762 and @xmath760 , it is easy to see that @xmath760 can be identified with the subalgebra of @xmath763 generated by @xmath764 .
a similar statement holds for @xmath761 and @xmath765 .
let @xmath766 be defined as in @xcite and @xmath767 be a @xmath603-basis of @xmath766 such that @xmath768 is a basis of @xmath769 for any @xmath770 .
let @xmath771 be the basis of @xmath766 dual to @xmath767 under the bilinear form @xmath772 introduced in @xcite .
then @xmath773 , where @xmath774 applying ( * ? ? ?
* ( d ) ) , we see that @xmath771 is homogeneous too .
moreover , @xmath775 and @xmath776 are bases of @xmath761 and @xmath760 respectively .
recall that @xmath777 is a part of the canonical bases of the @xmath601-module @xmath778 .
we refer the reader to ( * ? ? ?
* section 6 ) for the description of the whole canonical bases of @xmath778 and the action of @xmath601 on @xmath778 .
the important facts that we need here are @xmath779 and @xmath780 for any @xmath781 . for any @xmath782 and any @xmath783
, we see that @xmath784 must appear in every monomial occurring in @xmath785 .
in particular , this implies that @xmath786 since @xmath787 .
now we have that @xmath790 on the other hand , by construction , we have that @xmath791v_{j_1}\otimes\cdots\otimes v_{j_n}.\ ] ] suppose that @xmath792 . then we can find an @xmath793 , such that @xmath794v_{j_1}\diamond_c\cdots\diamond_c v_{j_n},\ ] ] where @xmath795 $ ] . applying @xmath743 to both sides of the above equality
, we get a contradiction to ( [ bar1 ] ) .
this proves that @xmath796 and the first statement of the lemma follows .
the second statement follows from the definition of @xmath797^{hi}$ ] ( see @xcite ) , the facts that @xmath798 for any @xmath390 and the action of the kashiwara operators @xmath799 are compatible with the natural embedding @xmath800 .
finally , note that for any @xmath801 $ ] , @xmath801^{hi}$ ] if and only if the canonical image of @xmath802 in @xmath803_{{\mathscr a}}/m[>\!\mu]_{{\mathscr a}}$ ] is mapped to a highest weight vector of some direct summand @xmath701 in the right - hand side of ( [ filtra3 ] ) .
it follows that @xmath804_{{\mathscr a}}=\sum_{\substack{b\in b[\nu]^{hi}\\ \mu<\nu\in\pi_0}}\mathbb{u}_{{\mathscr a}}(\mathfrak{sp}_{2m})b,\quad \hat{m}[>\!\mu]_{{\mathscr a}}=\sum_{\substack{b\in\hat{b}[\nu]^{hi}\\ \mu<\nu\in\hat\pi_0}}\mathbb{u}_{{\mathscr a}}(\mathfrak{gl}_{m})b.\ ] ] now the third statement follows from 2 ) and the fact for any @xmath805 , @xmath806 and every chevalley generator of @xmath618 acts on @xmath409 in the same way as the corresponding chevalley generator of @xmath807 .
now we are in position to prove the key lemma in this section .
the most difficult part is the second statement of the following lemma .
the main idea of its proof is to show that there exists a canonical basis element @xmath808^{hi}$ ] of the @xmath601-module @xmath666 such that @xmath809 always appears with coefficient @xmath235 in the linear expansion of @xmath810 into the specialization at @xmath616 of the canonical basis elements of @xmath666 .
the element @xmath802 will be identified with a canonical basis element in the @xmath669-module @xmath710 and hence eventually identified with a parabolic kazhdan - lusztig basis of certain permutation module over a type @xmath231 hecke algebra .
it is well - known that @xmath812 ( cf .
in particular , @xmath813 .
note that @xmath814 .
thus the statement 1 ) follows from lemma [ lmhx ] .
since @xmath815 , it follows easily that @xmath816 .
it is also well - known ( cf .
@xcite ) that @xmath817 as a right @xmath59-module .
in particular , @xmath818 which is independent of the field @xmath2 .
this proves the statement 3 ) .
it remains to prove the statement 2 ) .
we divide the proof into three steps : _ step 1 .
_ we claim that there exists a canonical basis element @xmath819^{hi}$ ] , such that @xmath820^{hi}}{c}'_{b'}b'\!\!\downarrow_{q=1}\!\!\!\pmod{\hat{m}[>\!{\lambda}]_{{\mathbb z}}},\ ] ] where @xmath821 for each @xmath822 . for each composition @xmath383 of @xmath28 ,
let @xmath823 be the corresponding weight subspace of @xmath646 ( with respect to the action of @xmath671 ) .
it is well known that there is a right @xmath630-module isomorphism : @xmath824 where @xmath825 is the one dimensional representation of @xmath826 which is defined on generators by @xmath827 for each @xmath828 .
let @xmath300 be the anti - linear involution on @xmath829 which is defined on generators by @xmath830 for each @xmath143 . by the main result of @xcite
, @xmath300 naturally induces an anti - linear involution on @xmath831 so that one can define the parabolic kazhdan - lusztig bases @xmath832 of @xmath831 .
recall that @xmath831 also has a standard basis @xmath833 . for each @xmath834
, we have that @xmath835\hat{y}_{{\lambda}}\hat{t}_{d'}},\ ] ] where
@xmath836 " is the usual bruhat order defined on the symmetric group @xmath200 .
we identify @xmath837 with @xmath838 via the isomorphism @xmath839 .
it follows from ( * ? ? ?
* theorem 2.5 ) that the bases @xmath832 coincide with the canonical bases of @xmath710 which are of weight @xmath24 .
here one should understand the notations @xmath840 in this paper as the notations @xmath841 in @xcite .
let @xmath842 such that @xmath843 .
for each @xmath834 , @xmath844 . note that if @xmath845 , then @xmath846 .
we want to use the isomorphism @xmath839 and ( [ can1 ] ) to express @xmath810 into a linear combination of the specializations at @xmath616 of some canonical basis elements . to this end
, we first express @xmath810 into a linear combination of standard basis elements .
let @xmath847 be the longest element in the young subgroup @xmath848 of @xmath200 .
note that @xmath849 is row - standard , which implies that @xmath850 . since @xmath851
, we have that @xmath852 . note
also that @xmath853 .
it follows that @xmath854 where @xmath855 for each @xmath856 and @xmath857 . since @xmath858 for any @xmath859
, we deduce that @xmath860 only if @xmath861 .
that is , @xmath862 recall that in this paper @xmath200 acts on @xmath18 by sign permutation action .
thus , @xmath863 , @xmath864 . note that @xmath865 , @xmath866 for each @xmath834 .
it follows from ( [ can1 ] ) , ( [ can2 ] ) and the isomorphism @xmath839 that @xmath867 where @xmath868 .
we set @xmath869 and write @xmath718 , where @xmath870 and @xmath871 .
then it is easy to check that @xmath872 which implies that @xmath873 by lemma [ lmkey41 ] .
hence @xmath874^{hi}$ ] as @xmath802 is a weight vector of weight @xmath24 .
note that @xmath875 only if @xmath822 is a weight vector of weight @xmath24 . on the other hand ,
if @xmath876 $ ] and @xmath822 is a weight vector of weight @xmath24 then we must have that @xmath24 is a weight of the weyl module of @xmath669 associated to @xmath383 , which implies that @xmath877 .
therefore , we can rewrite ( [ expan1 ] ) as @xmath820^{hi}}{c}'_{b'}b'\!\!\downarrow_{q=1}\!\!\!\pmod{\hat{m}[>\!{\lambda}]_{{\mathbb z}}},\ ] ] where @xmath878^{hi}$ ] , @xmath821 for each @xmath822 .
this proves our claim .
recall the equality ( [ filt6 ] ) we obtained in step 2 .
let @xmath700^{hi}$ ] be the canonical basis element we obtained in step 2 . using lemma 3.7 3 ) , we can get a sequence @xmath886 of dominant weights in @xmath689 and construct a weyl filtration of @xmath694 as follows : @xmath887 such that 1 . for each integer @xmath888
, @xmath889 is spanned by the canonical basis elements it contains ; 2 . for each integer @xmath888 , there is a @xmath697-module isomorphism : @xmath890 , such that if @xmath891 then the canonical image of @xmath822 in @xmath892 is mapped either to @xmath237 or to a canonical basis element of @xmath893 ; 3 .
@xmath894 only if @xmath404 ; 4 .
@xmath895^{hi}\setminus\{b\}\subseteq m^{{\mathscr a}}_{(i_0 - 1)}$ ] , where @xmath896 is the unique integer such that @xmath897 . in particular , @xmath898 as @xmath808^{hi}$ ] .
thus we can write @xmath899 , where @xmath900 , such that for each @xmath888 , the elements in the subset @xmath901 form an @xmath611-basis of @xmath889 . moreover , @xmath902 specializing at @xmath616 , we get a @xmath903-submodules filtration of @xmath100 .
we define a new @xmath903-submodules filtration of @xmath100 as follows : @xmath904 where @xmath905 we claim that ( [ filtra5 ] ) is a weyl filtration of @xmath906 .
in fact , to prove this claim , it suffices to show that for each @xmath907 , @xmath908 is isomorphic to some weyl module .
let @xmath536 be any field which is an @xmath80-algebra .
let @xmath909 .
replacing @xmath80 by @xmath536 in ( [ filtra5 ] ) , we get a @xmath910-submodules filtration of @xmath911 as follows : @xmath912 recall that @xmath810 is a maximal vector of weight @xmath24 . using ( [ filt6 ] ) , ( [ generator ] )
, together with the third and the fourth properties of the filtration ( [ filtra4 ] ) , we see that @xmath913 by the second property of the filtration ( [ filtra4 ] ) , we know that @xmath914 it follows that each @xmath915 must be a homomorphic image of @xmath916 , where @xmath917 noting that @xmath918 and comparing the dimensions , we deduce that the natural surjection from @xmath916 onto @xmath915 must be an isomorphism .
in particular , @xmath919 since @xmath810 is a maximal vector of weight @xmath24 , we have a natural surjection from @xmath920 onto @xmath921 , which induces a surjection from @xmath922 onto @xmath923 . on the other hand
, we also have a natural surjection from @xmath923 onto @xmath924 .
it follows again by comparing dimensions that the natural surjection from @xmath922 onto @xmath923 is an isomorphism for any field @xmath536 .
hence the natural surjection from @xmath920 onto @xmath921 must be an isomorphism as well . by a similar argument
, we can show that the natural surjection from @xmath925 onto @xmath908 must be an isomorphism as well for each @xmath130 . as a result
, the @xmath881-module @xmath926 has a weyl filtration .
this proves our claim . finally , by taking @xmath927 and noting that @xmath928 ( cf .
@xcite ) whenever @xmath2 is an algebraically closed field , we deduce that @xmath584 , and @xmath929 has a weyl filtration .
this completes the proof of the statement 2 ) .
[ cor45 ] let @xmath441 be an integer with @xmath22 $ ] and @xmath24 a partition of @xmath109 satisfying @xmath21 .
then there exists an embedding @xmath930 such that @xmath931 has a weyl filtration .
in particular , @xmath932 in the remaining part of this section , we fix an integer @xmath933 $ ] and a partition @xmath24 of @xmath109 with @xmath934 . for simplicity
, we shall write @xmath25 instead of @xmath935 . by lemma [ lmhx ] and specializing @xmath586 to @xmath235 , we get that @xmath25 is a non - zero maximal vector of weight @xmath24 with respect to the action of @xmath227 on @xmath18 .
that is , @xmath936 as a consequence , @xmath937 . on the other hand , we have @xmath938
since @xmath18 has a good filtration , it follows from lemma [ bp ] that the dimensions of @xmath939 and hence of @xmath445 are independent of @xmath2 .
therefore , to complete the proof of lemma [ hu2lem ] as well as the second part of theorem [ mainthm3 ] , it suffices to prove that @xmath940 in view of the above discussion and lemma [ lmk40 ] , to prove ( [ equaz1 ] ) , it suffices to show that @xmath943 . using corollary [ cor45 ]
, we have an embedding @xmath930 such that @xmath931 has a weyl filtration .
therefore , we have the following commutative diagram of homomorphisms : @xmath944 & \operatorname{hom}_{kg}\bigl(v^{\otimes n-2 g } , v^{\otimes n}{\mathfrak{b}}_n^{(g)}\bigr ) \ar[r]^{\,\quad\sim}\ar[d]_{\beta } & \operatorname{hom}_{kg}\bigl(v^{\otimes n-2 g } , v^{\otimes n}\bigr ) \ar@{>>}[d]\\ 0 \ar[r ] & \operatorname{hom}_{kg}\bigl(\delta({\lambda } ) , v^{\otimes n}{\mathfrak{b}}_n^{(g)}\bigr ) \ar@{^{(}->}[r ] & \operatorname{hom}_{kg}\bigl(\delta({\lambda } ) , v^{\otimes n}\bigr),}\ ] ] where by lemma [ keylem1 ] the top horizontal map is an isomorphism and the fact that @xmath945 forces that the right vertical map is a surjection .
since the bottom horizontal map is an injection , it follows that the left vertical map @xmath946 must be a surjection and the bottom horizontal map must be an isomorphism . now applying lemma [ keylem1 ] , we get that @xmath947 is spanned by @xmath948 for @xmath306 , where @xmath949 is defined in ( [ tauf ] ) .
therefore , we can deduce that the subspace of maximal vectors of weight @xmath24 in @xmath9 is spanned by all @xmath950 , where @xmath306 .
hence @xmath951 .
this completes the proof of lemma [ hu2lem ] as well as the second part of theorem [ mainthm3 ] . | let @xmath0 be a @xmath1-dimensional symplectic vector space over an algebraically closed field @xmath2 .
let @xmath3 be the two - sided ideal of the brauer algebra @xmath4 over @xmath2 generated by @xmath5 , where @xmath6 $ ] .
let @xmath7 be the subspace of partially harmonic tensors of valence @xmath8 in @xmath9 . in this paper , we prove that @xmath10 and @xmath11 are both independent of @xmath2 , and the natural homomorphism from @xmath12 to @xmath13 is always surjective .
we show that @xmath7 has a weyl filtration and is isomorphic to the dual of @xmath14 as a @xmath15-@xmath16-bimodule . we obtain a @xmath15-@xmath17-bimodules filtration of @xmath18 such that each successive quotient is isomorphic to some @xmath19 with @xmath20 , @xmath21 and @xmath22 $ ] , where @xmath23 is the co - weyl module associated to @xmath24 and @xmath25 is an explicitly constructed maximal vector of weight @xmath24 . as a byproduct ,
we show that each right @xmath17-module @xmath26 is integrally defined and stable under base change . |
observations at 850 , initially with the submillimeter common user bolometer array ( scuba ) and subsequently with the max planck millimeter bolometer ( mambo ) , have revealed galaxies that are strong sources of dust emission but faint at visible wavelengths ( smail , ivison & blain 1997 , hughes et al .
1998 , barger et al .
1998 , eales etal 1999 , clements et al . 2004 ) .
the scuba galaxy populations are important because they seem to represent the most luminous star - forming galaxies during the epoch when star formation rates were highest .
the space density of submillimeter galaxies is orders of magnitude greater than that of @xmath2@xmath3 galaxies at low redshift ( blain et al .
2002 ) , implying that strong evolution has greatly altered the submillimeter galaxy populations since @xmath4 . moreover , some or all of the scuba population may be the progenitors of today s massive elliptical galaxies .
the stars in nearby elliptical galaxies and in the bulges of disk galaxies are old and tend to be metal rich . at high redshifts ,
there ought to be a significant population of progenitor galaxies with star formation rates ( sfrs ) high enough to produce today s population of ellipticals and bulges ( lilly et al .
no such population has been detected in deep optical surveys , but the scuba galaxies , or some of them , could represent the long - sought progenitors .
understanding these so - called `` scuba galaxies '' and their place in cosmic evolution , however , requires complementary observations at other wavelengths and in particular some way to determine redshifts .
observing scuba galaxies in visible light is difficult .
the dust responsible for the submillimeter emission obscures the primary luminosity sources , making the scuba galaxies quite faint at visible wavelengths . even identifying the correct visible counterpart to a given scuba galaxy is difficult because of the large beamsize ( 14fwhm at 850 @xmath0 m ) of the james clerk maxwell telescope and other submillimeter facilities .
the unavoidable result is that at visible and even near - infrared wavelengths , there are numerous potential counterparts in the scuba beam , and there is no easy way to decide which of the candidates is the true counterpart .
one successful approach to finding counterparts has been to use radio interferometric observations to refine the scuba positions and thus select a specific visible object .
chapman et al . ( 2003 , 2005 ) , for example , have used this approach and have shown that radio - identified scuba galaxies have a median spectroscopic redshift of @xmath5 . however , about 1/3 of scuba sources are not detectable with current radio observations ( chapman et al . 2005 ) and could represent a distinct population .
it is therefore worthwhile seeking alternate ways to identify counterparts .
this contribution describes a new program to recover submm - selected galaxies in the mid - infrared , so as to exploit the advantages this passband offers for better identifications of the counterparts to these important objects .
our submm sample is comprised of 23 scuba - detected objects in the 14@xmath1 field of the canada - uk deep submillimetre survey ( cudss ; lilly et al . 1999 ,
eales et al .
2000 , webb et al .
we base our infrared counterpart search on the very deep 8.0 @xmath0 m _ spitzer _
mosaic of the 14@xmath1 field .
although the _ spitzer _ observations ( described below ) include coextensive observations at a number of passbands from 3.6 to 160 @xmath0 m , we use sources drawn specifically from the 8.0 @xmath0 m mosaic , for the following reasons . first , 8.0 @xmath0 m observations sample rest - frame _ k_-band emission at typical scuba galaxy redshifts @xmath6
. they are therefore much less affected than shorter - wavelength observations by extinction due to the dust known to be abundant in galaxies that are actively forming stars ( schmitt et al . 2006 ) .
this rest - frame _ k_-band emission is also a good proxy for stellar mass ( cole et al .
2001 ) ; it is therefore more sensitive to total star formation history than shorter - wavelength observations that are relatively more affected by recent star formation .
second , high - redshift galaxies like those selected at submm wavelengths typically have red infrared colors that are well - suited to detection at 8.0 @xmath0 m ( egami et al . 2004 , huang et al .
2004 , frayer et al .
2004b ) , or at least they have red optical and near - infrared colors ( pope et al .
2005 , frayer et al .
2004a , borys et al .
2004 , ivison et al .
lockman hole galaxies were detected in 450 and 850 @xmath0 m scuba maps with stacking analyses at 5.8 and 8.0 @xmath0 m by serjeant et al .
( 2004 ) but not at 3.6 or 4.5 @xmath0 m .
in addition , although the 3.6 and 4.5 @xmath0 m _ spitzer
_ observations in the 14@xmath1 field are somewhat more sensitive than the 8.0 @xmath0 m mosaic , they tend to detect more of the blue , relatively numerous , low - redshift galaxies that are unrelated to the scuba - selected population .
the resulting contamination problem is of course more severe at shorter wavelengths , e.g. , _ k_-band . at 8.0 @xmath0
m we avoid to a great degree the problems of false associations that affect these other bands , without missing sources : all potential 8.0 @xmath0 m counterparts we considered are also detected in the 3.6 @xmath0 m mosaic .
third , the 8.0 @xmath0 m mosaics themselves are very deep down to a 5@xmath7 point source sensitivity of 5.8 @xmath0jy .
we used two very simple models of star formation history to calculate that this depth is easily sufficient to detect typical scuba galaxies .
both models ( one a constant star formation rate model , the other a burst ) give approximately equivalent results and suggest that at redshifts of @xmath8 the 8.0 @xmath0 m mosaics will detect star formation rates of @xmath9 and @xmath10@xmath11 yr@xmath12 , respectively .
this is comfortably below the rates found for submm galaxies ( up to roughly @xmath13@xmath11 yr@xmath12 , e.g. , smail et al .
2004 , yun & carilli 2002 , lilly et al .
1999 ) , so we can expect _ a priori _ that their 8.0 @xmath0 m counterparts are detected in the _ spitzer _ data .
this paper describes the _ spitzer _ observations , the 8.0 @xmath0 m counterpart identification procedure , and the inferences to be drawn from the panchromatic seds of infrared - selected scuba counterparts .
a companion paper ( dye et al . 2006 ) describes the inverse process , whereby we determine the average submillimeter emission at the positions of the _ spitzer _ sources found in the field .
the next section describes the _ spitzer_/irac and mips observations and presents infrared images at the known locations of the scuba galaxies .
we then outline the scheme used to identify the most - likely infrared counterparts of the submillimeter galaxies in section [ sec : ids ] .
the infrared properties of the likely counterparts are analyzed in section [ sec : discussion ] .
the irac ( fazio et al . 2004 ) observations were carried out as part of spitzer guaranteed time observing program number 8 to deeply image the extended groth strip ( egs ) . the first visit to the field was in 2003 december .
the field was observed again during a second epoch in 2004 june at a different position angle . the irac exposures ( each of which covered a 512@xmath1412 field of view with @xmath15 pixels @xmath16 in size ) totaled 52 dithered 200 s exposures at 3.6 , 4.5 , and 5.8 @xmath0 m , together with 208 dithered 50 s exposures taken concurrently at 8.0 @xmath0 m at all positions in a @xmath17 map .
each dither cluster was separated from its neighbor by 290 , slightly less than the irac field of view . because the field was observed at position angles differing by roughly @xmath18 during the two epochs it was straightforward to remove most of the well - known instrumental artifacts ( multiplexer bleed , banding ) during mosaicing .
the resulting point - source sensitivity was 24.0 , 24.0 , 21.9 , and 22.0 mag ( ab , 5@xmath7 ) , in the 3.6 , 4.5 , 5.8 , and 8.0 @xmath0 m bands , respectively .
this corresponds to flux density levels of 0.9 , 0.9 , 6.3 , and 5.8 @xmath0jy in the four irac bands .
all magnitudes given in this work are on the ab system .
sources in the irac mosaics were identified and extracted using sextractor ( bertin & arnouts 1996 ) .
the software was configured with a 2.5@xmath7 detection threshold and a minimum object area of at least 5 pixels .
the result was a band - merged photometric catalog of objects for all four irac wavelengths ; photometry was performed using 3 diameter apertures centered at the positions computed at 3.6 .
all sources here are pointlike in the irac bands . to permit comparison with total magnitudes in the other bands , appropriate corrections were applied to obtain total magnitudes from the aperture magnitudes .
the density of 8.0 @xmath0 m sources brighter than 5.8 @xmath0jy is roughly 11 per square arcminute . the mips ( rieke et al .
2004 ) observations of the egs were carried out on 2004 june 19 and 20 .
the mips 24 @xmath0 m channel ( @xmath19 m ; @xmath20 m ) used a 128@xmath21128 si : as array with a pixel scale of 255 pixel@xmath12 , providing a field of view of @xmath2254 .
the 70 @xmath0 m channel ( @xmath23 m ; @xmath24 m ) uses a 32@xmath2132 ge : ga array with a pixel scale of 998 pixel@xmath12 , but one half of the array is not usable due to a high noise caused by a cabling problem .
the effective 70 @xmath0 m array size is therefore 32 @xmath2116 pixels , providing a field of view of 52@xmath2126 .
the area was observed in scan map mode at slow rate with scan legs 24 long .
this results in an integration time of 100 sec pixel@xmath12 per scan pass ( 10 frames@xmath2110 s ) at both 24 and 70 @xmath0 m . in scan mode images at these two wavelengths
are obtained simultaneously .
pairs of scan map observations , with a cross - scan offset of one full array ( 296 ) , were executed 6 times , with an offset of 21 between successive pairs .
the effective integration time is therefore @xmath251500 s ( 12 passes ) at 24 @xmath0 m for locations near the long centerline of the strip , decreasing to @xmath25700 s 5 from the centerline . at 70 @xmath0 m
, the integration time is only half as much due to the loss of half of the detector array .
the final scan maps cover an area 2.4@xmath2110 with an integration time @xmath26 sec pixel@xmath12 . a smaller 2@xmath216 strip in the center , encompassing most of the cudss field , was covered with an integration time of 13001600 seconds .
the data were reduced and mosaiced with the mips data analysis tool ( gordon et al .
the 24 @xmath0 m scan images were resampled and mosaiced at half the original instrument pixel scale ( 125 ) while at 70 @xmath0 m the original pixel scale was preserved .
the point spread functions in the mosaics have fwhms of 6 and 18 at 24 and 70 @xmath0 m , respectively .
the resulting 5@xmath7 point source sensitivities are respectively 70 @xmath0jy and 7 mjy at 24 and 70 @xmath0 m .
we have drawn from several pre - existing photometric catalogs in order to construct seds spanning as wide a wavelength range as possible for our counterparts .
for the u , b , v , and i band photometry , we used the canada - france deep field survey ( mccracken et al .
2001 , hereafter cfdf ) survey data .
cfdf magnitudes are calculated within 2.5 diameter apertures .
the 3@xmath7 cfdf limiting magnitudes are 27.71 , 26.23 , 25.98 , and 25.16 mag in u , b , v , and i , respectively . at r
we use the very deep image acquired at the subaru telescope with _ suprime _ ; our r - band photometry has a @xmath27 limit of 26.6 mag ( miyazaki 2006 , private communication ) . for z - band
we use the catalog constructed by brodwin et al .
( 2006 ) , converting vega to ab magnitudes by adding a constant 0.54 mag .
the limiting magnitude of the z - band data is 25.0 ( @xmath28 ) , within 2.5 diameter apertures .
the @xmath29-band photometry data are drawn from the observations described by webb et al .
( 2003 ) , converting to ab magnitudes using @xmath30 .
@xmath29-band magnitudes were measured within 3 apertures . because the @xmath29-band data were acquired at two facilities , the depth is nonuniform but it generally extends down to @xmath31 mag ( @xmath32 ) .
all visible and near - infrared sources considered here are small compared to the apertures used ; we therefore treat the aperture magnitudes as total magnitudes .
two of the submillimeter galaxies ( 14.13 and 14.18 ) have _ infrared space observatory _ ( iso ) 15 @xmath0 m fluxes reported by flores et al .
one of these ( 14.13 ) was observed using _
spitzer s _ _ infrared spectrograph _ ( irs ) in low - resolution mode by higdon
( 2004 ) , who also report a 16 @xmath0 m flux density from the irs peakup imager . finally
, we have also made use of the chandra x - ray observatory ( cxo ) source catalog compiled by nandra et al .
a montage showing our observations at r , 3.6 , 4.5 , 5.8 , 8.0 , and 24 @xmath0 m at the locations of each of the 23 scuba sources in the 14@xmath1 field is provided in figure [ fig : montage ] .
to avoid spurious sources we restrict our investigation to portions of the 14@xmath1 field covered by at least ten 8 @xmath0 m irac exposures . with this constraint
the irac coverage allows counterpart identifications for 20 of the 23 scuba galaxies in this field .
we identify irac counterparts of the scuba sources using a two - part approach , consisting first of a proximity search of the irac 8.0 @xmath0 m source catalog , followed by an examination of the infrared seds of all possible matches .
we first use a metric based on 8.0 @xmath0 m brightness and positional coincidence to select likely counterparts .
we restrict the search to irac objects within 10 of each scuba source , for several reasons .
the primary motivation is to adopt a ( conservative ) value slightly larger than the @xmath33 uncertainty in the scuba positions , so as not to miss any true counterparts that might lie at relatively large separations , while simultaneously avoiding such a large value that we include a large number of unassociated foreground objects in the search .
we find that a 10 search radius often selected just one ( or at most three ) candidates , and was therefore well - matched to our irac source density .
based on the source surface density down to the 5@xmath7 limit of our survey , on average one irac 8 @xmath0 m source should randomly reside within each 10 radius search area .
borys et al .
( 2004 ) found 11 counterparts within 7 of 19 submillimeter objects in the hubble deep field north , implying a substantial fraction may lie beyond this radius .
similarly , wang et al .
( 2004 ) identified five of their 20 x - ray counterparts and two of their six mid - infrared counterparts to submillimeter galaxies in the same field at separations greater than 7 .
by contrast lilly et al .
( 1999 ) found at least 90% of the visible - wavelength counterparts to the submillimeter galaxies in the cudss 14@xmath1 field are present within 7 , but their analysis includes only six sources .
our 10 criterion is also safely larger than the mostly - likely average 850 @xmath0m - infrared offset of 2 determined from the stacking analysis discussed by dye et al .
( 2006 ) .
all potential counterparts are listed in table [ 8micids ] . in seven cases ( see below ) a single , obvious best counterpart was identified . in instances where the identifications were not clear - cut ,
we applied a second criterion based on infrared color to refine the identifications .
specifically , we made use of @xmath29-band photometry from webb et al .
( 2003 ) as well as our own irac photometry to discriminate the most likely counterpart among multiple candidates . in doing so
, we made one important assumption .
we _ a priori _ assumed , based on spectroscopic studies ( chapman et al . 2003 , 2005 ) , that the true counterparts must reside at high redshift and that the seds must be consistent with this fact .
we applied color criteria developed by huang et al .
( 2004 ; see their table 2 ) , who use the @xmath29-to - irac colors to constrain a galaxy s redshift into local , intermediate , and high - redshift regimes .
those criteria are summarized as follows : a blue @xmath34 $ ] color corresponds to redshifts @xmath35 .
red @xmath34 $ ] color is consistent with either an intermediate ( @xmath36 ) or high redshift ( @xmath371.3 ) , depending on whether the mid - infrared @xmath38 - [ 4.5]$ ] color is blue or red , respectively . in section [ ssec :
notes ] we make frequent use of this technique to select the best ( highest - redshift ) candidate among multiple possibilities .
details of the identifications for these 20 accessible sources are provided in the following section .
seven sources have secure identifications : 14.1 , 14.3 , 14.13 , 14.17 , 14.18 , 14.19 , and 14.23 . of these , all but 14.17 and 14.23 have relatively small position uncertainties .
we have possible identifications for ten others : 14.2 , 14.4 , 14.7 , 14.8 , 14.10 , 14.12 , 14.14 , 14.15 , 14.20 , and 14.22 . because their infrared colors and r - band morphologies are not those of dusty , high - redshift galaxies , we regard both potential 8.0 @xmath0 m identifications for 14.11 as dubious and do not include it in subsequent analysis .
likewise 14.6 and 14.21 which have no significant 8.0 @xmath0 m counterpart within 10 of the scuba positions .
objects 14.1 , 14.3 , 14.13 , and 14.18 have 1.4 ghz positions available from fomalont et al .
a recalibration of the vla data by m. yun improved the effective beamwidth of this dataset to 4 fwhm and resulted in a radio identification for 14.19 ( webb et al .
2003 ) .
_ cudss 14.1_. our single 8.0 @xmath0 m identification is coincident with both the webb et al .
( 2003 ) @xmath29-selected candidate and the i - selected candidate from lilly et al .
( 1999 ) within the positional uncertainties .
there is also a mips 24 @xmath0 m source at this position . _
cudss 14.3_. our only 8.0 @xmath0 m candidate is coincident with the webb et al .
( 2003 ) candidate as well as the position observed by chapman et al .
2005 , who used keck to obtain a spectroscopic redshift of @xmath39 for this object .
this object is also an obvious mips 24 @xmath0 m source .
_ cudss 14.13_. for this object the best candidates at @xmath29 and 8.0 @xmath0 m agree within the positional uncertainties .
this object is also a strong mips 24 @xmath0 m source and was detected by iso at 7 and 15 @xmath0 m at the same position .
this is not surprising , because it is by far the brightest of all our irac candidates .
this object was also observed with the infrared spectrograph on the _ spitzer space telescope _
( higdon et al . 2004 ) and exhibits a continuum slope consistent with a seyfert 2 nucleus and no pah emission .
our best position is coincident with that observed by chapman et al .
( 2005 ) , who used keck to obtain a spectroscopic redshift of @xmath40 for this object ; this places it in the lowest quartile of the submm galaxy redshift distribution . _
cudss 14.18_. for this object our 8.0 @xmath0 m search selects the same candidate as webb et al
. found at @xmath29 .
this is identical to the i - selected candidate from lilly et al .
( 1999 ) as well . within the positional uncertainties , all three positions ( irac+@xmath29-selected+i - selected )
are also coincident with the iso 7 and 15 @xmath0 m detections .
the 8.0 @xmath0 m object is also a bright 24 @xmath0 m source . finally , the best irac position is coincident with that observed at keck by chapman et al .
( 2005 ) , who found a spectroscopic redshift of @xmath41 for this object .
this is a surprisingly low redshift compared to most scuba sources . _
cudss 14.19_. for this object our 8.0 @xmath0 m search selects a different best candidate than at @xmath29 .
the irac candidate position lies just 2 from the radio position and is coincident with significant 24 @xmath0 m emission .
we identify what we believe to be secure infrared identifications for two objects that lack radio counterparts , 14.17 and 14.23 .
_ cudss 14.17_. our 8.0 @xmath0 m search selects a single best candidate and disagrees with the @xmath29-selected candidate , which is not apparent in the mid - infrared mosaics .
the 8.0 @xmath0 m candidate falls just 08 from the iso detection and is detected by mips at 24 @xmath0 m .
thus despite the large offset ( 95 ) this identification appears secure .
_ cudss 14.23_. our 8.0 @xmath0 m search selects two candidates .
one ( b ) is fainter , blue in @xmath38-[4.5]$ ] , and offset 66 to the ne of the scuba position .
the other , brighter source is red in @xmath38-[4.5]$ ] and @xmath42 $ ] and is a strong source of 24 @xmath0 m emission .
it is offset only 2 from the scuba position . for these reasons
we prefer candidate a. it is the same as the @xmath29-selected candidate from webb et al .
we identify plausible infrared counterparts to ten sources , none of which has a radio identification ( but see discussion of 14.8 and 14.20 below ) .
_ cudss 14.2_. although we detect the webb et al .
( 2003 ) candidate ( 14.2a in table [ 8micids ] ) in all irac bands , we also find a closer , fainter irac object ( b ) only 21 from the reported scuba position .
however , candidate a is a source of mips 24 @xmath0 m emission and is much brighter at 8.0 @xmath0 m . for these reasons
we adopt a as the more likely counterpart to 14.2 .
the lilly et al . ( 1999 ) i - selected counterpart , which is distinct from both the @xmath29- and irac - selected positions , is not detected in the irac mosaic at 8.0 @xmath0 m .
_ cudss 14.4_. at 8.0 @xmath0 m there are three possible counterparts . of these , c is approximately coincident with the webb et al .
object c is also the faintest of the three candidates at 8.0 @xmath0 m and shows no significant 24 @xmath0 m emission .
it is blue in @xmath34 $ ] , suggesting this object resides at @xmath43 .
thus c seems unlikely to be the correct counterpart .
object a presents a red @xmath38 - [ 4.5 ] > 0
$ ] color , but the @xmath34 $ ] upper limit is not sufficiently sensitive to constrain the redshift .
object a is a mips 24 @xmath0 m source .
object b has no counterpart in the 24 @xmath0 m catalog , but this is most likely because it is blended with an even brighter object to the nw .
however object b presents a blue @xmath38 - [ 4.5]$ ] color , suggesting it lies at @xmath44 . thus candidate
a seems the most likely scuba counterpart among the irac sources , despite its relatively large separation ( 87 ) from the scuba position .
_ cudss 14.7_. there are two possible 8.0 @xmath0 m counterparts within the 10 search radius .
one of them ( a , the closer and fainter of the two ) coincides with the webb et al .
( 2003 ) @xmath29-selected candidate .
although both are detected in all irac bands , only the southern source ( b , the one more distant from the reported scuba position ) is unambiguously detected by mips .
the southern source exhibits multiple components in our deep r - band image .
it also has a redder @xmath45-[8.0]$ ] and @xmath46-[24.0]$ ] colors and is by far the brighter source at 8.0 @xmath0 m .
for these reasons we prefer candidate b. however , the scuba source is extended toward both irac objects .
in addition , both irac candidates have red @xmath29-to-4.5 @xmath0 m colors that suggest they both reside at @xmath47 .
thus while we prefer b we can not rule out the possibility that both objects may contribute to the submillimeter flux .
_ cudss 14.8_. our 8.0 @xmath0 m search selects two candidates ( a and b ) .
there is emission at the location of the @xmath29-selected candidate ( b ) , but it is present at less than 5@xmath7 significance .
candidate a is roughly 65 west of the best scuba position .
there is 24 @xmath0 m emission at this position but not at the position of candidate b. interestingly , the @xmath29-selected candidate may be a source of radio emission ( chapman et al .
2005 ) , and a spectroscopic redshift @xmath48 has been obtained for it .
the radio detection is regarded as somewhat marginal however ( @xmath49 , s. chapman 2005 , private communication ) , and so the association of candidate b with the scuba source is not established with certainty .
object b is detected at 3.6 and 4.5 @xmath0 m with flux densities of 5.4 and 5.6 @xmath0jy , respectively , but it exhibits no significant emission at 5.8 , 8 , or 24 @xmath0 m . because b has a blue mid - infrared sed and only a is significantly detected 8.0 @xmath0 m , we prefer a as the counterpart .
this object has red @xmath29-to-4.5 @xmath0 m colors that place it at @xmath47 , meaning it is possible that it could reside at the same redshift as candidate b as part of an interacting system .
_ cudss 14.10_. our 8.0 @xmath0 m search selects a different best candidate ( a ) than webb et al .
( b ) , 98 sw of the best scuba position . candidate a shows @xmath42 = 1.5 $ ] and @xmath38-[4.5]=-0.04 $ ] , suggesting it lies in the range @xmath36 .
candidate b , the @xmath29-selected object , though closer to the scuba position ( 44 away ) , is almost 1 magnitude fainter at k , is blue in @xmath38-[4.5]$ ] , and exhibits no significant emission at 5.8 , 8 , or 24 @xmath0 m . hence a is the more likely counterpart . _
cudss 14.12_. our 8.0 @xmath0 m diagnostic selects two candidates , neither of which is coincident with the @xmath29-selected candidate .
the @xmath29-selected candidate is not detected in the 8.0 @xmath0 m mosaic .
candidate a lies 55 w of the scuba coordinates and exhibits significant 24 @xmath0 m emission .
neither a nor b is a source of significant 5.8 @xmath0 m emission .
although b is brighter at 8.0 @xmath0 m than a , it exhibits no significant 24 @xmath0 m emission .
hence a is the more likely candidate , although the identification is by no means secure . _
cudss 14.14_. at 8.0 @xmath0 m there are two candidates .
neither is coincident with the @xmath29-selected candidate .
significant mips 24 @xmath0 m emission is associated only with candidate b. likewise only this candidate has significant detections in all four irac bands .
furthermore , only candidate b is red in @xmath38-[4.5]$ ] .
candidate b is our preferred counterpart on the basis of its relative brightness , the irac colors , and the presence of 24 @xmath0 m emission at its position . _
cudss 14.15_. there are two possible 8.0 @xmath0 m counterparts .
one of these ( a , the brightest ) is coincident with the best @xmath29-selected candidate and is a mips 24 @xmath0 m source .
candidate b , however , exhibits comparable 24 @xmath0 m emission , has a red @xmath38-[4.5]$ ] color , and has an sed that peaks at 5.8 @xmath0 m , suggestive of a galaxy at @xmath50 .
although a is brighter at 8.0 @xmath0 m , b is a more likely candidate because of these sed features . perhaps
because the @xmath29-band coverage at this location is relatively shallow , only a @xmath29 upper limit can be determined for candidate b. the r - band source seen in figure [ fig : montage ] close to the apparent position of candidate b is not the visible - wavelength counterpart .
only upper limits could be established in the bands blueward of i for this object .
_ cudss 14.20_. the 8.0 @xmath0 m search selects a different best candidate than webb et al . .
the irac candidate is not detected in our deep r - band image or the webb et al .
i - band image but is detected with @xmath51 mag .
it therefore has red @xmath29-to-4.5 @xmath0 m colors that place it at @xmath47 .
there is no significant mips 24 @xmath0 m source within 10 of the scuba position .
acquiring a visible - wavelength spectrum of this extremely optically - faint object would be very challenging .
as is the case for 14.8 , we select a different position than preferred by chapman et al .
( 2005 ) , who obtained a spectroscopic redshift of @xmath48 for the @xmath29-selected candidate , where they detect 1.4 ghz emission . however , the radio identification , which is made at roughly 4@xmath7 significance ( s. chapman 2005 , private communication ) may be incorrect . _
cudss 14.22_. the 8.0 @xmath0 m search finds two candidates .
one of these ( a ) is coincident with the the @xmath29-selected candidate to the nw .
even though both are red in @xmath42 $ ] , the fact that they are both blue in @xmath38-[4.5]$ ] suggests that they lie in a relatively nearby redshift range @xmath36 and that we may have not detected the true counterpart of this object .
however , only the southernmost candidate is a strong 24 @xmath0 m source , so we regard it as the more plausible irac counterpart . for the sake of completeness
, we examined the relatively faint 24 @xmath0 m source apparent in figure [ fig : montage ] 103 nnw of the scuba position , just outside the search radius .
it is blue throughout the irac bands and is a much weaker 24 @xmath0 m source ( 232 @xmath0jy ) than candidate b. it is therefore likely to be a low - redshift object and not associated with the scuba source . _ cudss 14.6_. despite the excellent irac coverage at the location of cudss 14.6 ( a total of 218 exposures ) , no sources in the irac 8.0@xmath0 m mosaic satisfy our selection criteria as a valid counterpart to cudss 14.6 . within 10 of the reported scuba position , and specifically at the position identified by webb et al .
( 2003 ) , there is no significant 5.8 , 8.0 , or 24 @xmath0 m emission .
moreover , the _ k _ to 4.5 @xmath0 m colors of the _
k_-selected counterpart are blue , suggesting that object lies nearby .
there is a faint source in the 8.0 @xmath0 m mosaic within the 10 search radius , 85 to the ne of the scuba position .
however , its low flux density ( 5.5 @xmath0jy ) puts it below the 5@xmath7 sensitivity of the 8.0 @xmath0 m mosaic . like the _ k_-selected counterpart it also has a blue @xmath42 $ ] color , implying that it resides relatively nearby .
the object appears in three of the irac mosaics ( it is not apparent at 5.8 @xmath0 m ) and is therefore unquestionably real , but it is too faint to satisfy our selection criteria . a significant 8.0 @xmath0 m detection ( 18 @xmath0jy ) is located just outside the search area , 101 to the ne of the scuba position .
it is red in @xmath42 $ ] but exhibits a blue @xmath38-[4.5]$ ] color , suggesting it lies at an intermediate ( @xmath36 ) redshift and is therefore unlikely to be the true counterpart .
both of the 8.0 @xmath0 m sources mentioned above are as likely as not to be due to chance superposition . in combination with the color information this strongly suggests that no plausible counterpart to the scuba source is present in the irac mosaics .
_ cudss 14.11_. at 8.0 @xmath0 m there are two indistinguishable best candidates , one of which ( b ) is coincident with the @xmath29-selected candidate .
candidate a is closer to the scuba position , only 26 away .
neither shows any significant 24 @xmath0 m emission , and both exhibit the morphologies of elliptical galaxies in our deep r - band image .
furthermore , the blue infrared colors ( @xmath42 = -0.53,-0.60 $ ] , and @xmath38-[4.5]=-0.48,-0.69 $ ] ) and _ r_-band morphologies are consistent with both objects being elliptical galaxies at @xmath35 . there is no significant 24 @xmath0 m emission within 10 of the scuba position .
for all these reasons , we conclude that we have not detected the true counterpart to this scuba source .
a speculative possibility is that the 850 @xmath0 m emission detected by scuba is lensed by the foreground , low - redshift objects .
the closest 24 @xmath0 m source ( 176 @xmath0jy ) lies 157 ne of the scuba position and has blue colors throughout the irac bands . _
cudss 14.21_. there are no significant 8.0 @xmath0 m sources within 10 of the scuba position of this source .
we therefore attempt no identification .
there is significant 3.6 and 4.5 @xmath0 m emission at the location of the @xmath29-selected counterpart .
this object is red in @xmath42 $ ] but blue in @xmath38-[4.5]$ ] , suggesting that it may be at @xmath36 . at 5.8 , 8.0 , and 24.0 @xmath0 m
only upper limits are available .
we regard cudss 14.21 as unidentified at 8 @xmath0 m . assuming that these sources are not highly dust - obscured galaxies at such extremely high redshift ( @xmath52 ) that their obscured rest - visible emission falls in the observed mid - infrared , we suggest three possible causes for the lack of plausible 8 @xmath0 m counterparts . in the case of cudss 14.11
, it may be that the two foreground elliptical galaxies are screening a high - redshift scuba source from view at visible - mid - infrared wavelengths .
a second alternative is that the submillimeter flux densities may include contributions from multiple weak components that lie within the scuba beam but not precisely at the reported scuba position .
the scuba maps are known to be highly confused below 2 mjy ( knudsen et al .
2006 , and references therein ) .
weak submillimeter sources that combine to yield an apparent scuba detection might well be individually too faint to detect even in the deep irac mosaic .
third , one of the scuba sources might be spurious , which would not be unreasonable given that the sample contains so many roughly 3@xmath7 sources .
cudss 14.6 , 14.11 , and 14.21 are detected in the 850 @xmath0 m maps at 4.2 , 3.5 , and 3.0@xmath7 significance , respectively .
three sources lie in a portion of the irac 8.0 @xmath0 m survey region for which fewer than 10 exposures were acquired . for these sources (
14.5 , 14.9 , and 14.16 ) the depth of coverage was insufficient to reliably detect an infrared counterpart . excluding ` dubious ' associations , potential counterparts for 17 of 20 scuba galaxies having deep 8.0 @xmath0 m coverage in the 14@xmath1 field
have been identified ( table 2 ) . of these , the seven secure identifications have a mean offset from the scuba positions of 4.4 `` and the ten possible identifications have a mean offset of 7.7 '' .
if one uses the available 1.4 ghz positions in preference to the scuba positions only the mean offset for the set of secure identifications changes ( because no scuba sources with ` possible ' identifications have radio positions ) , to 24 .
the mean offsets in right ascension and declination for the 17 counterparts are 25 and 09 ( scuba - 8.0 @xmath0 m ) , respectively .
all these offsets are smaller than the scuba positional uncertainties . of course , a small number of the 8.0 @xmath0m - selected counterparts are chance associations and not physically associated with the corresponding scuba source . in the following we address this issue in two ways .
first , we derive an estimate of the number of such chance associations using a statistical analysis .
second , we compare the infrared - submm colors of our counterparts to those that have been reliably identified in a deep irac survey of the lockman hole region by egami et al .
( 2004 ) .
we first employ a variant of the statistical formalism used by lilly et al .
( 1999 ) and others to assess the significance of possible matches .
specifically , we estimate a statistic , usually denoted @xmath53 , that a candidate irac source lies within the search distance ( 10 ) of a scuba source to which it is unrelated .
the statistic is @xmath54 where @xmath55 and @xmath56 is the surface density of irac sources at least as bright as the candidate .
thus lower values of @xmath53 suggest a higher likelihood that the association of the irac and scuba sources is significant .
unlike lilly et al .
( 1999 ) we treat all sources within the 10search radius equally with respect to their offset from the scuba position . in doing
so we avoid unduly biasing our selection toward sources at small offsets from the uncertain scuba positions .
we use table 1 of webb et al .
( 2003 ) for the scuba sources coordinates , except for cudss 14.1 , 14.3 , 14.13 , 14.18 , and 14.19 .
the first four of these objects have vla positions ( eales et al .
2000 ) with lower uncertainties ( roughly 2 ) .
similarly , we adopt the radio position reported for 14.19 by webb et al .
( 2003 ) , although it is offset from the nominal scuba position by 85 .
we found no significant differences in our lists of most - likely matches when we repeated the analysis using the less - certain scuba positions . in order to interpret the @xmath53-values
, we performed a monte carlo analysis in which randomly - generated catalogs of irac sources were searched for counterparts in a manner identical to that employed for the real sources .
the random catalogs were constructed to present brightness distributions and source densities identical to those of the real catalog .
the @xmath53-distributions for the random and real samples are compared in figure [ fig : pdist ] .
we calculate the probability @xmath57 of finding a random or unassociated source as bright as the observed candidate as @xmath58 , where @xmath59 is the ratio of the number of random sources with smaller values of @xmath53 than the candidate to the total number of random sources generated in the monte carlo calculation .
although the @xmath57 statistics are not meaningful for individual sources , the sum of the @xmath57 statistics tabulated in table [ 8micids ] is an estimate of the number of spurious identifications .
we can not say which _ particular _ identifications are wrong , but the sums for the seven secure and the ten possible identifications are 0.61 and 2.61 , respectively , implying that the bulk of the errors apply to the latter subset , as one would expect . for the sample as a whole ,
the sum is 3.22 , suggesting that about three of our 17 identifications are incorrect .
this estimate ( three false identifications ) is a pessimistic one , because the probability formalism from which it is derived does not account for our use of color information to winnow lists of multiple candidates down to single most - likely counterparts .
three is in fact the maximum number of likely false sources , because we have taken into account the @xmath29 to mid - infrared colors as well as the presence of 24 @xmath0 m mips emission as identification criteria .
when the estimate for the number of spurious sources is subtracted , our identification rate is ( conservatively ) 14/20 , or 70% .
this is comparable to or better than rates described elsewhere .
borys et al .
( 2004 ) recovered 11 of 19 ( 57% ) scuba source in the hdf - n .
pope et al .
( 2005 ) reported a 72% recovery rate for scuba sources in the goods - n field , although this figure does nt account for possible erroneous associations .
like pope et al .
( 2005 ) we recovered essentially all the scuba sources with secure radio positions ( but see the discussions of 14.8 and 14.20 above ) .
more straightforward comparisons can be made with lilly et al .
( 1999 ) and webb et al .
( 2003 ) , because the @xmath57 statistics they provide permit one to estimate the number of incorrect associations in a manner identical to that applied to our own sample .
lilly et al .
( 1999 ) identified counterparts for eight of the 12 submillimeter sources they observed at visible wavelengths .
the sum of the resulting @xmath57 statistics was @xmath60 , suggesting a total identification rate of 7/12 or 58% .
likewise the webb et al .
( 2003 ) @xmath29-band counterparts for which the @xmath57 statistics sum to @xmath61 , implying a total identification rate of 17/23 or 74% ( but see below ) .
egami et al .
( 2004 ) presented mid - infrared through submillimeter flux densities measured by spitzer for scuba sources in a @xmath62 area of the lockman hole east . because these sources have radio counterparts from ivison et al .
( 2002 ) their positions are well - localized , and the mid - infrared identifications are secure . for this reason , the egami et al . ( 2004 ) submillimeter sample provides a useful basis with which to assess our own larger sample of infrared - selected counterparts .
figure [ fig : irac_colors ] shows the distributions of submillimeter galaxies in the irac @xmath38-[4.5]$ ] versus @xmath45-[8.0]$ ] color - color space .
the portion of this space occupied by the lockman hole sources and the secure identifications of the cudss 14@xmath1 field sources are approximately coextensive ; they define a locus in which other counterparts sources may be expected to fall .
this is indeed the case for the ` possible ' sources taken as a group the distribution of these objects occupies a very similar region of this color - color space overall as do the secure sources . in general , when the counterpart is bright , the @xmath29- and 8.0 @xmath0m - selected counterparts are identical .
however , the mid - infrared technique selects different counterparts than webb et al .
( 2003 ) in 10 cases ( cudss 14.4 , 14.7 , 14.8 , 14.10 , 14.12 , 14.14 , 14.15 , 14.19 , 14.20 , and 14.22 ) . the @xmath29- and 8.0 @xmath0m - selected techniques disagree for three additional cases ( cudss 14.6 , 14.11 , and 14.21 ) for which the mid - infrared technique identifies no plausible counterparts .
cudss 14.11 is a nearby elliptical galaxy and is therefore unlikely to be the source of the submillimeter emission .
cudss 14.6 and 14.21 have no plausible counterparts in the 8.0 @xmath0 m mosaics . the @xmath29- and
8.0 @xmath0m - selected counterparts therefore disagree in a total of 13 cases ( slightly more than half the total ) . with the exception of
the counterpart to cudss 14.7a ( which as noted above may indeed be a contributor to the submillimeter emission ) , the distribution of the @xmath29-selected sources is clearly blue in both colors compared to the other objects . in particular
it lies close to the region occupied by a passively evolving elliptical galaxy .
only six @xmath29-selected counterparts are plotted because the remainder are undetected at both 5.8 and 8.0 @xmath0 m , however the @xmath38-[4.5]$ ] colors of all these other sources tend to be similarly blue in @xmath38-[4.5]$ ] and not like the distributions of the secure identifications ( table [ 8micids ] ) .
we conclude that the 8.0 @xmath0m - selected counterparts are therefore more likely to be the correct counterparts in these discrepant cases .
figure [ fig : mips_scuba ] shows the same populations but in @xmath63-[850]$ ] versus @xmath46-[24]$ ] color - color space .
again , the distributions of the lockman hole counterparts and the secure sources from this work are similar .
cudss 14.13 and 14.18 are understandable exceptions on account of their relatively low redshifts . and as before , the ` possible ' counterparts occupy a region of this color - color space that is reassuringly similar to that of the secure sources .
that is not the case for the discrepant @xmath29-selected counterparts .
these objects lie at the outskirts of the secure counterparts distribution .
based on the lockman hole sources and the secure 8.0 @xmath0m - selected sources ( not including cudss 14.13 and 14.18 ) , one would expect the typical @xmath63-[850]$ ] color of a scuba galaxy to be between 3 and 4 magnitudes .
not one of the discrepant @xmath29-selected counterparts lies in this range , and indeed only one ( 14.15a ) is even detected at 24 @xmath0 m .
analogous trends are apparent in @xmath38-[8.0]$ ] versus @xmath46-[24]$ ] color - color space shown in figure [ fig : mips_irac ] .
the secure 8.0 @xmath0m - selected sources occupy the same region of this color - color space as the lockman hole sources , and that region is spanned by the possible 8.0 @xmath0 m identifications .
the low - redshift objects are not outliers in this distribution .
as before , the discrepant @xmath29-selected counterparts tend to be blue in @xmath38-[8.0]$ ] and with the exception of 14.7a do not overlap with the securely identified counterparts ; half of these discrepant sources are so blue they are not even detected at 8 or 24 @xmath0 m .
figure [ fig : sedplot ] shows the seds of our most - likely irac counterparts to the scuba sources in this field , showing the diversity of spectral shapes present within the sample .
the galaxies divide roughly into two categories .
the majority ( 12 objects ) show evidence in the irac bands for the 1.6 @xmath0 m bump arising from the h@xmath64 opacity minimum .
in other words , these objects appear to be powered primarily by stars and not by a compact object .
cudss 14.18 is among these sources , in agreement with chapman et al .
( 2005 ) .
four objects in the sample ( sources 14.3 , 14.7 , 14.13 , and 14.19 ) show the monotonically rising seds indicative of agn - dominated emission .
these sources fulfill all three of the irac - color agn selection criteria proposed by @xcite , @xcite , and @xcite .
source 14.4 has an unusual sed , very faint in the three shortest - wavelength irac bands compared to the optical and 8 and 24 @xmath0 m .
this source also has a monotonic rise through the irac bands , so we consider it a tentative agn candidate .
simple color criteria developed from shallow irac surveys like those cited above are not adequate to separate agn from starburst galaxies at high redshift however .
we therefore chose to perform a simple sed fitting procedure to determine whether a galaxy is agn - dominated , in exactly the same way as @xcite did for other sources in the extended groth strip .
specifically , we fit power laws ( @xmath65 ) to irac flux densities as an empirical way of characterizing their seds .
seven sources have acceptable power - law fits .
source 14.14 has a blue power - law ( @xmath66 ) , source 14.18 has @xmath67 , and five others 14.3 , 14.7 , 14.12 , 14.13 , and 14.19 ) have good fits to red power - laws .
source 14.4 does not fit a power - law well because of a very steep rise in the sed at 8.0 @xmath0 m .
agns thus selected on the basis of the irac - mips seds show good agreement with agns discovered on the basis of their x - ray emission .
four objects with secure mid - infrared power - law agn signatures are x - ray sources ( 14.3 , 14.7 , and 14.19 are nandra et al
. 2005 objects c111 , c113 , and c128 respectively ; 14.13 is source c72 in the list of nandra et al . and source 23 in waskett et al .
cudss 14.12 and 14.4 ( the tentative agn candidate ) , show no x - ray emission . our classification of 14.13 as agn - dominated agrees with higdon et al .
( 2004 ) and le floch et al .
( 2006 , in preparation ) , although chapman et al .
( 2005 ) call it a starburst . given the small numbers involved , our finding that between five and six members of the sample are agn - dominated is not inconsistent with the assertion by alexander et al . (
2004 ) that at most 20% " of the luminosity of submillimeter galaxies arises from agn .
they found that bright ( @xmath68 mjy ) scuba sources host agns at twice this rate ( 38% ) , although star formation tends to dominate the energy output . in summary , the number of scuba counterparts apparently dominated by an agn in the mid - infrared could be as few as five ( cudss 14.3 , 14.7 , 14.12 , 14.13 , and 14.19 ) or as many as six ( if 14.4 is included ) .
there could be others not detectable using our approach : @xcite found that not all ( x ray selected ) agn show clear mid - infrared signatures .
figure [ fig : fluxhist ] shows the 850 @xmath0 m flux density distribution of the 23 scuba galaxies .
our 17 counterparts span the full range occupied by the sample , including all of the brightest scuba sources and all but one of the faintest .
there is no evidence that our recovery rate is biased toward either the high- or low - brightness scuba sources , based on a kolmogorov - smirnov ( ks ) test comparing a sample consisting of the retrieved sources with another sample consisting of all 23 scuba galaxies .
furthermore , agn - dominated objects occupy the middle of the distribution , without any obvious skew toward either the high- or low - brightness extreme of the distribution .
the infrared color distributions of the irac counterparts to the scuba galaxies are shown in figure [ fig : histocolor ] , and the means and dispersions are given in table [ tab : colors ] .
colors at longer wavelengths present greater diversity ( larger dispersions ) both for the scuba counterparts and for the full sample of all 8@xmath0m - detected galaxies in the field , but the color distributions are consistently tighter for the counterparts
. in particular , the counterparts do not overlap the blue wings of the distributions of the field galaxies .
the scuba counterparts are slightly redder in the irac bands relative to the full sample of all irac galaxies in the cudss 14@xmath1 field .
although the difference in any one color is small , it amounts to nearly 1 magnitude in @xmath38-[8.0]$ ] .
this is unlikely to be a selection effect .
the seven counterparts ( 14.1 , 14.3 , 14.13 , 14.17 , 14.18 , 14.19 , and 14.20 ) identified without recourse to any color criteria show no significant color differences from the larger sample .
given the dispersions , the differences in mean colors between scuba galaxies and field galaxies are only marginally significant .
the ks test , however , is sensitive to the lack of blue galaxies in the counterpart sample .
the ks - derived probabilities of drawing samples of scuba counterparts that are as different as observed from the sample consisting of all 8 @xmath0m - detected galaxies are @xmath69 , @xmath70 , and @xmath71 for the @xmath38-[4.5]$ ] , @xmath72-[5.8]$ ] , and @xmath45-[8.0]$ ] color distributions , respectively , if the underlying populations are the same . in other words ,
the color distributions are significantly different .
figure [ fig : mipscolor ] shows the [ 8.0 ] - [ 24.0 ] and _ k _ - [ 3.6 ] color distributions for the irac - detected galaxies in this field .
both distributions exhibit much larger dispersions than the irac - only colors plotted in figure [ fig : histocolor ] . despite the fact that in some cases the irac counterpart was selected on the basis of the presence and/or strength of mips 24 @xmath0 m emission , there are no significant differences in the mean [ 8.0 ] - [ 24.0 ] colors or dispersions of the full sample and the counterparts .
the _ k _ - [ 3.6 ] color distributions are likewise indistinguishable ( given the wide range of colors exhibited by the sample ) .
when we compare the color distributions using ks tests we find no evidence that the samples are different with respect to either _ k _ - [ 3.6 ] or [ 8.0 ] - [ 24.0 ] color .
deep irac mosaics identify infrared counterparts for 17 of 20 scuba galaxies in the cudss 14@xmath1 field . the recovery rate of scuba galaxies at 8.0 @xmath0 m is comparable to or better than that found in other counterpart surveys , once the likelihood of mistaken identifications is estimated ( three or fewer for the current work ) . for more than half of the sample ,
the most - likely 8.0 @xmath0 m counterpart is different than the _
k_-selected counterpart .
multiband irac - mips seds allow one to identify and avoid unlikely candidates in the scuba beam ( e.g. , elliptical galaxies ) .
this suggests that the mid - infrared regime in general ( and the irac 8.0 @xmath0 m band in particular ) offers a more reliable means than ground - based optical or near - infrared surveys for identifying the true counterparts to the high - redshift , dusty starforming galaxies detected in submillimeter surveys .
on the basis of power - law fits to the infrared irac - mips seds we infer that five or possibly six of the counterparts are agn - dominated .
the 8.0 @xmath0 m counterparts have redder infrared colors than the general population of irac - detected galaxies in the field .
five counterparts have irac - mips seds that are fit well by red power laws and are likely to be agn - dominated .
one more is fit well by a slightly blue power - law and may also harbor an active nucleus .
four of these six objects are detected in x - ray observations , lending further support to the agn hypothesis .
the remaining counterparts are likely to be dominated by starburst emission .
this work is based on observations made with the _
spitzer space telescope _ , which is operated by the jet propulsion laboratory , california institute of technology under nasa contract 1407 .
support for this work was provided by nasa through contract number 1256790 issued by jpl / caltech .
we are grateful to the anonymous referee , whose comments and suggestions greatly improved the clarity and presentation of this work .
iraf is distributed by the national optical astronomy observatories , which are operated by the association of universities for research in astronomy , inc .
, under cooperative agreement with the national science foundation .
facilities : .
lcrrrrrccll @xmath7314.1 & & 40.7 & 44.8 & 32.5 & 41.1 & 191 & 14 17 40.22 & 52 29 06.5 & 1.07(0.28 ) & 0.11 + @xmath7314.2 & a & 38.5 & 45.0 & 41.5 & 27.0 & 150 & 14 17 51.33 & 52 30 24.9 & 6.55 & 0.18 + 14.2 & b & 2.0 & 3.3 & 5.5 & 8.6 & @xmath74 & 14 17 51.93 & 52 30 30.8 & 2.12 & 0.45 + @xmath7314.3 & & 35.0 & 41.0 & 48.6 & 48.4 & 408 & 14 18 00.58 & 52 28 21.3 & 0.87(2.32 ) & 0.078 + @xmath7314.4 & a & 2.7 & 3.6 & 5.2 & 27.2 & 239 & 14 17 43.43 & 52 28 05.8 & 8.71 & 0.18 + 14.4 & b & 52.8 & 49.2 & 40.7 & 40.9 & & 14 17 42.41 & 52 28 11.6 & 9.07 & 0.11 + 14.4 & c & 7.4 & 8.7 & 9.5 & 6.2 & @xmath74 & 14 17 43.80 & 52 28 16.8 & 4.71 & 0.52 + 14.6 & & 1.8 & 1.8 & @xmath75 & 5.5 & @xmath74 & 14 17 57.02 & 52 29 14.7 & 8.50 & 0.51 + 14.7 & a & 8.4 & 10.6 & 11.7 & 12.9 & @xmath74 & 14 18 00.95 & 52 29 50.0 & 1.68 & 0.35 + @xmath7314.7 & b & 5.3 & 10.6 & 20.2 & 51.3 & 150 & 14 18 01.15 & 52 29 42.0 & 7.02 & 0.071 + @xmath7314.8 & a & 21.4 & 24.5 & 24.7 & 15.3 & 275 & 14 18 02.01 & 52 30 16.3 & 6.43 & 0.31 + 14.8 & b & 5.4 & 5.6 & 4.9 & 2.9 & @xmath74 & 14 18 02.70 & 52 30 10.4 & 5.25 & 0.76 + @xmath7314.10 & a & 50.1 & 48.4 & 34.8 & 38.3 & 307 & 14 18 03.00 & 52 29 33.2 & 9.81 & 0.12 + 14.10 & b & 4.3 & 3.2 & @xmath75 & @xmath76 & @xmath74 & 14 18 03.98 & 52 29 40.6 & 2.20 & 0.86 + 14.11 & a & 153.6 & 99.5 & 66.7 & 36.0 & @xmath74 & 14 17 46.89 & 52 32 36.1 & 2.69 & 0.13 + 14.11 & b & 225.3 & 140.8 & 92.7 & 50.3 & @xmath74 & 14 17 47.32 & 52 32 42.1 & 4.60 & 0.075 + @xmath7314.12 & a & 3.7 & 5.7 & @xmath75 & 8.0 & 80 & 14 18 04.70 & 52 28 56.0 & 5.48 & 0.47 + 14.12 & b & 3.1 & 4.2 & @xmath75 & 18.3 & @xmath74 & 14 18 05.69 & 52 29 03.3 & 8.60 & 0.33 + @xmath7314.13 & & 580.1 & 981.7 & 1448.5 & 2225.5 & 5646 & 14 17 41.88 & 52 28 23.5 & 0.67(6.46 ) & 0.0013 + 14.14 & a & 9.9 & 8.3 & 3.4 & 6.2 & @xmath74 & 14 18 09.06 & 52 31 00.5 & 4.78 & 0.52 + @xmath7314.14 & b & 9.4 & 10.7 & 7.8 & 6.9 & 120 & 14 18 08.21 & 52 31 08.1 & 6.16 & 0.50 + 14.15 & a & 30.8 & 26.7 & 19.7 & 17.1 & 103 & 14 17 29.87 & 52 28 21.5 & 5.83 & 0.28 + @xmath7314.15 & b & 7.4 & 10.8 & 15.3 & 12.1 & 93 & 14 17 28.44 & 52 28 19.8 & 7.90 & 0.37 + @xmath7314.17 & & 47.1 & 47.0 & 34.6 & 77.0 & 266 & 14 17 24.44 & 52 30 46.1 & 9.48 & 0.049 + @xmath7314.18 & & 159.7 & 138.1 & 143.0 & 164.7 & 1078 & 14 17 42.11 & 52 30 25.7 & 0.61(1.51 ) & 0.029 + @xmath7314.19 & & 15.3 & 20.6 & 21.1 & 25.2 & 84.0 & 14 18 11.26 & 52 30 12.3 & 2.33(8.59 ) & 0.20 + @xmath7314.20 & & 18.5 & 21.2 & 18.6 & 13.9 & @xmath74 & 14 17 50.51 & 52 30 55.2 & 8.87 & 0.34 + 14.22 & a & 29.2 & 28.1 & 19.6 & 19.1 & @xmath74 & 14 17 55.08 & 52 32 08.4 & 7.91 & 0.25 + @xmath7314.22 & b & 72.1 & 66.7 & 49.1 & 50.3 & 500 & 14 17 56.80 & 52 31 57.8 & 9.94 & 0.075 + @xmath7314.23 & a & 31.7 & 43.4 & 47.1 & 34.0 & 529 & 14 17 46.21 & 52 33 22.2 & 1.97 & 0.14 + 14.23 & b & 21.4 & 17.5 & 20.1 & 20.1 & @xmath74 & 14 17 46.68 & 52 33 29.6 & 6.60 & 0.24 lcccccccccc 14.1 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & + 14.2a & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & [email protected] & + 14.3 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & + 14.4a & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & @xmath78 & @xmath79 & + 14.7b & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & & [email protected] & + 14.8a & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & + 14.10a & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & + 14.12a & @xmath80 & @xmath81 & [email protected] & [email protected] & [email protected] & @xmath78 & @xmath82 & + 14.13 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 1.150 + 14.14b & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & + 14.15b & @xmath80 & @xmath81 & @xmath83 & @xmath84 & [email protected] & & @xmath85 & + 14.17 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & + 14.18 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & 0.661 + 14.19 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & + 14.20 & @xmath80 & @xmath81 & @xmath83 & 27.7@xmath86 & [email protected] & [email protected] & [email protected] & + 14.22b & @xmath80 & @xmath81 & @xmath83 & [email protected] & [email protected] & & [email protected] & + 14.23a & @xmath80 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & lrrrrrrr 8 @xmath0 m galaxies & 726 & @xmath87 & @xmath88 & @xmath89 & @xmath90 & @xmath91 + 8 @xmath0 m counterparts & 17 & @xmath92 & @xmath93 & @xmath94 & @xmath95 & @xmath96 + counterparts subsample & 7 & @xmath97 & @xmath93 & @xmath94 & @xmath98 & @xmath99 | we identify 17 possible 8.0 @xmath0m - selected counterparts to the submillimeter galaxies in the cudss 14@xmath1 field , derived from deep imaging carried out with the irac and mips instruments aboard the _ spitzer space telescope_. ten of the 17 counterparts are not the same as those previously identified at shorter wavelengths .
we argue that 8.0 @xmath0 m selection offers a better means for identifying counterparts to submillimeter galaxies than near - infrared or optical selection . based on the panchromatic seds ,
most counterparts appear to be powered by ongoing star formation .
power - law fits to the seds suggest that five objects in the 8.0 @xmath0m - selected counterpart sample harbor dominant agns ; a sixth object is identified as a possible agn .
the 3.6 - 8.0 @xmath0 m colors of the infrared - selected counterparts are significantly redder than the general irac galaxy population in the cudss 14@xmath1 field . |
the @xmath0 mass region is a particularly complex one , displaying transitional behavior such as prolate - oblate deformed shapes , @xmath5-unstability , triaxial deformation and/or coexistence of different configurations which present a daunting challenge to nuclear structure models . despite this complexity
, the @xmath0 mass region has been a rich source of empirical evidence for the existence of dynamical symmetries in nuclei both for even - even , odd - proton , odd - neutron and odd - odd nuclei , as well as supersymmetric pairs @xcite and quartets of nuclei @xcite . in this contribution , we present evidence for the existence of a new supersymmetric quartet in the @xmath0 mass region , consisting of the @xmath2os and @xmath3ir nuclei , and study correlations between different one- and two - nucleon transfer reactions .
dynamical supersymmetries ( susy ) were introduced in nuclear physics in the context of the interacting boson model ( ibm ) and its extensions @xcite .
the ibm describes collective excitations in even - even nuclei in terms of a system of interacting monopole ( @xmath6 ) and quadrupole ( @xmath7 ) bosons @xcite .
the bosons are associated with the number of correlated proton and neutron pairs , and hence the number of bosons @xmath8 is half the number of valence nucleons . for odd - mass nuclei
the ibm was extended to include single - particle degrees of freedom @xcite .
the ensuing interacting boson - fermion model ( ibfm ) has as its building blocks @xmath8 bosons with @xmath9 and @xmath10 fermion ( @xmath11 ) with @xmath12 @xcite .
the ibm and ibfm can be unified into a supersymmetry ( susy ) @xmath13 where @xmath14 is the dimension of the fermion space @xcite . in this framework ,
even - even and odd - even nuclei form the members of a supermultiplet which is characterized by @xmath15 , i.e. , the total number of bosons and fermions .
supersymmetry distinguishes itself from other symmetries in that it includes , in addition to transformations among fermions and among bosons , also transformations that change a boson into a fermion and _ vice versa_. the concept of nuclear susy was extended in 1985 to include the neutron - proton degree of freedom @xcite . in this case , a supermultiplet consists of an even - even , an odd - proton , an odd - neutron and an odd - odd nucleus .
the the best experimental evidence of a supersymmetric quartet is provided by the @xmath16pt and @xmath17au nuclei as an example of the @xmath18 supersymmetry @xcite , in which the odd neutron is allowed to occupy the @xmath19 , @xmath20 and @xmath21 orbits of the 82 - 126 shell , and the odd proton the @xmath22 orbit of the 50 - 82 shell .
this supermultiplet is characterized by @xmath23 and @xmath24 .
the excitation spectra of the nuclei belonging to the supersymmetric quartet are described simultaneously by the energy formula @xmath25 \nonumber\\ & & + b \left [ \sigma_1(\sigma_1 + 4)+\sigma_2(\sigma_2 + 2)+\sigma_3 ^ 2 \right ] + b ' \left [ \sigma_1(\sigma_1 + 4)+\sigma_2(\sigma_2 + 2)+\sigma_3 ^ 2 \right ] \nonumber\\ & & + c \left [ \tau_1(\tau_1 + 3)+\tau_2(\tau_2 + 1 ) \right ] + d \ ,
l(l+1 ) + e \ , j(j+1 ) ~. \label{npsusy}\end{aligned}\ ] ] the coefficients @xmath26 , @xmath27 , @xmath28 , @xmath29 , @xmath30 , and @xmath31 are determined in a simultaneous fit of the excitation energies of the four nuclei that make up the quartet .
recently , the structure of the odd - odd nucleus @xmath1ir was investigated by a series of transfer and neutron capture reactions @xcite . in particular , the new data from the polarized @xmath32 transfer reaction provided crucial new information about and insight into the structure of the spectrum of @xmath1ir which led to significant changes in the assignment of levels as compared to previous work @xcite .
the odd - odd nucleus @xmath1ir differs from @xmath33au by two protons , the number of neutrons being the same .
the latter is crucial , since the dominant interaction between the odd neutron and the core nucleus is of quadrupole type , which arises from a more general interaction in the ibfm for very special values of the occupation probabilities of the @xmath19 , @xmath20 and @xmath21 orbits , _
i.e. _ to the location of the fermi surface for the neutron orbits @xcite .
this situation is satisfied to a good approximation by the @xmath34pt and @xmath33au nuclei , and thus also for @xmath4os and @xmath1ir .
for this reason , it is reasonable to expect the odd - odd nucleus @xmath1ir to provide another example of the @xmath35 supersymmetry .
[ ir194 ] shows the negative parity levels of @xmath1ir in comparison with the theoretical spectrum in which it is assumed that these levels originate from the @xmath36 , @xmath37 , @xmath38 configuration .
the theoretical energy spectrum is calculated using the energy formula of eq .
( [ npsusy ] ) with @xmath39 , @xmath40 , @xmath41 , @xmath42 , @xmath43 , and @xmath44 ( all in kev ) .
given the complex nature of the spectrum of heavy odd - odd nuclei , the agreement is remarkable .
there is an almost one - to - one correlation between the experimental and theoretical level schemes @xcite .
ir.,width=302 ] the successful description of the odd - odd nucleus @xmath1ir opens the possibility of identifying a second quartet of nuclei in the @xmath0 mass region with @xmath35 supersymmetry .
the new quartet consists of the nuclei @xmath2os and @xmath3ir and is characterized by @xmath23 and @xmath45 .
whereas the @xmath46os and @xmath3ir nuclei are well - known experimentally , the available data for @xmath4os is rather scarce . in fig .
[ os193 ] we show the predicted spectrum for @xmath4os obtained from eq .
( [ npsusy ] ) using the same parameter set as for @xmath1ir .
we note , that the ground state of @xmath4os has spin and parity @xmath47 , which implies that the second band with labels @xmath48 $ ] , @xmath49 is the ground state band , rather than @xmath50 $ ] , @xmath51 .
the relative ordering of these bands is determined by the coefficients @xmath26 and @xmath52 . at present
, we are carrying out a simultaneous fit of the excitation energies of all four nuclei that make up the quartet to see whether it is possible to reproduce the relative ordering in @xmath4os without affecting the successful description of @xmath1ir @xcite os for the @xmath18 supersymmetry.,width=302 ]
the nuclei belonging to a supersymmetric quartet are described by a single hamiltonian , and hence the wave functions , transition and transfer rates are strongly correlated . as an example of these correlations , we consider here
the case of one - neutron transfer reactions between the pt and os nuclei . in a study of the @xmath1pt @xmath53 @xmath34pt stripping reaction
it was found @xcite that one - neutron @xmath54 , @xmath55 transfer reactions can be described in the @xmath35 supersymmetry scheme by the operator @xmath56 ~. \end{aligned}\ ] ] it is convenient to take ratios of intensities , since they do not depend on the value of the coefficient @xmath57 and hence provide a stringent test of the wave functions . for the stripping reaction @xmath1pt @xmath53 @xmath34pt ( ee @xmath53 on ) the ratio of intensities for the excitation of the @xmath58 , @xmath59 doublet with @xmath60 , @xmath55 belonging to the first excited band with @xmath61 $ ] ,
@xmath62 relative to that of the ground state band @xmath63 $ ] , @xmath64 is given by @xcite @xmath65 which gives @xmath66 for @xmath1pt @xmath53 @xmath34pt ( @xmath67 ) , in comparison with the experimental value of 19.0 for @xmath68 , and @xmath69 for @xmath46os @xmath53 @xmath4os ( @xmath70 ) . the equivalent ratio for the inverse pick - up reaction is given by @xmath71 which gives @xmath72 for @xmath34pt @xmath53 @xmath1pt ( @xmath73 and @xmath74 ) and @xmath75 for @xmath4os @xmath53 @xmath46os ( @xmath76 and @xmath74 ) .
this means that the mixed symmetry @xmath59 state is predicted to be excited more strongly than the first excited @xmath59 state .
this correlation between pick - up and stripping reactions can be derived in a general way only using the symmetry relations that exist between the wave functions of the even - even and odd - neutron nuclei of the supersymmetric quartet .
it is important to point out , that eqs .
( [ ratio ] and ( [ corr ] ) are parameter - independent predictions which are a direct consequence of nuclear susy and which can be tested experimentally .
in conclusion , we have presented evidence for the existence of a second quartet of nuclei in the @xmath77 region with @xmath78 supersymmetry , consisting of the @xmath2os and @xmath3ir nuclei .
the analysis is based on new experimental information on @xmath1ir . in particular , the @xmath32 reaction is important to establish the spin and parity assignments of the energy levels , and to provide insight into the structure of the spectrum of @xmath1ir . given the complexity of the @xmath0 mass region , the simple yet detailed description of @xmath1ir in a supersymmetry scheme is truly remarkable .
nuclear supersymmetry establishes precise links among the spectroscopic properties of different nuclei .
this relation has been used to predict the energies of @xmath4os .
since the wave functions of the members of a supermultiplet are connected by symmetry , there exists a high degree of correlation between different one- and two - nucleon transfer reactions not only between nuclei belonging to the same quartet , but also for nuclei from different multiplets @xcite . as an example
, we studied the correlations between one - neutron transfer reactions for the pt and os isotopes , and predicted that the @xmath59 mixed symmetry states in the even - even nucleus are populated much stronger than the first excited @xmath59 state . in order to establish the existence of a second supersymmetric quartet of nuclei in the @xmath0 mass region , it is crucial that the nucleus @xmath4os be studied in more detail experimentally .
the predictions for correlations between one - neutron transfer reactions in pt and os can be tested experimentally by combining for example @xmath79 stripping and @xmath80 pick - up reactions . | we present evidence for the existence of a new supersymmetric quartet of nuclei in the @xmath0 mass region .
the analysis is based on new experimental information on the odd - odd nucleus @xmath1ir from transfer and neutron capture reactions .
the new data allow the identification of a new supersymmetric quartet , consisting of the @xmath2os and @xmath3ir nuclei .
we make explicit predictions for @xmath4os , and suggest that its spectroscopic properties be measured in dedicated experiments .
finally , we study correlations between different transfer reactions .
address = icn - unam , ap 70 - 543 , 04510 mxico df , mxico address = center for theoretical physics , sloane physics laboratory , yale university , p.o . box 208210 , new haven , connecticut 06520 - 8120 , u.s.a .
address = icn - unam , ap 70 - 543 , 04510 mxico df , mxico address = sektion physik , ludwig - maximilians - universit " at m " unchen , d-85748 garching , germany address = fakultt fr physik , ludwig - maximilians - universit " at m " unchen , d-85748 garching , germany address = institut f " ur kernphysik , universit " at zu k " oln , d-50937 k " oln , germany address = sektion physik , ludwig - maximilians - universit " at m " unchen , d-85748 garching , germany |
the capacity of the interference channel remains one of the most challenging open problems in the domain of network information theory .
the capacity region is not known in general , except for a specific range of channel parameters .
for the two - user scalar gaussian interference channel , where the interference alignment is not required , the approximate capacity region to within one bit is known @xcite . for the channels where interference alignment is required such as the @xmath0-user gaussian interference channel @xcite and the gaussian x - channel @xcite , a tight characterization of the capacity region
is not known , even for symmetric channel cases .
a tractable approach to the capacity of interference channels is to consider partial connectivity of interference links and analyze the impact of topology on the capacity .
topological interference management @xcite approach gives important insights on the degrees - of - freedom ( dof ) of partially connected interference channels and their connection to index coding problems @xcite .
it is shown that the symmetric dof of a partially connected interference channel can be found by solving the corresponding index coding problem . in this paper
, we consider a class of three - user partially connected interference channels and characterize approximate capacity regions at finite snr .
we focus on the impact of interference topology , interference alignment , and interplay between interference and noise .
we choose a few representative topologies where we can achieve clear interference alignment gain .
for these topologies , z - channel type outer bounds are tight to within a constant gap from the corresponding inner bound . for each topology
, we present an achievable scheme based on rate - splitting , lattice alignment , and successive decoding . lattice coding based on nested lattices
is shown to achieve the capacity of the single user gaussian channel in @xcite .
the idea of lattice - based interference alignment by decoding the sum of lattice codewords appeared in the conference version of @xcite .
this lattice alignment technique is used to derive capacity bounds for three - user interference channel in @xcite .
the idea of decoding the sum of lattice codewords is also used in @xcite to derive the approximate capacity of the two - way relay channel .
an extended approach , compute - and - forward @xcite enables to first decode some linear combinations of lattice codewords and then solve the lattice equation to recover the desired messages .
this approach is also used in @xcite to characterize approximate sum - rate capacity of the fully connected @xmath0-user interference channel .
the idea of sending multiple copies of the same sub - message at different signal levels , so - called zigzag decoding , appeared in @xcite where receivers collect side information and use them for interference cancellation .
the @xmath0-user cyclic gaussian interference channel is considered in @xcite where an approximate capacity for the weak interference regime ( @xmath1 for all @xmath2 ) and the exact capacity for the strong interference regime ( @xmath3 for all @xmath2 ) are derived .
our type 4 and 5 channels are @xmath4 cases in _ mixed _ interference regimes , which were not considered in @xcite .
we consider five channel types defined in table [ tab : txrxsignals ] and described in fig .
[ fig : channeltype ] ( a)(e ) .
each channel type is a partially connected three - user gaussian interference channel .
each transmitter is subject to power constraint @xmath5\leq p_k = p$ ] .
let us denote the noise variance by @xmath6 $ ] . without loss of generality , we assume that @xmath7 .
the side information graph representation of an interference channel satisfies the following . *
a node represents a transmitter - receiver pair , or equivalently , the message .
* there is a directed edge from node @xmath8 to node @xmath9 if transmitter @xmath8 does not interfere at receiver @xmath9 .
the side information graphs for five channel types are described in fig .
[ fig : channeltype ] ( f)(j ) .
we state the main results in the following two theorems , of which the proofs will be given in the main body of the paper .
for the five channel types , if @xmath10 is achievable , it must satisfy @xmath11 for every subset @xmath12 of the nodes @xmath13 that does not include a directed cycle in the side information graph over the subset . + for any rate triple @xmath10 on the boundary of the outer bound region , the point @xmath14 is achievable .
.five channel types [ cols="^,<",options="header " , ] in this section , we present an outer bound on the capacity region of type 5 channel defined by @xmath15 = \left [ { \begin{array}{*{20}c } 1 & 1 & 0\\ 0 & 1 & 1\\ 1 & 0 & 1\\ \end{array } } \right ] \left [ { \begin{array}{*{20}c } x_1\\ x_2\\ x_3\\ \end{array } } \right ] + \left [ { \begin{array}{*{20}c } z_1\\ z_2\\ z_3\\ \end{array } } \right].\ ] ] this is a cyclic gaussian interference channel @xcite . we first show that channel type 5 is in the mixed interference regime . by normalizing the noise variances , we get the equivalent channel given by @xmath16 = \left [ { \begin{array}{*{20}c } \frac{1}{\sqrt{n_1 } } & \frac{1}{\sqrt{n_1 } } & 0\\ 0 & \frac{1}{\sqrt{n_2 } } & \frac{1}{\sqrt{n_2}}\\ \frac{1}{\sqrt{n_3 } } & 0 & \frac{1}{\sqrt{n_3}}\\ \end{array } } \right ] \left [ { \begin{array}{*{20}c } x_1\\ x_2\\ x_3\\ \end{array } } \right ] + \left [ { \begin{array}{*{20}c } z_1'\\ z_2'\\ z_3'\\ \end{array } } \right].\ ] ] we can see that @xmath17 we state the outer bound in the following theorem .
the capacity region of type 5 channel is contained in the following outer bound region : @xmath18 @xmath19 where we used the fact that @xmath20 .
@xmath21 where we used the fact that @xmath22 .
@xmath23 for ease of gap calculation , we also derive relaxed outer bounds . first , we can see that for @xmath24 , @xmath25 five outer bound theorems in this section , together with this inequality , give the sum - rate bound expression in theorem 1 .
next , we can assume that @xmath26 for @xmath27 .
otherwise , showing one - bit gap capacity is trivial as the capacity region is included in the unit hypercube , i.e. , @xmath28 . for @xmath26 , @xmath29 the resulting relaxed outer bounds @xmath30 are summarized in table [ tab : outerbounds ] .
given @xmath31 ^ 2 $ ] , the rate region @xmath32 is defined by @xmath33 where @xmath34 . and , @xmath35 is achievable where @xmath36 is convex hull operator .
lattice @xmath37 is a discrete subgroup of @xmath38 , @xmath39 where @xmath40 is a real generator matrix .
quantization with respect to @xmath37 is @xmath41 .
modulo operation with respect to @xmath37 is @xmath42{\textrm { mod } \lambda}=\mathbf{x}-q_{\lambda}(\mathbf{x})$ ] . for convenience ,
we use both notations @xmath43 and @xmath44{\textrm { mod } \lambda}$ ] interchangeably . fundamental voronoi region of @xmath37 is @xmath45 .
volume of the voronoi region of @xmath37 is @xmath46 .
normalized second moment of @xmath37 is @xmath47 where @xmath48
. lattices @xmath49 , @xmath50 and @xmath37 are said to be nested if @xmath51 . for nested lattices
@xmath52 , @xmath53 .
we briefly review the lattice decoding procedure in @xcite .
we use nested lattices @xmath54 with @xmath55 , @xmath56 , and @xmath57 .
the transmitter sends @xmath58{\textrm { mod } \lambda}$ ] over the point - to - point gaussian channel @xmath59 where the codeword @xmath60 , the dither signal @xmath61 , the transmit power @xmath62 and the noise @xmath63 .
the code rate is given by @xmath64 .
after linear scaling , dither removal , and mod-@xmath37 operation , we get @xmath65{\textrm { mod } \lambda}= \left[\mathbf{t}+\mathbf{z}_e\right]{\textrm { mod } \lambda}\end{aligned}\ ] ] where the effective noise is @xmath66 and its variance @xmath67=(\beta-1)^2 s+\beta^2 n$ ] . with the mmse scaling factor @xmath68 plugged in ,
we get @xmath69 .
the capacity of the mod-@xmath37 channel @xcite between @xmath70 and @xmath71 is @xmath72 where @xmath73 and @xmath74 are mutual information and differential entropy , respectively . for reliable decoding of @xmath70
, we have the code rate constraint @xmath75 . with the choice of lattice parameters , @xmath76 , @xmath77 and @xmath78 , @xmath79
thus , the constraint @xmath75 can be satisfied . by
_ lattice decoding _
@xcite , we can recover @xmath70 , i.e. , @xmath80 with probability @xmath81 where @xmath82\ ] ] is the probability of decoding error . if we choose @xmath37 to be poltyrev - good @xcite , then @xmath83 as @xmath84 .
we present an achievable scheme for the proof of _ theorem 8_. the achievable scheme is based on rate - splitting , lattice coding , and interference alignment .
message @xmath85 is split into two parts : @xmath86 and @xmath87 , so @xmath88 .
transmitter 1 sends @xmath89 where @xmath90 and @xmath91 are coded signals of @xmath92 and @xmath93 , respectively .
transmitters 2 and 3 send @xmath94 and @xmath95 , coded signals of @xmath96 and @xmath97 .
in particular , @xmath90 and @xmath95 are lattice - coded signals .
we use the lattice construction of @xcite with the lattice partition chain @xmath98 , so @xmath99 are nested lattices .
@xmath100 is the coding lattice for both @xmath90 and @xmath95 .
@xmath49 and @xmath101 are shaping lattices for @xmath90 and @xmath95 , respectively .
the lattice signals are formed by @xmath102{\textrm { mod } \lambda}_1\\ & & \ \mathbf{x}_3=[\mathbf{t}_3+\mathbf{d}_3]{\textrm { mod } \lambda}_3\end{aligned}\ ] ] where @xmath103 and @xmath104 are lattice codewords . the dither signals @xmath105 and @xmath106 are uniformly distributed over @xmath107 and @xmath108 , respectively . to satisfy power constraints , we choose @xmath109=n\sigma^2(\lambda_1)=(1-\alpha_1 ) np$ ] , @xmath110={\alpha}_1 np$ ] , @xmath111={\alpha}_2 np$ ] , @xmath112=n\sigma^2(\lambda_3)=np$ ] . with the choice of transmit signals ,
the received signals are given by @xmath113+\mathbf{x}_2+\mathbf{z}_2'\\ & & \mathbf{y}_3=\mathbf{x}_3+\mathbf{z}_3'.\end{aligned}\ ] ] where @xmath114 $ ] is the sum of interference , and @xmath115 and @xmath116 are the effective gaussian noise .
the signal scale diagram at each receiver is shown in fig .
[ fig : signalscale ] ( a ) . at the receivers , successive decoding
is performed in the following order : @xmath117 at receiver 1 , @xmath118 at receiver 2 , and receiver 3 only decodes @xmath95 .
note that the aligned lattice codewords @xmath119 , and @xmath120{\textrm { mod } \lambda}_1 \in\lambda_c\cap\mathcal{v}(\lambda_1)$ ] .
we state the relationship between @xmath121 and @xmath122 in the following lemmas .
the following holds .
@xmath123{\textrm { mod } \lambda}_1=\mathbf{t}_f\ ] ] where @xmath124 .
@xmath125{\textrm { mod } \lambda}_1\\ & & = [ m_{\lambda_1}(\mathbf{t}_{11}+\mathbf{d}_{11})+m_{\lambda_3}(\mathbf{t}_3+\mathbf{d}_3)-\mathbf{d}_f]{\textrm { mod } \lambda}_1\\ & & = [ m_{\lambda_1}(\mathbf{t}_{11}+\mathbf{d}_{11})+m_{\lambda_1}(\mathbf{t}_3+\mathbf{d}_3)-\mathbf{d}_f]{\textrm { mod } \lambda}_1\\ & & = [ \mathbf{t}_{11}+\mathbf{d}_{11}+\mathbf{t}_3+\mathbf{d}_3-\mathbf{d}_f]{\textrm { mod } \lambda}_1\\ & & = [ \mathbf{t}_{11}+\mathbf{t}_3]{\textrm { mod } \lambda}_1\\ & & = \mathbf{t}_f\end{aligned}\ ] ] the second and third equalities are due to distributive law and the identity in the following lemma . for any nested lattices
@xmath126 and + any @xmath127 , it holds that @xmath128{\textrm { mod } \lambda}_1=[\mathbf{x}]{\textrm { mod } \lambda}_1.\ ] ] @xmath129{\textrm { mod } \lambda}_1\\ & & = [ \mathbf{x}-\lambda_3]{\textrm { mod } \lambda}_1\\ & & = [ m_{\lambda_1}(\mathbf{x})-m_{\lambda_1}(\lambda_3)]{\textrm { mod } \lambda}_1\\ & & = [ m_{\lambda_1}(\mathbf{x})-\lambda_3 + q_{\lambda_1}(\lambda_3)]{\textrm { mod } \lambda}_1\\ & & = [ m_{\lambda_1}(\mathbf{x})]{\textrm { mod } \lambda}_1\\ & & = [ \mathbf{x}]{\textrm { mod } \lambda}_1\end{aligned}\ ] ] where @xmath130 , thus @xmath131 . the following holds .
@xmath132{\textrm { mod } \lambda}_1=[\mathbf{x}_f]{\textrm { mod } \lambda}_1.\ ] ] @xmath133{\textrm { mod } \lambda}_1\\ & & = [ m_{\lambda_1}(\mathbf{t}_{11}+\mathbf{t}_3)+\mathbf{d}_f]{\textrm { mod } \lambda}_1\\ & & = [ \mathbf{t}_{11}+\mathbf{t}_3+\mathbf{d}_f]{\textrm { mod } \lambda}_1\\ & & = [ m_{\lambda_1}(\mathbf{t}_{11}+\mathbf{d}_{11})+m_{\lambda_1}(\mathbf{t}_3+\mathbf{d}_3)]{\textrm { mod } \lambda}_1\\ & & = [ m_{\lambda_1}(\mathbf{t}_{11}+\mathbf{d}_{11})+m_{\lambda_3}(\mathbf{t}_3+\mathbf{d}_3)]{\textrm { mod } \lambda}_1\\ & & = [ \mathbf{x}_{11}+\mathbf{x}_3]{\textrm { mod } \lambda}_1\\ & & = [ \mathbf{x}_f]{\textrm { mod } \lambda}_1\end{aligned}\ ] ] receiver 2 does not need to recover the codewords @xmath134 and @xmath135 but the real sum @xmath121 to remove the interference from @xmath136 .
since @xmath137 , we first recover the modulo part and then the quantized part to cancel out @xmath121 .
this idea appeared in @xcite as an achievable scheme for the many - to - one interference channel .
the mod-@xmath49 channel between @xmath122 and @xmath138 is given by @xmath139{\textrm { mod } \lambda}_1\\ & & \ \ \ \ = [ \mathbf{x}_f-\mathbf{d}_f + \mathbf{z}_{e2}]{\textrm { mod } \lambda}_1\\ & & \ \ \ \ = [ \mathbf{t}_f + \mathbf{z}_{e2}]{\textrm { mod } \lambda}_1\end{aligned}\ ] ] where the effective noise @xmath140 .
note that @xmath141=(\bar{\alpha}_0 + 1)np$ ] , and the effective noise variance @xmath142=(\beta_2 - 1)^2(\bar{\alpha}_0 + 1)p+\beta_2 ^ 2 n_{e2}$ ] where @xmath143 . with the mmse scaling factor @xmath144 plugged in ,
we get @xmath145 .
the capacity of the mod-@xmath49 channel between @xmath122 and @xmath138 is @xmath146 for reliable decoding of @xmath122 at receiver 2 , we have the code rate constraint @xmath147 .
this also implies that @xmath148 . by lattice decoding
, we can recover the modulo sum of interference codewords @xmath122 from @xmath138 .
then , we can recover the real sum @xmath121 in the following way .
* recover @xmath149 by calculating @xmath150{\textrm { mod } \lambda}_1 $ ] ( lemma 3 ) .
* subtract it from the received signal , @xmath151 where @xmath152 .
* quantize it to recover @xmath153 , @xmath154 with probability @xmath81 where @xmath155\ ] ] is the probability of decoding error .
if we choose @xmath49 to be simultaneously rogers - good and poltyrev - good @xcite with @xmath156 , then @xmath83 as @xmath84 .
* recover @xmath121 by adding two vectors , @xmath157 we now proceed to decoding @xmath94 from @xmath158 . since @xmath94 is a codeword from an i.i.d .
random code for point - to - point channel , we can achieve rate up to @xmath159 at receiver 1 , we first decode @xmath90 while treating other signals @xmath160 as noise .
the effective noise in the mod-@xmath49 channel is @xmath161 with variance @xmath162=(\beta_1 - 1)^2 \bar{\alpha}_0 p+\beta_1 ^ 2 n_{e1}$ ] where @xmath163 .
for reliable decoding , the rate @xmath164 must satisfy @xmath165 where the mmse scaling parameter @xmath166 .
similarly , we have the other rate constraints at receiver 1 : @xmath167 at receiver 3 , the signal @xmath95 is decoded with the effective noise @xmath168 . for reliable decoding , @xmath169 must satisfy @xmath170 in summary , * @xmath90 decoded at receivers 1 and 2 @xmath171 where @xmath172 . *
@xmath91 decoded at receiver 1 @xmath173 * @xmath94 decoded at receivers 1 and 2 @xmath174 * @xmath95 decoded at receivers 2 and 3 @xmath175 where @xmath176 . note that @xmath177 , @xmath178 , and @xmath179 .
putting together , we can see that the following rate region is achievable .
@xmath180 where @xmath181 thus , _
theorem 8 _ is proved .
we choose the parameter @xmath182 , which is suboptimal but good enough to achieve a constant gap .
this choice of parameter , inspired by @xcite , ensures making efficient use of signal scale difference between @xmath183 and @xmath184 at receiver 1 , while keeping the interference of @xmath91 at the noise level @xmath184 at receiver 2 . by substitution
, we get @xmath185 since @xmath186 $ ] , it follows that @xmath187 , and @xmath188 .
starting from @xmath189 from table [ tab : outerbounds ] , we can express the two - dimensional outer bound region at @xmath190 as @xmath191 depending on the bottleneck of @xmath192 expressions , there are three cases : * @xmath193 * @xmath194 * @xmath195 . at @xmath196 , the outer bound region is @xmath197 depending on the bottleneck of @xmath192 expressions , we consider the following three cases : * @xmath198 * @xmath199 * @xmath200 .
_ case i _ ) @xmath198 : the outer bound region at @xmath196 is @xmath201 for comparison , let us take a look at the achievable rate region .
the first term of @xmath202 is lower bounded by @xmath203 we get the lower bounds : @xmath204 for fixed @xmath205 and @xmath206 , the two - dimensional achievable rate region is given by @xmath207 _ case ii _ )
@xmath199 : the outer bound region at @xmath196 is @xmath208 now , let us take a look at the achievable rate region .
we have the lower bounds : @xmath209 for fixed @xmath205 and @xmath206 , the two - dimensional achievable rate region is given by @xmath210 _ case iii _ ) @xmath200 : the outer bound region at @xmath196 is @xmath211 for this range of @xmath205 , the rate @xmath190 is small , i.e. , @xmath212 , and @xmath213 and @xmath169 are close to single user capacities @xmath214 and @xmath215 , respectively . let us take a look at the achievable rate region .
the first term of @xmath202 is lower bounded by @xmath216 we get the lower bounds : @xmath217 for fixed @xmath205 and @xmath206 , the following two - dimensional rate region is achievable .
@xmath218 in all three cases above , by comparing the inner and outer bound regions , we can see that @xmath219 , @xmath220 and @xmath221 .
therefore , we can conclude that the gap is to within one bit per message .
given @xmath222 $ ] , the region @xmath32 is defined by @xmath223 and @xmath224 is achievable . for this channel type ,
rate splitting is not necessary .
transmit signal @xmath225 is a coded signal of @xmath226 .
in particular , @xmath94 and @xmath95 are lattice - coded signals using the same pair of coding and shaping lattices . as a result ,
the sum @xmath227 is a dithered lattice codeword .
the power allocation satisfies @xmath228=\alpha_1 np$ ] , @xmath111=np$ ] , and @xmath112=np$ ] .
the received signals are @xmath229+\mathbf{x}_1+\mathbf{z}_1\\ & & \mathbf{y}_2=\mathbf{x}_2+\mathbf{x}_1+\mathbf{z}_2\\ & & \mathbf{y}_3=\mathbf{x}_3+\mathbf{x}_1+\mathbf{z}_3.\end{aligned}\ ] ] the signal scale diagram at each receiver is shown in fig .
[ fig : signalscale ] ( b ) .
decoding is performed in the following way .
* at receiver 1 , @xmath230 $ ] is first decoded while treating @xmath231 as noise .
next , @xmath232 is decoded from @xmath233=\mathbf{x}_1+\mathbf{z}_1 $ ] .
for reliable decoding , the code rates should satisfy @xmath234 * at receiver 2 , @xmath94 is decoded while treating @xmath235 as noise . similarly at receiver 3 ,
@xmath95 is decoded while treating @xmath236 as noise . for reliable decoding
, the code rates should satisfy @xmath237 putting together , we get @xmath238 where @xmath239 starting from @xmath189 from table [ tab : outerbounds ] , we can express the two - dimensional outer bound region at @xmath213 as @xmath240 depending on the bottleneck of @xmath192 expressions , there are three cases : * @xmath241 * @xmath242 * @xmath243 . at @xmath244 , the region can be expressed as @xmath245 depending on the bottleneck of @xmath192 expressions , we consider the following three cases .
_ case i _ ) @xmath246 : the two - dimensional outer bound region at @xmath244 is @xmath247 for fixed @xmath248 and @xmath249 , the following two - dimensional region is achievable .
_ case ii _ ) @xmath251 : the two - dimensional outer bound region at @xmath244 is @xmath252 for fixed @xmath248 and @xmath249 , the following two - dimensional region is achievable .
@xmath253 _ case iii _ )
@xmath254 : the two - dimensional outer bound region at @xmath244 is @xmath255 for fixed @xmath248 and @xmath249 , the following two - dimensional region is achievable .
@xmath256 in all three cases above , by comparing the inner and outer bounds , we can see that @xmath257 , @xmath258 , and @xmath259 .
we can conclude that the inner and outer bounds are to within one bit .
given @xmath260 $ ] , the region @xmath32 is defined by @xmath261 and @xmath262 is achievable . for this channel type , neither rate splitting nor aligned interference decoding
is necessary .
transmit signal @xmath225 is a coded signal of @xmath226 .
the power allocation satisfies @xmath228=\alpha np$ ] , @xmath111=\alpha np$ ] , and @xmath112=np$ ] .
the received signals are @xmath263 the signal scale diagram at each receiver is shown in fig .
[ fig : signalscale ] ( c ) .
decoding is performed in the following way .
* at receiver 1 , @xmath95 is first decoded while treating @xmath231 as noise .
next , @xmath232 is decoded from @xmath264 . for reliable decoding
, the code rates should satisfy @xmath265 * at receiver 2 , @xmath95 is first decoded while treating @xmath266 as noise .
next , @xmath94 is decoded from @xmath267 . for reliable decoding
, the code rates should satisfy @xmath268 * at receiver 3 , @xmath95 is decoded while treating @xmath269 as noise .
for reliable decoding , the code rates should satisfy @xmath270 putting together , we get @xmath271 where @xmath272 starting from @xmath189 from table [ tab : outerbounds ] , we can express the two - dimensional outer bound region at @xmath169 as @xmath273 depending on the bottleneck of @xmath192 expressions , there are two cases : @xmath274 and @xmath275 .
we assume that @xmath275 , equivalently @xmath276 .
we also assume that @xmath277 , equivalently @xmath278 .
the other cases are trivial .
the two - dimensional outer bound region at @xmath279 is @xmath280 for @xmath276 , the two - dimensional outer bound region is @xmath281 for @xmath278 , the two - dimensional achievable rate region at @xmath282 is @xmath283 by comparing the inner and outer bounds , we can see that @xmath284 , @xmath285 , and @xmath286 .
we can conclude that the inner and outer bounds are to within one bit .
the relaxed outer bound region @xmath30 given by @xmath287 the cross - sectional region at a given @xmath213 is described by @xmath288 depending on the bottleneck of @xmath192 expressions , there are three cases : * @xmath241 * @xmath242 * @xmath243 . in this section ,
we focus on the third case .
the other cases can be proved similarly .
if the sum of the righthand sides of @xmath190 and @xmath169 bounds is smaller than the righthand side of @xmath289 bound , i.e. , @xmath290 then the @xmath289 bound is not active at the @xmath213 .
this condition can be expressed as a threshold on @xmath213 given by @xmath291 for this relatively large @xmath213 , the cross - sectional region is a rectangle as described in fig .
[ fig : outerboundcrosssection4 ] ( a ) .
in contrast , for a relatively small @xmath213 , when the threshold condition does not hold , the cross - sectional region is a mac - like region as described in fig .
[ fig : outerboundcrosssection4 ] ( b ) . in the rest of the section , we present achievable schemes for each case . given @xmath292
^ 3 $ ] , the region @xmath32 is defined by @xmath293 where @xmath294 and @xmath295 , and
@xmath262 is achievable .
we present an achievable scheme for the case of @xmath296 .
message @xmath85 is split into three parts : @xmath87 , @xmath86 and @xmath297 , so @xmath298 .
we generate the signals in the following way : @xmath90 and @xmath299 are differently coded signals of @xmath92 , and @xmath91 and @xmath300 are coded signal of @xmath93 and @xmath301 , respectively .
the transmit signal is the sum @xmath302 the power allocation satisfies @xmath110=\alpha_0 np$ ] , @xmath109=\alpha_2 np$ ] , @xmath303=\alpha_1 np$ ] , and @xmath304=(1-\alpha_0-\alpha_1-\alpha_2 ) np$ ] .
the transmit signals @xmath94 and @xmath95 are coded signals of the messages @xmath96 and @xmath97 , satisfying @xmath111=\alpha_2 np$ ] and @xmath112=np$ ] . the signals @xmath299 and @xmath95 are lattice - coded signals using the same coding lattice but different shaping lattices . as a result ,
the sum @xmath305 is a dithered lattice codeword .
the received signals are @xmath306+\mathbf{x}_{12}+\mathbf{x}_{11}+\mathbf{x}_{10}+\mathbf{z}_1\\ & & \mathbf{y}_2=\mathbf{x}_{11}'+\mathbf{x}_{12}+\mathbf{x}_{11}+\mathbf{x}_{2}+\mathbf{x}_{10}+\mathbf{z}_2\\ & & \mathbf{y}_3=\mathbf{x}_3+\mathbf{x}_2+\mathbf{z}_3.\end{aligned}\ ] ] the signal scale diagram at each receiver is shown in fig .
[ fig : signalscale4 ] ( a ) .
decoding is performed in the following way .
* at receiver 1 , @xmath307 $ ] is first decoded while treating other signals as noise and removed from @xmath308 .
next , @xmath300 , @xmath90 , and @xmath91 are decoded successively . for reliable decoding
, the code rates should satisfy @xmath309 where @xmath310 and @xmath311 . note that @xmath177 , @xmath178 , and @xmath179 .
* at receiver 2 , @xmath299 is first decoded while treating other signals as noise .
having successfully recovered @xmath92 , receiver 2 can generate @xmath90 and @xmath299 , and cancel them from @xmath136 .
next , @xmath300 is decoded from @xmath312 .
finally , @xmath94 is decoded from @xmath313 .
for reliable decoding , the code rates should satisfy @xmath314 * at receiver 3 , @xmath95 is decoded while treating @xmath168 as noise .
reliable decoding is possible if @xmath315 putting together , we can see that given @xmath316 $ ] , the following rate region is achievable .
@xmath317 where @xmath318 we choose @xmath319 , @xmath248 and @xmath205 such that @xmath320 , that @xmath321 , that @xmath322 , and that @xmath323 .
it follows that @xmath324 , that @xmath325 , and that @xmath326 .
we get the lower bounds for each term of @xmath202 expression above .
@xmath327 and @xmath328 since @xmath329 , @xmath330 putting together , @xmath331 given @xmath248 , we choose @xmath205 that satisfies @xmath332 . as a result , we can write @xmath333 , and also @xmath334 since @xmath335 , @xmath336 the following rate region is achievable .
@xmath337 for fixed @xmath248 and @xmath338 , the two - dimensional rate region , given by @xmath339 is achievable . in comparison ,
the two - dimensional outer bound region at @xmath340 , given by @xmath341 as discussed above , the sum - rate bound on @xmath289 is loose for @xmath213 larger than the threshold , so the rate region is a rectangle . by comparing the inner and outer bound rate regions
, we can see that @xmath342 and @xmath343 .
therefore , we can conclude that the gap is to within one bit per message .
given @xmath292 ^ 3 $ ] , the region @xmath32 is defined by @xmath344 where @xmath345 and @xmath346 , and
@xmath262 is achievable . for the case of @xmath347
, we present the following achievable scheme . at transmitter 1
, we split @xmath348 into @xmath93 and @xmath92 , so @xmath349 .
the transmit signal is the sum @xmath350 the power allocation satisfies @xmath110=\alpha_0 np$ ] , @xmath109=(\alpha_1-\alpha_0 ) np$ ] , and @xmath304=(1-\alpha_1)np$ ] at receiver 1 , @xmath111=\alpha_2 np$ ] at receiver 2 , and @xmath112=np$ ] at receiver 3 . the signals @xmath299 and @xmath95 are lattice codewords using the same coding lattice but different shaping lattices . as a result ,
the sum @xmath305 is a lattice codeword .
the received signals are @xmath351+\mathbf{x}_{11}+\mathbf{x}_{10}+\mathbf{z}_1\\ & & \mathbf{y}_2 = \mathbf{x}_{11}'+\mathbf{x}_{11}+\mathbf{x}_2+\mathbf{x}_{10}+\mathbf{z}_2\\ & & \mathbf{y}_3 = \mathbf{x}_3+\mathbf{x}_2+\mathbf{z}_3.\end{aligned}\ ] ] the signal scale diagram at each receiver is shown in fig .
[ fig : signalscale4 ] ( b ) .
decoding is performed in the following way .
* at receiver 1 , @xmath307 $ ] is first decoded while treating other signals as noise and removed from @xmath308 .
next , @xmath90 and then @xmath91 is decoded successively . for reliable decoding , the code rates should satisfy @xmath352 where @xmath353 and @xmath354 .
note that @xmath177 , @xmath178 , and @xmath179 .
* at receiver 2 , @xmath299 is first decoded while treating other signals as noise .
having successfully recovered @xmath92 , receiver 1 can generate @xmath90 and @xmath299 , and cancel them from @xmath136 .
next , @xmath94 is decoded from @xmath313 . at receiver 2
, @xmath91 is not decoded . for reliable decoding
, the code rates should satisfy @xmath355 * at receiver 3 , @xmath95 is decoded while treating @xmath168 as noise .
reliable decoding is possible if @xmath315 putting together , we can see that given @xmath356 $ ] , the following rate region is achievable .
@xmath357 where @xmath358 we choose @xmath319 , @xmath248 , and @xmath205 such that @xmath359 , that @xmath360 , that @xmath322 , and that @xmath361 .
it follows that @xmath325 and that @xmath362 .
@xmath363 and @xmath364 putting together , @xmath365 let us define @xmath366 by the equality @xmath367 .
if we choose @xmath368 , then @xmath369 , and @xmath370 we can see that the following rate region is achievable .
@xmath371 for fixed @xmath372 $ ] and @xmath373 , the two - dimensional rate region @xmath32 , given by @xmath374 is achievable .
the union @xmath375}\mathcal{r}_\alpha$ ] is a mac - like region , given by @xmath376 this region is described in fig .
[ fig : maclike ] ( a ) . in comparison , the two - dimensional outer bound region at @xmath377 , given by @xmath378 since @xmath379 , @xmath380 and @xmath381
, we can conclude that the gap is to within one bit per message .
let us consider the relaxed outer bound region @xmath30 given by @xmath382 the cross - sectional region at a given @xmath190 is described by @xmath383 depending on the bottleneck of @xmath192 expressions , there are three cases : * @xmath193 * @xmath194 * @xmath195 . in this section ,
we focus on the third case .
the other cases can be proved similarly .
if the sum of the righthand sides of @xmath213 and @xmath169 bounds is smaller than the righthand side of @xmath384 bound , i.e. , @xmath385 then the @xmath384 bound is not active at the @xmath190 . by rearranging ,
the threshold condition is given by @xmath386 note that @xmath387 is roughly half of @xmath388 .
for this relatively large @xmath190 , the cross - sectional region is a rectangle as described in fig .
[ fig : outerboundcrosssection5 ] ( a ) .
in contrast , for a relatively small @xmath213 , when the threshold condition does not hold , the cross - sectional region is a mac - like region as described in fig .
[ fig : outerboundcrosssection5 ] ( b ) . in the following subsections , we present achievable schemes for each case . given @xmath389
^ 3 $ ] , the region @xmath32 is defined by @xmath390 where @xmath391 and @xmath392 , and
@xmath262 is achievable .
we present an achievable scheme for the case of @xmath393 .
message @xmath96 for receiver 2 is split into two parts : @xmath394 and @xmath395 , so @xmath396 .
we generate the signals in the following way : @xmath397 and @xmath398 are differently coded signals of @xmath399 , and @xmath400 is a coded signal of @xmath401 .
the transmit signal is the sum @xmath402 the power allocation satisfies @xmath228=\alpha_1 np$ ] , at receiver 1 , @xmath403=\alpha_2 ' np$ ] , @xmath404=\alpha_2 np$ ] , and @xmath405=(1-\alpha_2-\alpha_2')p$ ] at receiver 2 , and @xmath112=np$ ] at receiver 3 . the signals @xmath398 and @xmath95 are lattice codewords using the same coding lattice but different shaping lattices . as a result ,
the sum @xmath406 is a lattice codeword .
the received signals are @xmath407+\mathbf{x}_{22}+\mathbf{x}_{21}+\mathbf{z}_2\\ & & \mathbf{y}_3=\mathbf{x}_3+\mathbf{x}_1+\mathbf{z}_3.\end{aligned}\ ] ] the signal scale diagram at each receiver is shown in fig .
[ fig : signalscale5 ] ( a ) .
decoding is performed in the following way .
* at receiver 1 , @xmath398 is first decoded while treating other signals as noise .
having successfully recovered @xmath399 , receiver 1 can generate @xmath397 and @xmath398 , and cancel them from @xmath308 .
next , @xmath400 is decoded from @xmath408 .
finally , @xmath232 is decoded from @xmath231 . for reliable decoding
, the code rates should satisfy @xmath409 * at receiver 2 , @xmath410 $ ] first decoded while treating other signals as noise and removed from @xmath136 .
next , @xmath400 and @xmath397 are decoded successively . for reliable decoding
, the code rates should satisfy @xmath411 where @xmath412 and @xmath413 .
note that @xmath414 , @xmath415 , and @xmath179 .
* at receiver 3 , @xmath95 is decoded while treating @xmath236 as noise .
reliable decoding is possible if @xmath416 putting together , we can see that given @xmath417 $ ] , the following rate region is achievable .
@xmath418 where @xmath419 we choose @xmath248 and @xmath205 such that @xmath420 , that @xmath198 , that @xmath421 , and that @xmath422 .
it follows that @xmath423 .
we get the lower bounds for each term of @xmath424 expression above .
@xmath425 the first entry of @xmath192 in @xmath426 is lower bounded as follows . @xmath427
the second entry of @xmath428 is lower bounded as follows .
@xmath429 putting together , we get the lower bound @xmath430 given @xmath205 , we choose @xmath248 that satisfies @xmath431 . as a result , we can write @xmath432 . we also have @xmath433 putting together , we can see that the following rate region is achievable .
@xmath434 for fixed @xmath205 and @xmath435 , the two - dimensional rate region , given by @xmath436 is achievable . in comparison , the two - dimensional outer bound region at @xmath437
is given by @xmath438 as discussed above , the sum - rate bound on @xmath384 is loose for @xmath190 larger than the threshold , so the rate region is a rectangle . by comparing the inner and outer bound rate regions
, we can see that @xmath439 and @xmath440 .
therefore , we can conclude that the gap is to within one bit per message .
given @xmath441 ^ 2 $ ] , the region @xmath32 is defined by @xmath442 where @xmath443 and @xmath444 , and
@xmath262 is achievable . for the case of @xmath445 , we present the following scheme . at transmitter 2 ,
rate splitting is not necessary .
the transmit signal is the sum @xmath446 where @xmath397 and @xmath398 are differently coded versions of the same message @xmath96 .
the power allocation : @xmath228=\alpha_1 np$ ] at receiver 1 , @xmath403=\alpha_2 np$ ] , and @xmath405=(1-\alpha_2)np$ ] at receiver 2 , and @xmath112=np$ ] at receiver 3 .
the signals @xmath398 and @xmath95 are lattice codewords using the same coding lattice but different shaping lattices . as a result ,
the sum @xmath406 is a lattice codeword .
the received signals are @xmath447+\mathbf{x}_{21}+\mathbf{z}_2\\ & & \mathbf{y}_3=\mathbf{x}_3+\mathbf{x}_1+\mathbf{z}_3.\end{aligned}\ ] ] the signal scale diagram at each receiver is shown in fig .
[ fig : signalscale5 ] ( b ) .
decoding is performed in the following way .
* at receiver 1 , @xmath398 is first decoded while treating other signals as noise .
having successfully recovered @xmath399 , receiver 1 can generate @xmath397 and @xmath398 , and cancel them from @xmath308 .
next , @xmath232 is decoded from @xmath231 . for reliable decoding
, the code rates should satisfy @xmath448 * at receiver 2 , @xmath410 $ ] first decoded while treating other signals as noise and removed from @xmath136 .
next , @xmath397 is decoded from @xmath449 . for reliable decoding
, the code rates should satisfy @xmath450 where @xmath451 and @xmath452 .
note that @xmath414 , @xmath415 , and @xmath179 .
* at receiver 3 , @xmath95 is decoded while treating @xmath236 as noise .
reliable decoding is possible if @xmath416 putting together , we get @xmath453 where @xmath454 we choose @xmath248 and @xmath205 such that @xmath420 , that @xmath198 , that @xmath422 , and that @xmath455 .
it follows that @xmath423 .
we get the lower bound @xmath456 and @xmath457 let us define @xmath458 by the equality @xmath459 . if we choose @xmath460 , then @xmath461 .
we can see that the following rate region is achievable .
@xmath462 for fixed @xmath463 $ ] and @xmath435 ,
the two - dimensional rate region @xmath32 , given by @xmath464 is achievable .
the union @xmath465}\mathcal{r}_\alpha$ ] is a mac - like region , given by @xmath466 in comparison , the two - dimensional outer bound region at @xmath437 is given by @xmath467 since @xmath468 , @xmath469 and @xmath470 , we can conclude that the gap is to within one bit per message .
we presented approximate capacity region of five important cases of partially connected interference channels .
the outer bounds based on @xmath471-channel type argument are derived .
achievable schemes are developed and shown to approximately achieve the capacity to within a constant bit . for future work , the channels with fully general coefficients may be considered . in this paper
, we presented different schemes for each channel type although they share some principle .
a universal scheme is to be developed for unified capacity characterization of all possible topologies .
the connection between interference channel and index coding problems is much to explore . in particular , the results on the capacity region for index coding in @xcite seem to have an interesting connection to our work .
at transmitter 1 , message @xmath348 is split into three parts @xmath472 , and the transmit signal is @xmath473 .
the signals satisfy @xmath303=n(p - n_2-n_3)$ ] , @xmath109=nn_3 $ ] , and @xmath110=nn_2 $ ] . at transmitter 2 ,
message @xmath474 is split into three parts @xmath475 , and the transmit signal is @xmath476 .
the signals satisfy @xmath403=n(p - n_3)$ ] and @xmath477=nn_3 $ ] .
rate - splitting is not performed at transmitter 3 , and @xmath112=np$ ] .
the top layer codewords @xmath478 are from a joint random codebook for @xmath479 .
the mid - layer codewords @xmath480 are from a joint random codebook for @xmath481 .
the bottom layer codeword @xmath91 is from a single - user random codebook for @xmath93 .
the received signals are @xmath482 decoding is performed from the top layer to the bottom layer . at receiver 1 , simultaneous decoding of @xmath483 is performed while treating other signals as noise . and
then , @xmath90 and @xmath91 are decoded successively . at receiver 2 , simultaneous decoding of @xmath484
is performed while treating other signals as noise . and then , simultaneous decoding of @xmath480 is performed . at receiver 3 , simultaneous decoding of @xmath485
is performed while treating other signals as noise . for reliable decoding
, code rates should satisfy @xmath486 at receiver 1 , @xmath487 at receiver 2 , @xmath488 at receiver 3 .
putting together , @xmath489 at the top layer , @xmath490 at the mid - layer , @xmath491 at the bottom layer .
note that the rate variables are not coupled between layers .
we get the achievable rate region @xmath492 this region includes the following region .
@xmath493 therefore , we can conclude the capacity region to within one bit .
transmit signal construction is the same as the one for channel type 4 .
the received signals are @xmath494 decoding is performed from the top layer to the bottom layer . at receiver 1 , simultaneous decoding of @xmath484 is performed while treating other signals as noise . and then , simultaneous decoding of @xmath90 and @xmath495 is performed .
lastly , @xmath91 is decoded . at receiver 2 , simultaneous decoding of @xmath496
is performed while treating other signals as noise . and then , @xmath495 is decoded . at receiver 3 , simultaneous decoding of @xmath497
is performed while treating other signals as noise . and
then , @xmath90 and @xmath91 are decoded successively . for reliable decoding
, code rates should satisfy @xmath498 at receiver 1 , @xmath499 at receiver 2 , @xmath500 at receiver 3 .
putting together , @xmath501 at the top layer , @xmath502 at the mid - layer , @xmath503 at the bottom layer .
note that the rate variables are not coupled between layers .
we get the achievable rate region @xmath492 this region includes the following region .
@xmath504 therefore , we can conclude the capacity region to within one bit .
s. sridharan , a. jafarian , s. vishwanath , s. jafar , and s. shamai , `` a layered lattice coding scheme for a class of three user gaussian interference channel , '' _
46th annual allerton conference on communication , control , and computing , _ pp . 531538 , 2008 .
g. bresler , a. parekh , and d. n. c. tse , `` the approximate capacity of the many - to - one and one - to - many gaussian interference channels , '' _ ieee trans .
theory , _ vol .
9 , pp . 45664592 , sep . 2010 .
a. s. motahari , s. o .- gharan , m .- a .
maddah - ali , and a. k. khandani , `` real interference alignment : exploiting the potential of single antenna systems , '' _ ieee trans .
theory , _ vol .
47994810 , aug .
2014 .
m. p. wilson , k. narayanan , h. pfister , and a. sprintson , `` joint physical layer coding and network coding for bidirectional relaying , '' _ ieee trans .
theory , _ vol .
56415654 , nov . | we derive inner and outer bounds on the capacity region for a class of three - user partially connected interference channels .
we focus on the impact of topology , interference alignment , and interplay between interference and noise .
the representative channels we consider are the ones that have clear interference alignment gain . for these channels ,
z - channel type outer bounds are tight to within a constant gap from capacity .
we present near - optimal achievable schemes based on rate - splitting and lattice alignment .
interference channel , interference alignment , nested lattice code , side information graph , topological interference management . |
recently @xcite , we have studied the collision of two @xmath0-dimensional aichelburg - sexl ( as ) shock waves @xcite , using a method first developed ( in @xmath2 ) by death and payne @xcite , with the goal of obtaining the radiated energy .
this method is conceptually and technically elaborate , involving both analytical and numerical studies . remarkably , the fraction of radiated energy - which we refer to as the _ inelasticity of the collision _ - agrees in first order perturbation theory , within the numerical error of the method ( less than 0.1% ) , with the simple formula _
1st order= - .
[ miracle ] asymptotically , @xmath3 , which agrees with the _ bound _
, @xmath4 , obtained by computing the apparent horizon ( ah ) on the past light cone @xcite ( or on the future one @xcite ) for a head - on collision of two as shock waves : @xmath5 where @xmath6 is the volume of the unit @xmath7-sphere .
moreover , the trend with @xmath0 observed from @xmath8 agrees , qualitatively , with that of @xmath9 ( cf .
fig . 3 in @xcite ) .
an immediate question is if the appeal and simplicity of the result is kept in higher order perturbation theory . to answer it , the @xmath0 dimensional formalism developed in @xcite
must be extended beyond linearised theory and , in particular , so does the radiation extraction method . in @xcite , we have approached the problem of extracting the gravitational radiation by using the landau - lifshitz pseudo - tensor , which was straightforward to apply for the setup therein .
indeed , the first order calculation amounts to approximating the outgoing radiation by an isotropic flux , with a value obtained as the limit of the flux at the axis computed in linearised theory .
this allowed us to obtain a relatively simple expression from the pseudo - tensor , since only the radiation in the direction of the symmetry axis had to be computed . in higher order perturbation theory , on the other hand , we will have to include higher order contributions to the news function , following @xcite , which means that the pseudo - tensor components become more complicated , making this method less manageable and clear
. we shall therefore , in this paper , discuss the higher order perturbative formalism with a different radiation extraction method , valid for generic axially symmetric spacetimes , based on the bondi mass loss formula in @xmath0 dimensions @xcite .
in particular , we shall obtain a formula for the inelasticity which is valid to all orders ( cf . ) , and which makes closer contact with the original payne and death computation @xcite , since it uses the natural generalisation to higher @xmath0 of the bondi news function used by these authors .
in linearised theory , of course , the result obtained using the bondi mass loss formula coincides with the pseudo - tensor method .
the second part of this paper is devoted to the study of a collision of shocks obtained from infinitely boosted charged particles .
such shocks were constructed in @xcite and considered in @xcite to estimate an upper bound on the inelasticity , motivated by tev gravity scenarios @xcite ( see @xcite , sec . 4 , for a recent review ) .
in performing this study , our goal is to probe the validity of a perturbative method to study a process that , at its core , includes a non - perturbative phenomenon - black hole formation .
as it turns out , this example illustrates quite well how the method ceases to work , and creates a clear contrast with the neutral case .
the reason why the method fails in such a case is that the repulsive nature of the charged gravitational source implies that an important contribution to the radiation , in the perturbative approach , is obtained from the strong field region . in reality , this contribution , or at least an important part of it , should be cloaked by a horizon , as it would be manifest in a non - perturbative computation .
thus , in this example , the perturbative approach entangles the physical signal with a significative spurious signal , making the method uninformative .
by contrast , in the neutral case , the spurious signal appears to be subleading .
this paper is organised as follows . in section [ eq : generalframework ]
we shall set up the higher dimensional perturbation theory for the problem at hand and derive the formula for the inelasticity , leaving some technical details , related to the transformations between the various coordinate systems involved , to appendix [ app : bondi ] . in section [ sec_charge ] we discuss the charged case .
althought we have tried to make the discussion self - contained , the construction uses some details provided in @xcite .
two technical issues , the gauge fixing to de donder coordinates and the evaluation of the integral solution for the perturbative metric , were organized as appendices [ gaugetransformation ] and [ app : simplify ] .
we close with some final remarks - section [ sec_final_rem ] .
in @xcite we have shown that , in a boosted frame , the metric on the future light cone of the collision of two higher dimensional aichelburg - sexl shock waves is given by a perturbative series : @xmath10=\nu^{\frac{2}{d-3}}\left[\eta _ { \mu \nu}+\sum_{i=1}^\infty\left(\frac{\lambda}{\nu}\right)^i h_{\mu\nu}^{(i)}\right]\ .\label{eq : pertexpansion}\ ] ] here , @xmath11 are the energy parameters of the weak / strong shock in the boosted frame , respectively , and the background flat metric in null coordinates is @xmath12 @xmath13 is the line element of the unit @xmath14 sphere .
we call this coordinate system _ brinkmann coordinates _
@xcite , where the retarded and advanced times @xmath15 are @xmath16 in terms of minkowski coordinates , and @xmath17 are the remaining cartesian coordinates on the plane of the shocks , @xmath18 , such that the transverse radius is @xmath19 . in this section
we will have two further types of coordinates : i ) _ de donder _ coordinates and ii ) _ bondi _ coordinates , both to be introduced and explained below .
since the de donder coordinates coincide with brinkmann coordinates outside the future light cone of the collision , we adopt the same notation for brinkmann and de donder coordinates ( as in @xcite ) .
the superposition of two shock waves produces boundary conditions on @xmath20 ( the location of the strong shock ) which go up to second order .
thus , in second order perturbation theory , the boundary conditions are exact . suggestively , for @xmath2 , the second order result coincides with the outcome of numerical relativity simulations @xcite .
this sharpens the motivation to pursue this computation ( at least ) to second order .
the general perturbative method consists of the following steps .
once we have the boundary data for the perturbation @xmath21 , we insert the ansatz into the einstein equations and equate order by order .
the components of the metric perturbations do not decouple immediately , but we can perform a gauge transformation to the so called de donder coordinates so that they indeed decouple .
we take the perturbative gauge transformation in the form @xmath22 where the vector @xmath23 is to be determined order by order so that the de donder gauge condition is obeyed ; i.e. for @xmath24 @xmath25 ( barred quantities are trace reversed )
. now @xmath26 are the metric perturbations in de donder coordinates . using and ,
the @xmath7-th order perturbation components obey decoupled wave equations @xmath27 \ , \ ] ] where the source on the right hand side is generated by the lower order perturbations , and can be computed explicitely order by order .
the general integral solution of can be written using the green s function method ( see theorem 6.3.1 of @xcite ) @xmath28 \ , \ ] ] where @xmath29 denotes the finite part of the integral , @xmath30 and we have used the fact that the source only has support in @xmath24 . due to the axial symmetry of the problem , a basis of vectors and tensors on the transverse plane , constructed from @xmath31 and @xmath32 , is @xmath33 where we have chosen the last tensor to be traceless .
then , the metric perturbations in de donder coordinates are decomposed into seven functions of @xmath34 , here denoted @xmath35 , in the following way : @xmath36 with this setup , using the boundary condition on @xmath37 and the solution , one can find the metric perturbations by solving for these scalars after suitable contractions of with the tensors . as mentioned in the introduction
, we shall here construct an energy extraction method which is the higher @xmath0 generalisation of the original method used by death and payne , based on bondi s news function .
the @xmath0 dimensional extension of the news function formalism has been recently addressed in @xcite , where the following mass loss formula for the bondi mass @xmath38 was derived : @xmath39}\dot{\mathfrak{h}}^{[1]\hat i\hat j}d\omega_{d-2 } \ .\ ] ] the dot denotes derivative with respect to the retarded time @xmath40 ; the @xmath41 latin indices are raised with the metric components @xmath42 ; all these quantities , together with the metric functions appearing inside the integral , will be defined in the following .
the geometry we are considering in de donder coordinates has the generic form @xmath43 we can transform it to coordinates @xmath44 , where @xmath45 , @xmath46 and @xmath47 is the angle with the @xmath48 axis and @xmath49 are the angles on the transverse plane , and we have used an index with a prime to denote these new coordinates . then the metric reads : @xmath50 where the various components @xmath51 are defined in appendix [ app : bondi ] in terms of the seven aforementioned scalar functions .
next , we change to bondi coordinates @xmath52 , related to these intermediate coordinates through the transformation @xmath53 such that the new @xmath54 and @xmath55 metric components vanish , and the form of the metric becomes @xmath56 here @xmath57 and the radial coordinate @xmath58 is chosen such that @xmath59 once found such coordinates , assuming that we have an asymptotically flat spacetime with gravitational radiation , it follows that asymptotically @xcite @xmath60}}{\hat r^{d/2+k-1 } } \ , \ ] ] where @xmath61 is the metric on the unit @xmath62-sphere and @xmath63 runs over all integers for @xmath0 even and over semi - integers for @xmath0 odd .
this defines the @xmath64}$ ] components appearing in the mass loss formula . due to the axial symmetry of our problem one needs
not , in fact , transform the angles on the transverse plane , i.e. @xmath65 , and @xmath66 from condition , we can then eliminate @xmath67 in terms of @xmath68 and find that asymptotically @xmath69 so the mass loss formula can be written as a @xmath70 angular integral of the following angular power flux @xmath71 ^ 2 \ ; .\ ] ] using the general form of our metric in de donder coordinates , we have constructed the coordinate transformation in appendix [ app : bondi ] .
one then shows that @xmath72 ^ 2 \ , \ ] ] where all functions are evaluated with bondi coordinates .
an important remark is that our derivation does not rely on the metric being perturbative , and therefore it is valid _ non - perturbatively_. the assumptions are simply : i ) the metric is expressed in de donder coordinates as in eq . ;
ii ) the metric is axisymmetric , cf .
iii ) the spacetime is asymptotically flat and contains gravitational radiation , cf .
eq . .
furthermore , asymptotically , the bondi coordinates used in will approach de donder coordinates , from the construction in appendix [ app : bondi ] .
so , in general , we can express the mass loss formula in de donder coordinates as = - _ r + ^2 .
[ eq : bondiadapted ] observe that the @xmath73 terms correspond to fluxes across @xmath74 surfaces , which are supposed to vanish on @xmath75 , as argued in @xcite , for the problem of gravitational shock wave collisions ( indeed we have checked this numerically ) .
we shall now specialize the general formula , as to facilitate the application of this result to the perturbative problem of shock wave collisions .
we assume spacetime has an adm energy scale @xmath76 , with which we construct a length scale @xmath77 .
then , taking units with @xmath78 , and dividing by the total adm energy scale , the inelasticity factor , corresponding to the fraction of radiated energy into gravitational waves , is @xmath79 with @xmath80 @xmath81 is the generalisation of the bondi news function ( integrated over @xmath82 ) to higher dimensions . in the appropriate limit , this ( more general ) result reduces to the one obtained in @xcite using the landau - lifschitz method . in the particular case of the perturbative method for shock wave collisions ,
the framework is valid for @xmath47 close to @xmath83 ( @xmath84 ) .
then , the approximation taken is usually to expand around the axis , and extrapolate off the axis by integrating the truncated expansion over @xmath47 . besides being axially symmetric ,
our system is invariant under reflections @xmath85 so @xmath81 must be even .
then @xmath86 if the news function @xmath81 is analytic , then the expansion close to the axis is indeed sufficient , provided that @xmath87 .
furthermore , since we have an extra suppression factor in , we expect higher orders to become increasingly less important .
the approximation used by death and payne @xcite in @xmath2 , corresponds to an isotropy approximation so only @xmath88 is used in their result which give ( to second order in perturbation theory ) a result of @xmath89 .
this is in agreement with the latest numerical relativity simulations of ultra - relativistic particle or black hole collisions at large boost @xcite , so it seems to indicate that the angular corrections ( @xmath90 for @xmath91 ) are small in @xmath2 .
[ section2 ] in this section we shall analyze an example of shock wave collisions with an electric charge parameter .
we will apply the formalism developed in the previous section while testing the assumptions of the formalism and commenting on its limitations .
this example will make clear that the perturbative construction is only applicable if the bulk of the gravitational radiation is generated far away from the strongly curved region of space - time .
the geometry of the @xmath0-dimensional reissner - nordstrm solution with mass @xmath92 and charge @xmath93 is @xcite @xmath94 where @xmath95 it is intuitive , as first argued by pirani @xcite , that the gravitational field of a fast - moving mass should become increasingly similar to that of a gravitational plane - wave , as the speed is increased . for the case of a rn ` particle ' , the corresponding aichelburg - sexl shock wave is found by boosting this black hole and then taking simultaneously the limit of infinite boost @xmath96 and vanishing mass and charge , keeping fixed @xcite @xmath97 the resulting geometry for a particle moving in the @xmath98 direction in brinkmann coordinates is @xmath99 where @xmath100 .
the function @xmath101 depends only on @xmath102 and takes the form @xcite @xmath103 where @xmath104 the above coordinates are discontinuous at the shock . transforming to rosen coordinates @xmath105 @xcite , which are continuous at the shock ( see @xcite for the explicit transformation ) , the geometry for two oppositely directed shock waves , with equal charge parameter @xmath106 in , and equal energy @xmath107 may be written everywhere as a simple superposition of the two individual geometries , except in the future of the collision .
moreover , in a boosted frame , moving with respect to the @xmath108 chart with velocity @xmath109 in the @xmath110 direction , the oppositely directed shock waves keep their form , but acquire new energy parameters , respectively , @xmath111 where @xmath112 . in this boosted frame ,
the geometry reads @xmath113d\trho^{2 }
\\ + \trho^2\left[\big(1 + \frac{\nu \tu \ , \theta(\tu)}{\sqrt{2 } \bar \rho}\phi'\big)^2 + \big(1 + \frac{\lambda \tv \ , \theta(\tv)}{\sqrt{2 } \bar \rho}\phi'\big)^2 - 1\right ] d\omega^2_{d-3 } \ , \label{collision}\end{gathered}\ ] ] which is valid everywhere except in the future light cone of @xmath114 .
to set up a perturbative computation and derive the geometry in the future light cone of the collision , we proceed as in @xcite . in the boosted frame one shock
carries much more energy than the other and we thus face the weak shock ( traveling in the @xmath110 direction ) as a perturbation of the geometry of the strong shock ( traveling in the @xmath98 direction ) . the geometry of the latter is flat for @xmath115 ; thus we make a perturbative expansion of the einstein equations around flat space - time in this region .
we shall make the perturbative expansion in brinkmann coordinates .
moreover , we choose to work with the following rescaled dimensionless coordinates @xmath116 and rescaled profile function @xmath117 .
thus , hereafter , @xmath118 and all coordinates are dimensionless .
the boundary conditions for the perturbative computation are given by the geometry ( [ collision ] ) in the limit @xmath119 , yielding only the first two orders in ( notice that these boundary conditions are exact , albeit written in a perturbative form ) : after the coordinate transformation .
this behaviour at @xmath120 is consistent with the propagation of the initial data from @xmath37 , as can be checked numerically . ]
@xmath121 and @xmath122 where @xmath123 . for @xmath124
they are at @xmath125 ; then the generators undergo a discontinuity in @xmath126 at @xmath37 , which is @xmath127 dependent and negative for small @xmath127 .
they jump to the collision surface ( green and brown lines ) .
generically , the rays gain shear and : i ) for large @xmath127 ( green part of the curve ) focus along the caustic ( red line ) ; ii ) for small @xmath127 ( brown part of the curve ) diverge .
undeflected rays are drawn in green . ]
the step function in the previous equations jumps at the collision , which , in brinkmann coordinates , occurs at @xmath37 , @xmath128 . as the weak shock null generators ( traveling along @xmath129 ) reach the strong shock at @xmath37
there is a discontinuity in brinkmann coordinates . to understand what occurs
, we consider the trajectory of a weak shock null generator that , before the collision , obeys the parametric equations u()= , v()=0 , ( ) = .
then , after the shock , its trajectory is given by @xcite u()= , v ( ) = ( ) ( + ) , ( ) = ( 1- ) .
indeed , the @xmath126 coordinate jumps from @xmath125 to the surface @xmath130 , i.e. the _ collision surface _ - fig .
[ lightcone ] . after this jump the @xmath126 coordinate of the trajectory increases , _ except _ at points where @xmath131 . in the uncharged case
there were no extrema of this profile function .
but in the charged case there is one maximum .
the ray incident at the corresponding value of @xmath132 will follow a path of @xmath74 after the collision .
this is possible for a null trajectory because such ray is _
undeflected_. in general the rays are deflected by an angle @xmath133 in the @xmath134 plane , such that = ( 1- ) .
it follows that the weak shock null generators become convergent ( towards the symmetry axis ) for @xmath135 , divergent for @xmath136 and there are undeflected generators for @xmath137 - fig .
this is qualitatively very different from the uncharged case where all rays are convergent ( see fig . 4 in @xcite ) ; the divergent behaviour is caused by the repulsive gravitational effect of the charge .
finally , for an observation point @xmath138 far from the collision and near the axis , fig . [ rays ] suggests four bursts of radiation , associated to the four optical paths that reach @xmath138 .
the order in which the four rays should arrive at @xmath138 defines their numbering .
the first ( second ) ray comes from the same ( opposite ) side of the axis as @xmath138 and from a low redshift region ( cf .
[ lightcone ] ) .
the third ( fourth ) ray comes from the same ( opposite ) side of the axis as @xmath138 and from a high redshift region . .
the introduction of charge leads to a divergent behaviour of the rays for @xmath136 and the existence of the undeflected rays ( green lines ) .
moreover , amongst the convergent rays there is one family of maximal deflection ( dashed red and blue lines ) .
the convergent rays will meet at the axis forming a caustic . for points outside the axis ,
far away from the collision , such as point @xmath138 , the diagram suggests four radiation peaks associated to rays 1 - 4 .
we shall see this interpretation matches the wave forms exhibited in section [ sec_num_rel ] .
an analogous figure was presented in @xcite . ] to the future of @xmath37 , we use the perturbative expansion introduced in section [ eq : generalframework ] in de donder gauge . in first order perturbation theory
, the gauge fixing condition does not affect the radiative components in @xmath139 ; we refer the interested reader to appendix [ gaugetransformation ] for the details . using the results in appendix [ gaugetransformation ]
we find the initial conditions @xmath140 where @xmath141 are , respectively , the charge ( in)dependent parts of @xmath101 . to obtain the relevant scalar functions in first order , @xmath142 ,
in the future ligth cone of the collision , we use the integral solution @xmath143 \ .\ ] ] the source that must be considered in first order perturbation theory , @xmath144 , is the energy momentum tensor associated to the maxwell field of the _ background _ geometry . by taking
the limit described at the beginning of section [ dmetric ] also for the maxwell field , one computes that the energy momentum tensor associated to the shock with support at @xmath37 has a single non - vanishing component : t_uu=(u ) .
[ bsource ] first observe that this energy - momentum tensor _ does not _ have support on @xmath24 .
second , this energy - momentum tensor does not source the radiative components of the first order perturbation @xmath145 , since only the @xmath146 component is non - vanishing .
finally , this energy - momentum tensor is completely taken into account already by the strong shock geometry , and hence by the initial conditions considered .
indeed , the einstein tensor computed from ( [ aise ] ) reads g_uu=(u ) , which , of course , solves the einstein equations with the source ( [ bsource ] ) .
as for the weak shock , it sources an energy momentum tensor with support on @xmath120 that , in principle , contributes to the radiative components via the first term in . in this paper we shall focus on the second term in , which suffices to demonstrate the difficulties in applying the perturbative method to shock waves with charge .
a summary of the determination of the radiative scalar functions @xmath147 , is described in appendix [ app : simplify ] .
we can now discuss the domain for the time integration in ( [ epsilon ] ) .
in the uncharged case , we observed in @xcite that both the beginning of the radiation burst and its peak , as observed at some space - time point @xmath138 in the future of the collision , could be understood by a simple ray analysis , similar to that displayed in figs .
[ lightcone ] and [ rays ] , together with an analysis of the intersection of the past light - cone of @xmath138 with the collision surface .
a similar reasoning for the charged case can be made .
let @xmath138 have space - time coordinates @xmath34 , or , equivalently @xmath148 .
one now observes that an observation point @xmath138 , specified by coordinates @xmath149 , close to the axis and far away from the collision is struck by four rays - fig .
these arrive at retarded times @xmath150 , @xmath151 , where @xmath152 are determined by solving @xmath153 with @xmath154 , for @xmath155 and @xmath156 , and @xmath157 , for @xmath158 and @xmath159 , simultaneously determining the auxiliary variable @xmath160 , which now ( unlike the uncharged case ) has two solutions for each of the two values of @xmath161 , reflecting the two rays coming from each side ( cf .
[ rays ] ) .
a qualitative difference with respect to the uncharged case is , however , that the past light cone of @xmath138 will have a non - vanishing intersection with the collision surface at _ all _ retarded times , corresponding to points very close to the axis .
we shall now argue , however , that this contribution is unphysical and should be neglected .. ] in rosen coordinates ( cf .
section [ section2 ] ) , the collision occurs at @xmath162 . in these ( continuous ) coordinates ,
the future light cone of the collision has therefore two branches : @xmath163 and @xmath164 . in terms of the brinkmann coordinates these conditions read : u=0 , v > , u>0 , v=+u , where = |(1 + ) .
[ branches ] the second of these branches determines a surface which intersects the worldline of @xmath138 at the retarded times @xmath150 computed from ( [ retardedtimes ] ) .
in particular , prior to @xmath155 , the worldline of @xmath138 is _ not _ in the future light - cone of the collision .
thus , the radiation observed along the worldline of @xmath138 before @xmath155 is not causally connected to the collision and consequently we neglect it . + in fig .
[ fig : redwfcharged ] , we display some wave forms for @xmath165 obtained from the numerical integration of the contributions in eqs . , , and .
the integration was done using the numerical code developed in @xcite .
we have found two radiation signals as expected from the geometrical optics analysis .
the left hand side plots , corresponding to the signal coming from rays 1 and 2 in fig .
[ rays ] , coincide precisely with the wave forms computed without charge in @xcite .
the second signal ( right hand side plots ) , is associated to rays 3 and 4 in fig .
[ rays ] ; these are the rays that are incident close to the axis . in a non - linear computation with horizon formation
, a considerable part of this signal should be caught inside the black hole horizon , and should therefore be absent from the viewpoint of an asymptotic observer .
the evidence that supports this statement comes from considering the apparent horizon . indeed , cutting off the @xmath166 integration region inside the exterior apparent horizon ( defined in @xcite )
, the second burst of radiation ( between @xmath156 and @xmath159 ) vanishes .
both re - scaled signals are actually independent of the charge parameter for large @xmath167 .
this should not be suprising for rays that are incident far away from the center of the collision , since for large @xmath127 , the charge contribution to the gravitational field decays faster . however , the second ( anomalous ) re - scaled signal is also independent of the charge parameter so it gives a constant contribution when integrated in time , to @xmath168 . in particular
this means that the result is discontinuous in the @xmath169 limit , as compared to the @xmath170 result .
this is clearly related to the fact that the source is always repulsive for non - vanishing @xmath106 .
we have extracted the two contributions to @xmath168 for large @xmath167 , and found that the contribution from the first wave form coincides ( within a numerical error of less that 1% ) with the @xmath170 computation .
for @xmath171 we found @xmath172 and for @xmath173 , we found @xmath174 . if we add the contribution from the anomalous wave form signal , we get respectively @xmath175 and @xmath176 , independently of @xmath106 . up to now we have focused on the even @xmath0 case .
for @xmath0 odd , we expect the method to become even less meaningful since the green s function for odd @xmath0 has support not only on the past light cone , but also _ inside _ the past light cone . since the shock wave profile becomes repulsive at the center ( see fig .
[ lightcone ] ) , we will get contributions to the integrals which come from highly deflected rays that went through the highly curved and non - linear region of spacetime .
we should also mention that for @xmath2 , the integration of the wave form does not even converge for large @xmath167 to extract a finite @xmath168 .
these results , altogether , indicate the break down of the perturbative method , and clarify its regime of validity . to summarise
, the perturbative method should only capture the relevant physics whenever the optical rays arriving at the observation point go through weak field regions only @xcite .
an ultra - relativistic particle collision is a highly dynamical process and also , if a black hole forms as a result , a strong field process .
it is then quite remarkable if we may use a perturbative method , such as the one originally proposed in @xcite , to extract relevant physics , such as the inelasticity of the process . a good way to test
the method is to generalise it . in @xcite
we have extended it to higher dimensions , revealing a remarkably simple pattern in first order perturbation theory .
second order perturbation theory is then the next goal and , in this problem , one of particular importance . on the one hand
, it will test if the simple pattern observed in @xcite is a special property of linear theory . on the other hand ,
the initial data is exact in second order .
furthermore , to this order , an agreement is found ( in @xmath2 ) with numerical relativity simulations . in this paper
we have paved the road for this second order computation , by presenting the setup for higher order perturbation theory and formulas for extracting the inelasticity , based on a generalisation of the bondi mass formula to higher @xmath0 .
it remains to compute the relevant scalar functions in second order and to perform the numerical integrations .
we shall report on this elsewhere .
another generalisation considered in this paper was the collision of shock waves with a charge parameter , reminiscent of the reissner - nordstrm black holes from which they where obtained .
these collisions have been considered before for apparent horizon computations @xcite , from which bounds on the inelasticity of the process can be obtained .
the generic observed behaviour can be described as follows ( for simplicity we consider only shock waves with equal charge parameter ) .
firstly , including the charge parameter increases the value of @xmath9 , suggesting that a larger fraction of the energy is radiated away .
intuitively , this may be associated to the fact that the charge term in the reissner - nordstrm ( rn ) black hole yields a repulsive effect .
secondly , beyond a certain value of the charge parameter no apparent horizon can be seen .
again , this is expected from the rn solution , which has a limit for the charge to mass ratio , beyond which no event horizon exists .
thirdly , these results are independent of the relative sign of the initial charged particles that were infinitely boosted , in agreement with the observation that gravity is the dominant interaction in trans - planckian scattering @xcite .
the observed behaviour using the perturbative method to first order shows a qualitative agreement with the first and third observations above .
but , as emphasized already , the perturbative method lacks legitimacy in this problem , since there is a considerable amount of radiation that originates in the strong field region , where we have no reason to believe the method .
thus , a more reasonable stance is to regard this charged example as an illustration of how the perturbative method can fail .
, m.s . and c.r .
are funded by fct through the grants sfrh / bd/60272/2009 , sfrh / bpd/69971/2010 and sfrh / bpd/77223/2011 .
the work in this paper is also supported by the fct grants cern / fp/116341/2010 , ptdc / fis/098962/2008 , ptdc / fis/098025/2008 , ptdc / fis/116625/2010 , the marie curie action nrhep295189-fp7-people-2011-irses , the fct strategic project pest - c / ctm / la0025/2011 and by fundao calouste gulbenkian through the `` programa de estmulo investigao . ''
in this appendix we shall address various details related to the coordinate transformations mentioned in section [ eq : generalframework ] . we shall keep the discussion as general as possible , to facilitate its generalization to higher orders in perturbation theory .
the metric perturbations in the coordinates @xmath177 of section [ eq : generalframework ] , are expressed in terms of the scalar functions introduced therein as @xmath178 \ ,
\label{eq : metrictaur}\\ h_{\theta \theta}&=&r^2\left[(\tfrac{a+g}{2}-c ) \sin^2\theta+\sqrt{2}(b - f)\sin\theta\cos\theta+(h-(d-3)e)\cos^2\theta\right ] \ , \nonumber\\ h_{r\theta}&=&r\left[\tfrac{1}{2}\left(a(1-\cos\theta)-g(1+\cos\theta)\right)\sin\theta+(c+h-(d-3)e)\sin\theta\cos\theta+\right .
\nonumber\\ & & + \left.\tfrac{b+f}{\sqrt{2}}\cos\theta+\tfrac{b - f}{\sqrt{2}}\left(\sin^2\theta-\cos^2\theta\right)\right]\ , \nonumber\\ h_{\phi\phi}&=&e+h \ .\nonumber\end{aligned}\ ] ] we consider a generic gauge transformation from de donder coordinates to bondi coordinates @xmath179 assuming only that @xmath180 decays sufficiently fast with some power of @xmath181 , such that bondi coordinates match , asymptotically , de donder coordinates . inserting this transformation in the explicit and implicit dependence on @xmath182 in equation one
obtains that the new metric perturbation in bondi coordinates is asymptotically @xmath183 where the dots denote terms which decay faster asymptotically , the right hand side terms contain the perturbative metric , and the minkowski metric evaluated with bondi coordinates as emphasized by the barred indices .
the three conditions to satisfy are @xmath184 with @xmath185 defined analogously to @xmath186 .
the last condition is only valid , strictly speaking , asymptotically . using a transformation with @xmath187 , we obtain @xmath188 \ , \end{aligned}\ ] ] where @xmath189 and @xmath190 are two arbitrary integration functions and we have used the symbol @xmath191 to denote the primitivation with respect to @xmath58 ( note that the metric functions under the integral decay as an inverse power of @xmath58 ) .
such integrating functions are further constrained by requiring that the new metric functions in bondi coordinates decay at large @xmath58 as in eq . .
the component we need is finally @xmath192 to have the correct asymptotic decay , the contribution from @xmath109 must be zero ; equating the differential operator acting on @xmath193 to zero gives that @xmath194 .
the same applies for the @xmath96 contribution in @xmath195 ; for @xmath2 , however , @xmath96 remains arbitrary .
this is the well known _
freedom referred to in @xcite .
since we are interested in extending the bondi mass loss formula to higher dimensions , and neither of these arbitrary functions affect the components entering the radiative metric components , we set both contributions to zero in the remainder .
the final quantity that we need is @xmath196 , so all that remains is to take @xmath40 derivatives and use the metric functions in evaluated in bondi coordinates .
after doing so , the result will still depend on many of the metric functions we have introduced in .
such functions are constrained by the de donder gauge conditions .
writing down those conditions , changing to @xmath197 coordinates , replacing by hatted coordinates @xmath198 , and taking the asymptotic limit one obtains that @xmath199&=&0 \ , \nonumber\\ \lim_{\hat r\rightarrow+\infty } \hat r^{\frac{d-2}{2}}\left[(1+\cos\hat\theta)\dot g+\sqrt{2}\sin\hat\theta\dot f+\frac{d-2}{2}(1-\cos\theta)\dot h\right]&=&0 \ , \label{eq : gaugeasympt}\\ \lim_{\hat r\rightarrow+\infty } \hat r^{\frac{d-2}{2}}\left[(1-\cos\hat\theta)\dot b+(1+\cos\hat\theta)\dot f+\sqrt{2}\sin\hat\theta\left(\dot c-(d-3)\dot e-\tfrac{d-4}{2}\dot h\right)\right]&=&0\nonumber \ .
\ ; \;\end{aligned}\ ] ] finally one takes @xmath40 derivatives of evaluated with hatted coordinates , and inserts relations to obtain @xmath200 the final result is then @xmath201 which we have used in .
in this appendix we explicitly show that the gauge fixing discussed in section [ sec_gauge_fix ] does not affect the radiative components of the metric . for simplicity of notation
we shall , herein , omit the superscript @xmath202 , in the first order metric perturbations . using the integral solution
we can show that the de donder gauge is preserved in @xmath24 if the following is obeyed on @xmath37 : |h^n _ , v=0 ; [ ddpreserve ] here @xmath203 is a reminder that this is the new de donder metric perturbation .
this condition can be written in terms of the boundary values of the perturbations at @xmath37 : @xmath204 to eliminate the @xmath205derivative , we use the einstein equations @xmath206 \ , \ ] ] to obtain @xmath207}+\frac{1}{2}\eta_{\alpha v}\box\xi^{\gamma}_{,\gamma}=g_{\alpha v}+\frac{1}{2}h_{,\alpha v}-\frac{1}{4}\eta_{\alpha v}\box h+h_{u\alpha , vv}-h_{i\alpha , iv}+\frac{1}{2}h_{\alpha v , ii}\ . \label{a3}\ ] ] moreover , the vanishing of the trace of the einstein tensor implies that @xmath208 ; hence ( [ a3 ] ) becomes @xmath207}=g_{\alpha v}+\frac{1}{2}h_{,\alpha v}+h_{u\alpha , vv}-h_{i\alpha , iv}+\frac{1}{2}h_{\alpha v , ii } \ .\ ] ] let us analyse the various components of this equation . for @xmath209 , we have an identity @xmath210 for @xmath211 , @xmath212}&=&g_{jv}+\frac{1}{2}h_{,jv}+h_{uj , vv}-h_{ij , iv}\nonumber \\ & = & -(2d-5)(d-3)\frac{a}{\kappa^2}\frac{x_j}{\rho^{2d-3}}\sqrt{2}\theta(\sqrt{2}v-\phi)\ ; \end{aligned}\ ] ] for @xmath213 , we need to eliminate the @xmath214 derivative of the trace ; we can do this by using @xmath215 thus @xmath213 reads @xmath216 where @xmath217 \ .\end{aligned}\ ] ] we look for a solution which has a power series expansion around @xmath37 : @xmath218 if @xmath219 , @xmath220 ( at @xmath37 ) , and we get @xmath221 one solution is @xmath222 and : @xmath223 where @xmath224 the only metric components that change under this gauge transformation are @xmath225 and the trace remains unchanged , @xmath226 .
thus , the components of the trace - reversed metric perturbation are @xmath227 since only the transverse components will be relevant for the computation of the radiation , we conclude that the coordinate transformation is of no relevance for that matter .
in this appendix we shall obtain the first order solution for the scalars @xmath228 , obeying the initial conditions presented in section [ sec_gauge_fix ] for the collision of charged shocks .
these are the necessary quantities to compute the inelasticity ( cf .
section [ sec_ext_gr ] ) . using eq . and
the definitions of the scalars , we obtain explicit integral solutions for the scalars e^(1 ) = d()^d-4_-1 ^ 1dx^()(v - v _ ) , h^(1)= d_2()^d-4_-1 ^ 1 dx ( 1-x^2)^^()(v - v _ ) , with @xmath229 then , taking into account the scaling properties of the derivatives of the delta function , we obtain e^(1)=_0^d()^_-1 ^ 1dx^()(x - x _ ) , h^(1)=_0^ d_2()^_-1 ^ 1 dx ( 1-x^2)^^()(x - x _ ) , where @xmath230 similarly , after taking a @xmath126 derivative and using the scaling properties of the derivative of the delta distribution , we obtain e^(1)_,v=_0^d()^_-1 ^ 1dx^()(x - x _ ) , [ wf1 ] h^(1)_,v=_0^ d_2()^_-1 ^ 1 dx ( 1-x^2)^^()(x - x _ ) .
[ wf2 ] finally , @xmath231\frac{\rho^2 - 2\rho\rho'x+\rho'^2}{2u^2}\delta^{\left(\frac{d-4}{2}\right)}(x - x_{\star})\ , \ ; \ ; \ ; \ ; \label{eq : ecommau}\\ h^{(1)}_{,u}&=&-\frac{d-2}{2u}h^{(1)}-\frac{(d-3)(d-4)}{2(d-2)}\frac{\sqrt{2}\omega_{d-4}}{(2\pi\rho)^{\frac{d-2}{2}}}\times\nonumber\\ & & \times\int_0^{\infty } d\rho'\phi'_2(\rho')\rho'^{\frac{d-6}{2}}\int_{-1}^{1 } dx ( 1-x^2)^{\frac{d-5}{2}}\frac{\rho^2 - 2\rho\rho'x+\rho'^2}{2u^2}\delta^{\left(\frac{d-4}{2}\right)}\left(x - x_\star\right)\ .
\label{eq : hcommau}\end{aligned}\ ] ] one important point resulting from the analysis performed so far is that , as for the uncharged case , the relevant functions for the radiation extraction ( [ wf1 ] ) and ( [ wf2 ] ) have an argument with a ( fractional ) derivative of a delta function .
the vanishing of the argument of the delta function has the same interpretation as in the uncharged case , namely the intersection of the past light cone of the observation point with the collision curve . to deal with the @xmath232 integration
it is convenient to separate even and odd @xmath0 . in the following
we shall only consider the even @xmath0 case , since it suffices to illustrate the problems with the method .
@xmath233\delta^{\left(\frac{d-6}{2}\right)}(x - x_{\star})&=&(-1)^\frac{d-2}{2}\dfrac{d^\frac{d-2}{2}}{dx_\star^\frac{d-2}{2}}\left[\left(1-x_\star^2\right)^\frac{d-1}{2}\right]\theta(1-x_\star)\theta(1+x_\star ) \nonumber\\ & \equiv & ( -1)^\frac{d-2}{2}\sqrt{1-x_\star^2}\sum_{n=0}^{\frac{d-2}{2}}a_n x_\star^n\;\theta(1-x_\star)\theta(1+x_\star)\nonumber \ ; \\\end{aligned}\ ] ] @xmath234\theta(1-x_\star)\theta(1+x_\star)+\pi\theta(-1-x_\star ) & , \ d=4\ , \\&\\ ( -1)^\frac{d-2}{2}\dfrac{d^\frac{d-6}{2}}{dx_\star^\frac{d-6}{2}}\left[\left(1-x_\star^2\right)^\frac{d-5}{2}\right]\theta(1-x_\star)\theta(1+x_\star ) & , \
d\geq 6 \ , \end{array}\right.\nonumber\end{gathered}\ ] ] with @xmath235\equiv \sqrt{1-x_\star^2}\sum_{n=0}^\frac{d-6}{2}b_n x_\star^n \ ; \ ] ] @xmath233\delta^{\left(\frac{d-4}{2}\right)}(x - x_{\star})&=&(-1)^\frac{d}{2}\dfrac{d^\frac{d}{2}}{dx_\star^\frac{d}{2}}\left[\left(1-x_\star^2\right)^\frac{d-1}{2}\right]\theta(1-x_\star)\theta(1+x_\star)\nonumber\\ & \equiv&(-1)^\frac{d}{2}\dfrac{1}{\sqrt{1-x_\star^2}}\sum_{n=0}^\frac{d}{2}c_n x_\star^n\,\theta(1-x_\star)\theta(1+x_\star ) \
; \nonumber\\\end{aligned}\ ] ] @xmath236\theta(1-x_\star)\theta(1+x_\star)\nonumber\\ & \equiv&(-1)^\frac{d-4}{2}\dfrac{1}{\sqrt{1-x_\star^2}}\sum_{n=0}^\frac{d-4}{2}d_n x_\star^n\,\theta(1-x_\star)\theta(1+x_\star ) \ ; \end{aligned}\ ] ] @xmath237\frac{\rho^2 - 2\rho\rho'x+\rho'^2}{2u^2}\delta^{\left(\frac{d-4}{2}\right)}(x - x_{\star})\nonumber\\ & & \hspace{2cm}=(-1)^\frac{d-4}{2}\dfrac{d^\frac{d-4}{2}}{dx_\star^\frac{d-4}{2}}\left[\dfrac{d^2}{dx_\star^2}\left(\left(1-x_\star^2\right)^\frac{d-1}{2}\right)\frac{\rho^2 - 2\rho\rho'x_\star+\rho'^2}{2u^2}\right]\theta(1-x_\star)\theta(1+x_\star ) \nonumber\\ & & \hspace{2cm}\equiv ( -1)^\frac{d-4}{2}\dfrac{1}{\sqrt{1-x_\star^2}}\sum_{n=0}^\frac{d+2}{2}g_n x_\star^n\,\theta(1-x_\star)\theta(1+x_\star ) \ ; \end{aligned}\ ] ] @xmath238\theta(1-x_\star)\theta(1+x_\star ) \nonumber\\ & & \hspace{1cm}\equiv ( -1)^\frac{d-4}{2}\dfrac{1}{\sqrt{1-x_\star^2}}\sum_{n=0}^\frac{d-2}{2}h_n x_\star^n\,\theta(1-x_\star)\theta(1+x_\star ) \ .\end{aligned}\ ] ] c. herdeiro , m. o. p. sampaio and c. rebelo , `` radiation from a d - dimensional collision of shock waves : first order perturbation theory , '' jhep * 1107 * ( 2011 ) 121 [ arxiv:1105.2298 [ hep - th ] ] .
f. s. coelho , c. herdeiro and m. o. p. sampaio , phys .
* 108 * ( 2012 ) 181102 [ arxiv:1203.5355 [ hep - th ] ] .
p. c. aichelburg and r. u. sexl ,
`` on the gravitational field of a massless particle , '' gen .
* 2 * , 303 ( 1971 ) .
d. death , p. n. payne , `` gravitational radiation in high speed black hole collisions .
2 . reduction to two independent variables and calculation of the second order news function , '' phys .
* d46 * ( 1992 ) 675 - 693 .
k. tanabe , s. kinoshita and t. shiromizu , `` asymptotic flatness at null infinity in arbitrary dimensions , '' phys .
d * 84 * ( 2011 ) 044055 [ arxiv:1104.0303 [ gr - qc ] ] .
k. tanabe , t. shiromizu and s. kinoshita , `` angular momentum at null infinity in higher dimensions , '' arxiv:1203.0452 [ gr - qc ] .
c. o. lousto and n. g. sanchez , `` the curved shock wave space - time of ultrarelativistic charged particles and their scattering , '' int .
j. mod .
a * 5 * ( 1990 ) 915 .
m. ortaggio , `` ultrarelativistic boost of spinning and charged black rings , '' j. phys .
* 33 * ( 2006 ) 386 [ gr - qc/0601093 ] .
h. yoshino and r. b. mann , `` black hole formation in the head - on collision of ultrarelativistic charges , '' phys .
d * 74 * ( 2006 ) 044003 [ arxiv : gr - qc/0605131 ] .
d. m. gingrich , `` effect of charged partons on black hole production at the large hadron collider , '' jhep * 0702 * ( 2007 )
098 [ arxiv : hep - ph/0612105 ] . n. arkani - hamed , s. dimopoulos , g. r. dvali , `` the hierarchy problem and new dimensions at a millimeter , '' phys .
* b429 * ( 1998 ) 263 - 272 .
[ hep - ph/9803315 ] .
v. cardoso , l. gualtieri , c. herdeiro , u. sperhake , p. m. chesler , l. lehner , s. c. park and h. s. reall _ et al .
_ , `` nr / hep : roadmap for the future , '' arxiv:1201.5118 [ hep - th ] . h. w. brinkmann , `` einstein spaces which are mapped conformally on each other , '' math . ann .
* 18 * ( 1925 ) 119 ; also proc . natl .
sci . u.s .
* 9 * ( 1923 ) 1 . | the collision of two @xmath0-dimensional , ultra - relativistic particles , described in general relativity as aichelberg - sexl shock waves , is inelastic . in first order perturbation theory ,
the fraction of the initial centre of mass energy radiated away was recently shown to be @xmath1 . here ,
we extend the formalism to higher orders in perturbation theory , and derive a general expression to extract the inelasticity , valid non - perturbatively , based on the bondi mass loss formula .
then , to clarify why perturbation theory captures relevant physics of a strong field process in this problem , we provide one variation of the problem where the perturbative framework breaks down : the collision of ultra - relativistic _ charged _ particles .
the addition of charge , and the associated repulsive nature of the source , originates an extra radiation burst , which we argue to be an artifact of the perturbative framework , veiling the relevant physics . |
the topic of relaxation driven by stellar encounters in star systems has a long and distinguished history @xcite and has important implications for the evolution of star clusters @xcite and of active galactic nuclei @xcite .
however , it is widely believed that relaxation through star - star encounters occurs at a negligible rate in the bulk of galaxies , and therefore @xmath0-body simulations of galaxies should mimic this collisionless property .
the study presented here focuses on just one small aspect of the general problem of relaxation in simulations .
although the computational power available to simulators has risen steadily over time , calculations with billions of particles are still not routinely possible and most simulations employ fewer , more massive particles .
a rough estimate of the relaxation time @xmath3 in a simulation of @xmath0 equal mass particles is @xcite @xmath4 with @xmath5 being a typical crossing time .
this , as other more precise expressions for fokker - planck diffusion , contains the coulomb logarithm that arises from integration over impact parameters , implying that every decade of impact parameter makes an equal contribution to the integral .
it should be noted that the approximate formula ( 1 ) applies for pressure - supported systems , not discs @xcite , and neglects collective effects .
furthermore , it was derived for nominally point mass particles , since the argument of the logarithm comes from the ratio of the system half - mass radius to the scale on which scattering causes large deflections , whose contributions would otherwise be overestimated .
it therefore reflects spatial resolution , which is usually determined by other factors , such as particle softening or grid resolution in @xmath0-body codes .
therefore , @xmath3 should be simply proportional to @xmath0 when resolution is held fixed .
numerical methods used to compute the gravitational field in simulations that aspire to be collisionless fall into three broad categories .
the most popular direct method is some type of tree code @xcite that effectively sums the attraction of every particle pair , with a softening kernel to limit the magnitude of the acceleration at short range .
the far more efficient @xcite particle - mesh methods ( pm , * ? ? ?
* ) determine the gravitational field on a raster of points that has some appropriate geometry ; forces at the actual particle positions are computed by interpolation between grid points . finally , the least popular are field methods @xcite that expand the density and potential in a basis set of functions that should be chosen such that truncating the expansion at low order yields an adequate approximation to the field .
weinberg , in a series of papers @xcite has argued that field methods are inherently superior to other @xmath0-body methods in their ability to hide the lumpiness of the potential from a set of point masses . in particular , @xcite assert that field methods have `` relaxation times ... orders of magnitude longer '' than in tree codes .
this claim was based on a lengthy calculation using hamiltonian mechanics that i review below . since a well - chosen basis can yield an adequate approximation to the total field from a small number of terms
, it may seem reasonable to expect the resulting potential to be smoother than that computed by other methods .
however , this argument is beguiling for the following reasons .
only the lowest order monopole term has a large value about which shot noise fluctuations are small , while the amplitudes of the aspherical terms , which are oscillatory , depend on the almost - complete cancellation of the @xmath0 contributions , and are entirely noise - driven in a spherical model , for example .
it is also true that the number of values that define the potential in pm codes is @xmath6 and , since each mesh point typically hosts @xmath7 particles , each separate value will be subject to a greater degree of shot noise . in direct methods , there are @xmath0 separate contributions to the field . but note that the potential is the double integral of the density ; in effect , the potential kernel , which is monotonic and has infinite range , implies the field at each point is the sum of contributions from all the sources .
thus potential variations are far smoother than could be supported by an arbitrary function defined by the same number of values .
* hereafter hb90 ) used a king model for an experimental comparison between the relaxation rates in simulations when the gravitational field was determined by three different methods , and @xcite extended their results to include a field method .
they found that the rate of energy diffusion of particles was only mildly affected by the method used . while their evidence was quite strong , they generally employed only 4096 particles and gravity softening spoiled the equilibrium of some of their initial models .
@xcite showed that grid and field methods performed equally well in the specific problem of bar - halo friction . here
i present a more general study of fully self - consistent equilibrium spheres that uses two distinct measures of the `` relaxation '' rate : the energy diffusion rate reported by hb90 , and a measure of the rate of energy exchange between particle species of differing masses .
the former measure includes all sources of relaxation , especially collective effects , while the latter is a more direct consequence of two - body encounters .
in order to illustrate the importance of collective effects , i here report measurements in three different mass models .
all three are spheres with ergodic ( isotropic ) distribution functions ( dfs ) that are therefore stable equilibria .
the mass models are : a. _ hernquist model _
@xcite developed a simple spherical model having the centrally cusped density profile @xmath8 which has the potential @xmath9 . here
@xmath10 is a length scale and @xmath11 the finite total mass integrated to infinity .
hernquist also gave the equilibrium isotropic df for this density and potential .
i truncate this model so that no particle has sufficient energy to stray beyond @xmath12 , which causes the density profile to taper smoothly to zero at that radius , and discards @xmath13% of the mass , but leaves the density profile as given by eq .
( [ eq.hdens ] ) over the range @xmath14 .
b. _ plummer sphere _
@xcite introduced one of the most widely used spherical mass models in astronomy .
it has the cored density profile @xmath15 where @xmath16 and @xmath11 is the total mass .
the potential is @xmath17 , and the isotropic df is that of a polytrope of index 5 ( see bt08 ) .
applying an energy truncation so that no particle passes outside @xmath12 discards only @xmath18% of the mass in this model with its lower density envelope . c. _ uniform sphere _ with density @xmath19 , where @xmath10 is the outer radius .
@xcite give the isotropic df for a homogeneous sphere with a sharp outer boundary .
the harmonic potential in the interior of this unusual model implies that all particles have the same orbital frequencies ; this model , therefore , affords a dramatic illustration of the role of collective effects . in all cases ,
i employ two equally numerous sets of particles drawn from the df : the masses of particles , @xmath20 are such that those of one sample have @xmath21 and the other have @xmath22 , and @xmath23 is chosen such that the combined density profile is that given by the above expressions .
* appendix a ) describe an optimal method of drawing particle coordinates from a df in such a way as to reduce shot noise in the distribution of energies .
@xcite reports results that show the material benefit of this strategy .
the models are evolved for 100 dynamical times , where @xmath24 , with a timestep of @xmath25 for the uniform sphere and plummer models .
the basic step for the hernquist model is @xmath26 but time steps are increased , in this case only , by four successive factors of 2 at appropriate radii .
i save the instantaneous energy of a representative set of particles ( @xmath27 or all , whichever is the less ) after every dynamical time .
i employ four different numerical methods to determine the gravitational field from the particles : a. * bht * a tree code that uses the scheme first proposed by @xcite .
i include dipole terms and particle groupings are opened when they subtend an angle @xmath28 radians .
forces are softened at short range only using the kernel advocated by @xcite , with @xmath29 for the uniform sphere and @xmath30 for the other two models .
b. * s3d * a spherical grid , which is a hybrid pm expansion method that expands the non - spherically symmetric part in surface harmonics on a set of spherical shells @xcite .
force discontinuities , which arise when particle radii cross , are eliminated by adopting linear interpolation between radial shells .
i tabulate the expansion coefficients at 100 logarithmically spaced radii , and expand in azimuth up to @xmath31 . c. * sfp * a field method that employs a biorthonormal set of basis functions @xcite , which i name sfp ( for smooth field - particle ) but is also known as scf .
i expand in azimuth up to @xmath31 and employ radial functions up to @xmath32 .
i use this method for the plummer and hernquist models only .
d. * c3d * i do not use the sfp method for the uniform sphere , but employ a pm method that uses 3d cartesian grid @xcite , and i set @xmath33 mesh spaces .
the grid has @xmath34 points , except for the lowest @xmath0 case where it was @xmath35 points in order to allow plenty of room for expansion of the particle distribution .
linear interpolation results in the inter - particle force given in ( * ? ? ?
* appendix ) , which is well approximated by cubic density kernel with @xmath36 grid spaces @xcite .
more details of all these methods are given in the on - line manual @xcite .
the choices of numerical parameters in each case are somewhat arbitrary , but values are typical of those used in practice and are not varied as the particle number is changed .
the rate of relaxation will depend only weakly on spatial resolution , since only short range scattering is affected by changes to the softening length , for example , leading to a slight change in the value of the coulomb logarithm . truncating the expansion at low order in field methods smooths the mass distribution , and more aggressive truncation
will give rise to a smoother potential , which must reduce the relaxation rate ; in the extreme case of a single term , the particles will be moving in an almost fixed potential , and no evolution and little relaxation could occur . since the purpose of simulations is to compute the self - consistent evolution as the density changes ,
i include sufficient terms to be able to follow changes that might be expected in an evolving model .
the tree code uses explicit particle softening , while short - range forces are implicitly smoothed in the cartesian grid .
these methods therefore do not yield the exact newtonian potential of the mass distribution . in order to ensure that the initial model is in equilibrium in the tree code , i use the largest @xmath0 simulation of each type to tabulate the difference between the spherically averaged central attraction of the particles at @xmath37 and the analytic central attraction at a 1d array of points
; i then interpolate from this table a supplementary central attraction that is added to the tree - determined force on each particle before its motion is advanced .
i use a similar procedure for the 3d cartesian grid , but in order to avoid shot noise in the ( very small ) permanent part of central attraction , the numerical force is computed from a smooth density distribution assigned to the grid .
these generally small corrections are not needed for s3d or sfp methods . to with these fixed correction terms in c3d and bht , and with a coordinate centre for the s3d and sfp methods , it is important to ensure that the initial set of particles is at rest and centred in this coordinate frame . after creating the initial set of particles ,
i therefore adjust the positions and speeds by a small amount to ensure the centre of mass is at the coordinate centre and the model has no net momentum . an alternative strategy that would achieve the same outcome would be to shift the grid ( or expansion )
centre at frequent intervals .
i also experimented with inserting mirror pairs of particles having coordinates @xmath38 and @xmath39 , in a step towards a quiet start , but this strategy had the undesirable ( for this study ) effects of eliminating any lop - sided , and emphasizing the bi - symmetric , contributions to the total field . the instantaneous energy of the @xmath40th particle is approximately @xmath41 , \label{eq.edef}\ ] ] where @xmath42 is the specific energy of the particle , or energy per unit mass , @xmath43 is the estimated gravitational potential at the particle position , @xmath44 , and @xmath45 is the scalar speed of the particle .
this definition is exact only for a particle of infinitesimal mass , since a finite mass particle contributes to @xmath43 .
the energy required to disperse a system of gravitating particles , the total energy , is @xmath46 , where @xmath47 and @xmath48 , and the @xmath49 term is halved because the summation over the @xmath43 values includes every pair of particles twice .
the total energy , defined this way is very well conserved in the simulations , whereas the sum of the energies ( eq . [ eq.edef ] ) , @xmath50 , is clearly _ not _ the total energy , and is not conserved . in order to test for virial equilibrium of an @xmath0-body simulation , it is better to measure the virial of clausius , @xmath51 , since the inter - particle forces are not perfectly newtonian , especially at short range .
it is easy to show that @xmath52 for precisely inverse - square law accelerations .
the particle distribution is in equilibrium when @xmath53 .
it is convenient to chose units such that @xmath54 , so that the dynamical time @xmath55 , for example .
a convenient scaling to physical units for the inhomogeneous models is to choose @xmath56kpc and @xmath57myr , which implies @xmath58m@xmath59 and velocities scale as @xmath60km / s .
figure [ fig.rho ] shows the final ( @xmath61 ) density profiles in all 27 simulations .
the three panels show the different mass models . within each panel
the curves drawn in red are from the lowest @xmath62 , those in green employ @xmath63 particles , and those in blue @xmath64 , with @xmath65 in each mass species so that the initial densities from each species differ by factors of nine , as shown .
the field is determined on the s3d grid for the full - drawn curves , the bht method for the dotted curves , while the dashed curves are from either the sfp , for the hernquist and plummer models , or c3d for the uniform sphere .
the density profiles of the models evolve insignificantly for the largest @xmath0 ( blue curves ) , confirming that the models are stable equilibria .
as expected , relaxation drives the greatest changes in the simulations with the smallest @xmath0 ( red curves ) where , in a few cases , there are hints of some slight segregation of the particles of different masses over the time interval computed . to figure [ fig.vir ] shows the time evolution of the virial ratio , @xmath66 in all 27 simulations reported here .
the three panels show the different mass models and the colours and line styles are used to distinguish the particle number and field determination method , as in fig .
[ fig.rho ] .
it is clear from figure [ fig.vir ] that the virial ratio @xmath66 remains close to @xmath67 for the duration of all these simulations , and the fluctuations around this ratio are largest for the lowest @xmath0 ( red curves ) and smallest for the highest @xmath0 ( blue curves ) .
the fluctuations are aperiodic for the inhomogeneous models , but are periodic in the uniform sphere , where they have very nearly the same period , @xmath68 , in all nine simulations .
furthermore , the initial behaviour of the curves for models with the same @xmath0 ( colour ) is very similar for the different field determination methods , because the models were set - up using particles having the same positions and velocities .
the red curves diverge quite noticeably , but the blue curves for the scf and s3d methods remain barely distinguishable to the end ; a slightly different value of @xmath69 arises in the bht code because the forces differ due to softening and corrective terms , which is the reason the dotted curves remain distinct , particularly in the hernquist model .
it is helpful to think of these fluctuations as the consequences of collective modes that are driven by shot - noise in the finite number of particles , as was recognized long ago by @xcite in the context of collisionless plasmas , and has been studied in gravitating systems by @xcite and others .
the simplifying concept here is to distinguish the ideal collisionless system , which has a set of neutral and/or damped modes , from the noise spectrum arising from the particles that excites the modes . in principle
, this approach would enable the evolution to be calculated by perturbation theory .
naturally , such an analysis should not differ in its predictions from the direct evolution of an @xmath0-body system composed of equal mass particles , which is exactly what the simulations compute .
the results presented below reveal that relaxation is largely independent of particle mass , indicating that global modes are the dominant relaxation mechanism even when not every particle has the same mass .
the amplitudes of the modes should scale with the shot noise , and the rms scatter in @xmath66 about the value 0.5 indeed scales approximately as @xmath2 .
the modes have essentially the same frequency in the uniform sphere where they appear to be almost undamped .
modes in collisionless systems can be damped only through resonant exchanges between the mode and the particles , and conditions for resonant damping are highly unfavorable because all particles have the same frequencies in this harmonic potential .
the aperiodic fluctuations in the inhomogeneous models ( left and middle panels ) reflect the broader range of frequencies among the particles in these models .
collective oscillations that are excited by shot noise in these cases are almost certainly quickly damped at resonances , yet the amplitudes fluctuate greatly with neither a clear decaying , nor a growing , trend .
particles of each mass when forces are determined using the field method ( sfp ) .
the left panel shows the mean square change of energy of the particles , while the right panel shows the evolution of the mean specific energy ( per unit mass ) of the particles of each mass species . ]
the potential in a hypothetical simulation with infinitely many particles would be smooth and steady , assuming the model is a stable equilibrium , and the specific energy @xmath70 of each particle would be conserved .
therefore the time evolution of the rms changes in @xmath70 is one convenient measure of relaxation in a system with finite @xmath0 .
figure [ fig.typical ] presents a typical set of energy measurements from the particles in a simulation ; values for heavy particles are drawn in red , while green is used for the light particles .
the left panel gives the value of @xmath71 ^ 2\rangle$ ] , the mean square change since the start in the measured specific energy of the particles .
it can be seen that the value of this quantity rises roughly linearly with time , as hb90 found , indicating that the values are changing through a diffusive process .
i fit a straight line to the last 90 values ( ignoring the first 11 , where the rise is often a little steeper ) , and henceforth report only this fitted slope as the mean square change of @xmath70 per dynamical time @xmath72 , and its associated uncertainty .
the changes presented in figure [ fig.typical ] are averages over all the particles .
naturally , the rms changes are larger for those particles whose orbit periods are shorter . as this trend is quite gradual in the inhomogeneous models and non - existent in the uniform models , the global averages i report in all cases are representative of those at intermediate radii where the bulk of the particles are located . the total angular momentum , @xmath73 , of each particle is also conserved in a smooth potential , and its changes afford another measure of relaxation .
i have found that the mean - squared changes in @xmath74 also rise roughly linearly with time , but not quite as smoothly as the @xmath70 changes in the left panel of figure [ fig.typical ] .
furthermore , mass segregation is still less evident in the angular momentum changes .
i therefore focus on the @xmath70 changes for the remainder of the paper .
the right panel of figure [ fig.typical ] gives the time evolution of the mean specific energy @xmath75 of the particles of the separate masses .
, is somewhat higher than that expected from the df , @xmath76 , because the potential well in the simulation determined from the selected particles , which are only 74% of the total mass , is not as deep as the analytic potential of the untruncated model . ]
the rapid variations have the same sign for each particle species because they arise from variations in @xmath77 , as discussed in 3 , that are related to the virial fluctuations ( fig . [ fig.vir ] ) .
they result from evolving potential changes seeded by shot - noise driven fluctuations among the particles .
in addition to these fluctuations , the right panel of figure [ fig.typical ] displays a slow divergence of the mean energies of the heavy and light particles , which represents the gradual exchange of energy that would in the long - run cause the heavy particles to settle to the centre and the lighter to populate the envelope of the model .
the slopes of the straight lines fitted to these data should differ in magnitude by the ratio of the particle masses , since total energy conservation requires the energy lost by the heavy particles to be taken up by the light . in practice , this symmetry is imperfect because of the large measurement uncertainties .
henceforth , i report only the slopes , multiplied by @xmath78 , and their statistical uncertainties .
in systems of gravitating particles , heavy particles lose energy to light particles through dynamical friction @xcite .
physically , a particle is braked by the attraction from its wake
the response of the surrounding sea of particles to its motion . in a system of equal mass particles ,
the frictional drag on each particle is balanced , on average , by the scattering accelerations @xcite and no secular changes occur . but
differences between the braking and acceleration terms cause mass segregation when particle masses are unequal .
were energy exchange between particles of the different mass species the only source of relaxation , the trends in the right panel of figure [ fig.typical ] would not be strongly masked by short term oscillations .
that short - term changes are larger than the gradual diverging trend is evidence that relaxation , measured in the left - hand panel , is being driven mostly by the collective oscillations of the model discussed above that are independent of particle masses .
figure [ fig.main ] summarizes the relaxation measurements from all 27 simulations .
the three rows show the three different mass models ; within each panel there are three different numbers of particles , and for each case , the evolution was computed by three separate force determination methods .
the measured rates from the heavy particles are shown in red , while those from the light particles are marked in green , and the uncertainties in the slopes are indicated by the error bars , that are often too short to be visible .
the different force determination methods used are distinguished by the different symbol types as indicated in the figure caption , and offset horizontally from each other for clarity even though the particle numbers are the same .
these conventions are the same in the left and right panels .
the left panels show the energy diffusion rate , defined as @xmath79 ^ 2\rangle \label{eq.edif}\ ] ] in units of @xmath80 .
the adopted values for @xmath81 , which are @xmath82 , @xmath83 , @xmath84 for the hernquist , plummer , and uniform sphere respectively , are the same for all nine simulations with each model .
the straight dotted line has slope @xmath85 while the slope of the dashed line is @xmath86 ; these lines are not fits to the data and are for comparison only .
the right panels show @xmath87 , the factor @xmath0 for each sub - population is included for clarity in reality , the slopes roughly decrease as @xmath88 .
there are many conclusions to be drawn from these data .
first , the differences between the energy diffusion rates ( left panels ) for the heavy ( red ) and light ( green ) particles within one simulation are generally small .
this is a further indication of the dominance of coherent potential variations , which cause deflections that are independent of particle mass , in driving this measure of relaxation .
second , the energy diffusion rate declines roughly as @xmath2 in the uniform sphere , whereas in the inhomogeneous models it declines more or less with the expected @xmath1 dependence for collisional relaxation at fixed resolution ( see 1 ) .
if two - body effects were dominant in all cases , the variation with @xmath0 should be the same in all three mass models .
third , the rates of energy exchange shown in the right hand panels generally have opposite signs , and vary roughly as @xmath1 , since they are approximately constant when multiplied by @xmath0 .
the values from the highest @xmath0 experiments are quite uncertain , because the decreasing variation in the mean as @xmath0 rises is masked by short - term changes ( see the right panel of fig .
[ fig.typical ] ) the trends decrease into the noise .
this is particularly problematic for the uniform sphere ( bottom right panel ) , where the potential changes associated with pulsations of the model dominate over the energy exchange rate at the highest @xmath0 , making an accurate measurement over the time interval simulated impossible .
fourth , the different methods used to compute the gravitational field yield broadly similar behaviour in all three mass models , and the variation of the rates with @xmath0 in both the left and right panels is similar for each method in each of the different models .
fifth , the energy diffusion rate ( left panels ) is consistently lower , but by a factor @xmath89 , for the sfp method ( diamonds ) than for the tree code ( crosses ) and the spherical grid ( circles ) in the hernquist and plummer models .
this systematic difference is also present in the variations of @xmath74 .
@xcite also reported a slightly lower energy diffusion rate when they used a field method , although it was unclear whether the more rapid diffusion in their tree code , for example , resulted mainly from the mild disequilbrium of their initial model caused by gravity softening .
however , the mass segregation rate ( right panels ) , which is a more direct consequence of two - body scattering , is no less for sfp than for the other methods .
sixth , relaxation in the uniform sphere probably is completely dominated by the oscillatory modes excited by the shot noise in the particle distribution .
the initial amplitudes of the modes should scale as @xmath2 , and the decline in relaxation rate with approximately this dependence in the uniform sphere is a direct indication of their dominance in this case .
collective modes are present in all models , but the different @xmath0-dependence in the inhomogeneous models is probably because modes are rapidly damped in those cases .
seventh , short - range gravity softening appears to have little effect on the relaxation rate .
there is no explicit softening in sfp , and the only smoothing on the spherical grid is linear interpolation in radius .
force softening is explicit in the tree code and implicit in the c3d grid , where forces are slightly sharper @xmath90 while @xmath91 for the bht code for the uniform sphere . yet
the relaxation rates scarcely differ between the various codes .
this emphasizes the dominance of distant encounters , enhanced by collective oscillations , in driving relaxation and that softening s only value is to avoid large accelerations during close encounters between particles , which would require short time steps to integrate the motions accurately . once again
, the quantities shown in figure [ fig.main ] are averages over all the particles , and these conclusions therefore apply to the particles in the bulk of the model .
the small number of more tightly bound particles have the largest energy changes , but by factors of only a few , and therefore have little overall effect on the global average .
note also that their rate of energy change should be reckoned on the time - scale of their shorter orbit periods .
the finding stated in point four above is in strong disagreement with that of @xcite , who argued that the relaxation time - scale is orders of magnitude longer in field methods .
it should be noted that their analysis addressed the more limited problem of resonant exchanges between particles and a perturbing bar potential , not the more general relaxation studied here .
however , a discrepancy remains to be explained since @xcite , in a study that addressed their specific prediction , also found that field methods were not superior to grid methods .
their calculation is highly technical but , in essence , they began by separating large- from small - scale noise and calculated separate contributions to the orbit deflections from both these `` components '' of the noise . rather surprisingly , they found that deflections due to large - scale noise were much weaker than those from small - scale noise .
they then argued that direct codes possess noise on both scales while field methods are affected only by large - scale noise , which led them to conclude that relaxation should be far slower in field methods .
however , it is well - known that noise is present on all scales and , since every decade of impact parameter contributes equally to the coulomb logarithm @xcite , there can be no clear distinction between large- and small - scale noise .
this also implies that relaxation from large - scale noise should be comparable to that from small - scale noise .
@xcite found good agreement between experimental relaxation rates and theoretical predictions from fokker - planck diffusion coefficients , suggesting that self - gravity of the collective modes does not boost the relaxation rate significantly .
furthermore , his predicted rate for the plummer sphere is in good agreement with the value i find ( fig .
[ fig.main ] , middle row , left panel ) .
however , such studies usually consider inhomogeneous models .
the anomalous results presented here for the uniform sphere show that relaxation _ is _ greatly enhanced by collective modes in this unusual case where all orbits have the same period .
note that @xcite also found enhanced relaxation through collective effects in their study of a system of low - mass particles in near keplerian motion about a central mass . in their case
, it was important that particles with the same @xmath74 had the same period .
the reason for the slightly lower diffusion rate in the field method ( point five above ) is unclear .
both the sfp and s3d methods employ an expansion in surface harmonics up to @xmath31 to capture any angular variations , and the principal difference between them is radial resolution : the sfp method employs 11 radial functions ( @xmath92 ) , while the s3d uses a grid of @xmath93 radial shells .
it therefore seemed that the marginally higher energy diffusion rate when the s3d grid was used was because that grid could support more oscillatory modes .
however , increasing the number of radial functions used in the sfp method to 51 , in order to enable the field method also to support more collective modes , led to little change in either measure of the relaxation rate , confounding this theory !
additional experiments with @xmath94 and @xmath95 caused very slight changes in the measured relaxation rates in the expected sense , but the systematic discrepancy between the two methods persisted .
while the puzzle remains , the negative outcome of these tests is further evidence of the unimportance of self - gravity for collective oscillations in inhomogeneous models .
the main result here confirms that previously found by @xcite and by @xcite that the rate of relaxation in @xmath0-body simulations that aspire to be collisionless is very largely independent of the method used to compute the gravitational field .
this is especially true when the relaxation rate is assayed as the rate of energy exchange between particles of different masses see the right hand panels of fig .
[ fig.main ] .
this result is physically reasonable , since relaxation is dominated by distant encounters and any method that correctly yields the field from distant particles must faithfully include their stochastic contributions .
the rate of relaxation arises from at least two distinguishable sources : a slow diffusion of the integrals caused by coherent potential oscillations of the system that are largely independent of particle mass , and mass segregation that is more directly caused by two - body encounters .
the good agreement ( 5.2 ) between fokker - planck diffusion and @xmath0-body experiment in the inhomogeneous models indicates that self - gravity of the collective modes adds little to the relaxation rate .
the rate of energy exchange between particles of differing masses becomes more difficult to measure as @xmath0 rises because mean energies scatter with potential variations that scale as @xmath2 , while the trends in the mean energies of the different mass species diverge as @xmath1 .
the different @xmath0-dependence in the uniform sphere ( bottom left panel of fig .
[ fig.main ] ) is clear evidence that collective modes _ do _ dominate the relaxation in this special case .
this model differs from the other two by having a harmonic potential throughout in which the orbit frequencies of all particles are the same ; this difference permits undamped collective oscillations , whereas collective modes are damped in the other cases . the slightly lower energy diffusion rates that result from use of the field method ( diamonds in top left and middle left panels of fig .
[ fig.main ] ) is bought at a high price .
the leading term in the basis used for both cases was a perfect match to the equilibrium model , and the same basis would be less suited were the density to evolve , or were it used for any other model , requiring more terms to yield the correct total potential .
thus field methods lack the versatility to follow arbitrary changes to the distribution of mass within a model unless the expansion is taken to higher order , and their slightly better relaxation rate can be achieved in any of the more general methods simply by employing a few times more particles .
a real advantage of field methods , featured by @xcite , is that they offer an elegant comparison of simulations with perturbation theory when computing first order changes to an equilibrium model that could be unstable or externally perturbed .
the energy diffusion time - scale , the inverse of the rate defined in eq .
( [ eq.edif ] ) and plotted in fig .
[ fig.main ] , is @xmath96 dynamical times ( @xmath97 yr for the suggested scaling ) for only @xmath98 particles in the inhomogeneous models ( it is shorter for the exotic case of the uniform sphere ) . by this measure ,
relaxation times in simulations of inhomogeneous 3d models using any valid code are already significantly longer than the ages of the galaxies being simulated , and relaxation considerations alone do not require much larger numbers of particles .
the author wishes to thank tad pryor for many insightful conversations and the referee for a perceptive report .
i also thank eugene vasiliev for helpful comments on an earlier draft and martin weinberg for an extensive e - mail correspondence .
this work was supported by nsf grant ast/1108977 . | i present empirical measurements of the rate of relaxation in @xmath0-body simulations of stable spherical systems and distinguish two separate types of relaxation : energy diffusion that is largely independent of particle mass , and energy exchange between particles of differing masses .
while diffusion is generally regarded as a fokker - planck process , it can equivalently be viewed as the consequence of collective oscillations that are driven by shot noise .
empirical diffusion rates scale as @xmath1 in inhomogeneous models , in agreement with fokker - planck predictions , but collective effects cause relaxation to scale more nearly as @xmath2 in the special case of a uniform sphere .
i use four different methods to compute the gravitational field , and a 100-fold range in the numbers of particles in each case .
i find the rate at which energy is exchanged between particles of differing masses does not depend at all on the force determination method , but i do find the energy diffusion rate is marginally lower when a field method is used .
the relaxation rate in 3d is virtually independent of the method used because it is dominated by distant encounters ; any method to estimate the gravitational field that correctly captures the contributions from distant particles must also capture their statistical fluctuations and the collective modes they drive .
[ firstpage ] galaxies : kinematics and dynamics methods : numerical |
symmetry reduces complexity . in physical systems realizing quantum computers ,
the highest degree of symmetry is therefore not the most desirable .
a quantum computer needs to be sufficiently simple and robust to be controllable in an experiment yet complex enough to be universal .
one may therefore ask the question `` how much symmetry does a quantum computer allow for ? '' .
in fact , a number of physical systems considered for the realization of a quantum computer such as optical lattices @xcite or arrays of micro - lenses @xcite are translation invariant , and the above question acquires a practical aspect . quite surprisingly , it turns out that universal quantum computation can tolerate a fair amount of symmetry .
recently , a scheme of quantum computation using the rotation - invariant measurement of the total ` spin ' of two qubits as the only gate has been devised @xcite .
furthermore , a translation - invariant computation scheme has been described @xcite .
the computational power of translation - invariant or nearly translation invariant quantum systems was revealed in lloyd s @xcite and watrous @xcite work on quantum cellular automata ( qca ) . in @xcite
it was shown that a one - dimensional qca can simulate any quantum turing machine .
translation invariance is broken only by the initial state which encodes the program .
the schemes @xcite and @xcite allow to simulate quantum circuits using a chain of qubits with a generic translation - invariant interaction .
they require different species of qubits in a periodic arrangement and local addressability at one end - point of the chain . in the scheme
@xcite such individual addressing is only required in the initialization .
the method proposed in @xcite is completely translation invariant in space .
it requires homogeneous one- and two - local operations on 5-level systems . here
i describe a scheme for universal quantum computation via translation - invariant operations on a chain of qubits .
no individual addressability is required .
the scheme uses an ising - type interaction and spatially uniform one - qubit gates .
the qubits are all of the same species .
cold atoms in optical lattices @xcite , where the requirement of local control adds to the overall technological challenge , are a candidate for the realization of the presented scheme .
consider a one - dimensional chain of @xmath0 qubits initialized in the state @xmath1 which is repetitively updated according to the transition function @xmath2 that is , in each elementary step of the evolution first a hadamard gate is applied to each qubit and second , conditional phase gates are simultaneously applied to all pairs of neighboring qubits .
this qca transition function has previously been discussed in @xcite . between the transitions one may apply translation - invariant unitary transformations of the form
@xmath3 with @xmath4 .
( note that the subscript ` @xmath5 ' labels the site .
the same operation is applied to each qubit . )
these requirements are equivalent to bang - bang pulses of form ( [ pulses ] ) and a permanent ising - type interaction @xmath6 .
let us first observe that @xmath7 where @xmath8 is the reflection operator that sends the state of the qubit chain into its mirror image .
thus , despite the fact that the qubits at the end points need not be addressed , it is relevant that the chain _ has _ ends .
( for adaption of the scheme to a ring of qutrits , see remark 1 in section [ cr ] ) . a proof of ( [ rev ] ) is given in the appendix .
apart from its use in the computational scheme described here , the bit reversal operation @xmath8 is interesting in its own .
recently , proposals for both approximate and perfect bit reversal in qubit chains with a heisenberg- and @xmath9- interaction have been made ; see @xcite-@xcite and references therein . for perfect mirror reflection in an @xmath9-chain the coupling strength needs to vary with position @xcite , but only mildly @xcite .
relation ( [ rev ] ) represents a method to achieve spatial reflection in systems with an ising - type interaction . here , the interaction strength is independent of position , but additional stroboscopic pulses ( [ pulses ] ) are required to realize the uniform hadamard transformations . [
[ universal - set - of - gates . ] ] universal set of gates .
+ + + + + + + + + + + + + + + + + + + + + + + the key to the construction of a universal set of gates for the described automaton is displayed in fig .
the @xmath10-fold repetition of the elementary transition function constitutes half a clock cycle of the automaton . within this period each local pauli
observable goes through the phases of expansion , transmission , and contraction to the mirror image of the initial position . during expansion and contraction ,
the propagated observable is susceptible to a global @xmath11-pulse @xmath12 namely , it picks up a sign factor under conjugation .
contrarily , during the phase of transmission where the observables behave as left- or right - movers @xcite a @xmath11-pulse has no effect .
the duration of expansion , transmission and contraction phase depend on the initial position of the local pauli observable .
therefore , each local pauli observable shows a characteristic response to sequences of @xmath11-pulses within the half - cycle . in this way
, temporal control can be translated into spatial control . in the construction described below , suitably tailored sign - flips of pauli observables are used to reverse rotation angles .
depending on whether a rotation angle is reversed or not , two matching rotations will either cancel or amplify each others effect .
the degree of spatial control obtained suffices to simulate a quantum logic network .
one constraint arises : every operation applied to the qubit chain is reflection - symmetric .
therefore , qubit @xmath5 can not be addressed separately from its mirror image at location @xmath13 .
to cope with this constraint , the same network is simulated twice on the chain , once on the left side and once as mirror image on the right .
this doubles the required length of the chain and also influences the readout process , as will be discussed in section [ construct ] @xmath14 . to probe the available spatial control ,
consider the sequence @xmath15 where @xmath16 is a binary vector . to analyze the effect of this sequence ,
the pauli operators @xmath17 are propagated backwards in time until they reach the @xmath18-gates . there
they accumulate to @xmath19 , @xmath20 then , @xmath21 where we have used that @xmath22 , and @xmath23=0\\1 , & \mbox{if}\ ; \{\overline{y},z_i\ } = 0\end{array}\right . .\ ] ] in this way , temporal control has been converted into spatial control , provided that for suitable choices of @xmath24the binary variables @xmath25 do indeed vary with @xmath5 .
the @xmath26 are easily computed in the stabilizer formalism @xcite-@xcite .
following @xcite , we write pauli operators @xmath27 in the form @xmath28 , where @xmath29 is a 2@xmath0-component binary vector , @xmath30 , @xmath31 ; @xmath32 , and @xmath33 . the evolution of @xmath27 under conjugation by our clifford unitary @xmath34 , @xmath35 ,
may then be followed in terms of @xmath36 , @xmath37 and @xmath38 . the scalars @xmath36 and @xmath37 have no influence on the sign factor @xmath39 ( [ exchange ] ) , and we need to consider the update of @xmath38 only , @xmath40 .
therein , @xmath41 is a @xmath42 binary symplectic matrix which takes the form @xmath43 @xmath44 is the adjacency matrix of the interaction graph ( the line graph ) .
further , denote by @xmath45 the @xmath46-matrix @xmath47 and observe that @xmath48 .
now , the vector @xmath49 carrying the information about the sign flips under conjugation by @xmath19 is related to the vector @xmath50 describing the temporal sequence of @xmath11-pulses , via @xmath51 .
the matrix @xmath52 encodes how temporal control is converted to spatial control .
its elements are given by @xmath53 the binary quantities @xmath54 can be defined for all @xmath55 ( @xmath56 ) .
but the matrix @xmath52 is the collection of @xmath54 only within the half - cycle @xmath57 . for the interval
@xmath58 $ ] ( including @xmath59 $ ] ) the @xmath54 take the values @xmath60 therein , the addition is modulo 2 and the step function @xmath61 is equal to one for all non - negative arguments , and zero otherwise . to prove ( [ spatial2 ] ) one may
in addition to @xmath52define a matrix @xmath62 that encodes the sign factor acquired by the @xmath55-steps - backward - propagated observable @xmath63 under conjugation with @xmath19 , @xmath64 .
then , because of @xmath65 and @xmath66 , the @xmath54 and @xmath67 obey the recursion relations @xmath68 the solution of the recursion relation for @xmath54 is unique once the boundary conditions on two consecutive time slices are specified . in the discussed case ,
the boundary conditions are @xmath69 , for all @xmath56 .
the expression on the r.h.s of ( [ spatial2 ] ) obeys the recursion relation ( [ recrel ] ) in the interval @xmath70 and the boundary conditions , and is thus the correct expression for @xmath52 . to perform a @xmath71-rotation on qubit @xmath5
, one may apply a @xmath11-pulse at times @xmath72 and @xmath5 . for @xmath73 , with ( [ spatial2 ] ) ,
one obtains @xmath74 .
therefore , the pulse for a @xmath71-rotation of qubit @xmath5 is @xmath75 with @xmath76 the mirror site of @xmath5 . in this way ,
two out of @xmath0 @xmath71-rotations can be selected .
a further discrimination between two sites @xmath5 and @xmath77 is not possible because of the reflection symmetry of every sequence @xmath78 . as a result of this symmetry ,
each quantum algorithm is run on the qubit chain in two copies , one being the mirror image of the other .
this accounts for a factor of two in spatial overhead . from ( [ zgate ] )
the remaining universal gates can be deduced easily .
relation ( [ zgate ] ) is conjugated by @xmath79 . using @xmath80
, one finds @xmath81 conjugating relation ( [ xgate ] ) again by @xmath79 and noting that @xmath82 one obtains @xmath83 now , the length of the chain is doubled a second time .
specifically , the logical qubits in state @xmath84 are interlaced by ancilla qubits which remain in the state @xmath85 throughout the computation , @xmath86 at this point it is suitable to introduce a logical coordinate @xmath87 $ ] that is related to the physical location within the qubit chain via @xmath87=2j-1 $ ] , for @xmath88 , such that e.g. @xmath89,[2], .. ,[n]}\otimes |0 .. 0\rangle_{2,4, .. ,2n-2}$ ] . then , the action of the @xmath90 gate is equivalent to a two - qubit next - neighbor entangling gate , @xmath91}x_{[j+1]}\right)|\psi^\prime\rangle.\ ] ] the gates ( [ zgate ] ) , ( [ xgate ] ) and ( [ xxgate ] ) form a gate set that can be easily converted into the standard universal set @xcite . [ [ readout . ] ] readout . + + + + + + + + as an example for global readout i consider a measurement of the @xmath71-component of the total spin , i.e. , of the observable @xmath92 this is a model for readout e.g. of atomic qubits via resonance fluorescence spectroscopy , if the atoms are well within the lamb - dicke limit ( their separation and fluctuation of position are much smaller than the optical wave length ) . in this case
the underlying physics prohibits the extraction of local information .
the readout method described below equally works for a scenario where a local readout could in principle be performed but is not pursued due to technological limitations .
for specification of such a measurement model , see remark 3 in section [ cr ] . with the capability to perform arbitrary unitary evolution on a quantum register the total spin observable @xmath93 acting on that register
may be conjugated into any desired observable , and the standard network readout of the individual @xmath94 should be feasible . however , for the setting described here the readout procedure is , in the case of probabilistic algorithms , slightly complicated by the fact that two copies of the algorithm are run simultaneously and the readout measurement couples them . for the readout each of the two circuits on the chain
requires one additional logical qubit that is in the state @xmath85 until the readout starts .
the location of this qubit within the logical quantum register is denoted as @xmath95 $ ] which shall correspond to the physical position @xmath96 in the qubit chain ( see fig .
[ seq]a ) .
the readout consists of three steps .
first , @xmath93 is measured and the outcome @xmath97 is recorded .
second , for all logical qubits @xmath98 the following procedure is performed : 2.1 ) apply cnot gates @xmath99,[0]}\otimes \lambda(x)_{\overline{[j]},\overline{[0]}}$ ] ( [ 0 ] , @xmath100}$ ] are the target qubits ) .
2.2 ) measure the observable @xmath93 and record the outcome @xmath101 .
2.3 ) apply the cnot gates @xmath99,[0]}\otimes \lambda(x)_{\overline{[j]},\overline{[0]}}$ ] again .
third , denote by @xmath102 those @xmath103 for which @xmath104 . if @xmath105 the readout is finished
otherwise , for all @xmath106 perform the following procedure : 3.1 ) apply the toffoli - gates @xmath107,[j_k];[0 ] } \otimes \lambda_2(x)_{\overline{[j_1]},\overline{[j_k]};\overline{[0]}}$ ] ( [ 0 ] , @xmath100}$ ] are the target qubits ) , 3.2 ) measure @xmath93 and record the measurement outcome @xmath108 .
3.3 ) apply the toffoli - gates @xmath107,[j_k];[0 ] } \otimes \lambda_2(x)_{\overline{[j_1]},\overline{[j_k]};\overline{[0]}}$ ] again .
the conditional phase- and toffoli gates are their own inverse such that the above protocol amounts to a sequence of measurements where the first one is of @xmath93 and the following ones are of conjugated observables .
these observables mutually commute because they are all diagonal in the computational basis , and the final state is a simultaneous eigenstate of them .
the observables measured in step 2 are @xmath109}z_{[j]}+z_{\overline{[0]}}z_{\overline{[j]}}+\sum_{i \neq [ 0],\overline{[0]}}z_i$ ] . since the qubits at locations @xmath95 $ ] and @xmath100}$ ] are individually in the state @xmath85 , and the observable @xmath110 already has been measured in step 1 , step 2 effectively acts as the measurement of the observables @xmath111}+z_{\overline{[j]}}$ ] . similarly , step 3 amounts to the measurement of the observables @xmath112,[j_k]}\langle11| + |11\rangle_{\overline{[j_1]},\overline{[j_k]}}\langle11|$ ] , @xmath113 .
this measurement assigns a sharp value to @xmath114}r_{[j_k ] } + \overline{r}_{[j_1]}\overline{r}_{[j_k]}$ ] .
its purpose is , in combination with the information gathered in step 2 , to discriminate between the two cases @xmath115}=r_{[j_k ] } , \overline{r}_{[j_1]}=\overline{r}_{[j_k]}\}$ ] and @xmath115 } \neq r_{[j_k ] } , \overline{r}_{[j_1 ] } \neq \overline{r}_{[j_k]}\}$ ] , for each @xmath116 .
the state of the two quantum registers after the measurement , leaving out all the ancillas in the state @xmath85 , may be written as a superposition @xmath117 of computational basis states @xmath118 , where @xmath119},
.. ,r_{[n]})^t$ ] , @xmath120 } } , .. , r_{\overline{[n]}})^t$ ] .
the measurements then impose constraints upon which basis states are admissible , i.e. , occur with nonzero coefficients @xmath121 .
the constraints imposed by the measurements in step 2 read @xmath122}=\overline{r}_{[j]}=0 , & \mbox{for}\,\ , m - m(j)=0,\\ r_{[j]}+\overline{r}_{[j ] } \mod 2=1 , & \mbox{for}\,\ , m - m(j)=2,\\ r_{[j]}=\overline{r}_{[j]}=1 , & \mbox{for}\,\ , m - m(j)=4 , \end{array}\ ] ] for all @xmath123 .
the constraints that arise from the measurements in step 3 are @xmath124}+r_{[j_k ] } \mod 2=0 , & \mbox{for}\,\ , m - m(j_1,j_k)=2,\\ r_{[j_1]}+r_{[j_k ] } \mod 2=1 , & \mbox{for}\,\ , m - m(j_1,j_k)=0 , \end{array}\ ] ] for @xmath125,[j_k ] \in j$ ] , @xmath106 .
the system ( [ constr1 ] , [ constr2 ] ) of linear equations has a unique solution if @xmath126 and two solutions otherwise . in the latter case ,
the two solutions @xmath127 and @xmath128 correspond to two states @xmath129 and @xmath130 , and the state of the quantum register after readout is their reflection - symmetric linear combination .
the solutions @xmath127 and @xmath128 are , modulo interchange , uniquely specified by the measurement outcomes @xmath97 , @xmath131 and @xmath132 . for a probabilistic algorithm , the automaton produces two potential solutions in each run .
[ [ overhead . ] ] overhead .
+ + + + + + + + + for the described method of universal computation via translation - invariant interaction it is relevant that the elementary transition function @xmath34 repeated a sufficient number of times reduces to the identity operation .
such an elementary ` clock cycle ' takes @xmath133 time steps where @xmath0 is the length of the chain .
but as described so far , one still ends up with a set of local and next - neighbor gates and thus pays the price of next - neighbor slowdown twice .
this is not necessary .
before the resources are counted , a long - distance entangling gate is described .
as compared to a logically equivalent composition of next - neighbor and local gates , it cuts the required operational resources by a factor of order @xmath134 .
the natural gate set for the described scheme of quantum computation does not only contain local and next - neighbor gates , but all gates of the form @xmath135}\right).\ ] ] as before , the logical position @xmath136 $ ] corresponds to the physical location @xmath137 in the chain .
the sequence for implementing @xmath138 is @xmath139 with @xmath140 , @xmath141 and @xmath142 .
now note that , with @xmath143 , @xmath144 thus , a long - distance @xmath145-controlled spin - flip can be implemented with 4 elementary gates ( [ elmx ] ) .
these four operations can be grouped into two pairs , with constant value of @xmath146 in each pair .
either pair of operations can be implemented in one clock cycle because the @xmath11-pulse sequences of the two operations coincide .
an @xmath145-controlled spin - flip thus takes two clock cycles irrespective of the distance between control and target qubit .
now counting the resources , a quantum computation on @xmath134 qubits in the described scheme requires a qubit chain of length @xmath147 .
each clock cycle takes @xmath148 elementary time steps ( applications of @xmath34 ) , and each of the universal gates @xmath149 , @xmath150 and @xmath151 takes at most two clock cycles .
as compared to the network model , there arises a constant overhead of @xmath152 in the spatial and a linear overhead of @xmath153 in temporal resources .
universal quantum computation can be performed by translation - invariant operations on a chain of qubits initialized in the state @xmath1 .
the described method requires translation - invariant interaction of ising type and spatially uniform local rotations .
as compared to the network model , there occurs a constant spatial and a linear temporal overhead .
three remarks : 1 ) in the described setting translation invariance is broken by the finite extension of the qubit chain . to create a setting with perfect translational invariance
, one may replace the qubits by qutrits , where the additional level @xmath154 does not take part in any interaction , and consider a @xmath10-qubit ring geometry instead of a chain .
set @xmath155 . with either the translation - invariant superposition @xmath156 or mixture @xmath157 as the initial state the scheme works as before .
+ 2 ) it may occur that a qubit chain capable of the required interaction has been created but its length @xmath0 is unknown .
it can be found out easily by repetitions of the following protocol : initialize the chain in the state @xmath1 , apply the transition function @xmath34 t times and subsequently measure the spin observable @xmath158 .
the received signal @xmath159 carries a characteristic signature of @xmath0 , namely @xmath160 .
+ 3 ) deviating from the setting discussed in section [ construct]@xmath161 , the available measurement may be such that local information could in principle be retrieved but is not retrieved , due to technological limitations .
as an example , consider cold atoms in an optical lattice which are read out via resonance fluorescence spectroscopy and whose separation is larger than the wave length of the probing laser . in this case
, the readout can be modeled by the map @xmath162 with the projectors @xmath163 , and @xmath164 .
every atom is individually measured but only the _ global _ signal @xmath165 is recorded . for measurements of this type
the readout procedure is the same as discussed in section [ construct ] and has the same efficiency .
i thank frank verstraete , luming duan , gm jeremia , robin blume - kohout , gavin brennen , ren stock and sergey bravyi for discussions .
this work was supported by the multidisciplinary university research initiative program under grant no daad19 - 00 - 1 - 0374 .
99 d. jaksch , c. bruder , j.i .
cirac , c.w .
gardiner , and p. zoller , phys .
lett . * 81 * , 3108 ( 1998 ) .
g. birkl , f.b.j .
buchkremer , r. dumke , w. ertmer , opt . comm . * 191 * , 67 ( 2001 ) .
t. rudolph and s.s .
virmani , new j. phys . * 7 * , 228 ( 2005 ) .
vollbrecht , e. solano , and j.i .
cirac , phys .
lett . * 93 * , 220502 ( 2004 ) ; k.g.h .
vollbrecht and j.i .
cirac , quant - ph/0502143 ( 2005 ) .
s. lloyd , science * 261 * , 1569 ( 1993 ) .
j. watrous , in _ proceedings of the 36th ieee symposium on foundations of computer science _
, 528 ( 1995 ) .
benjamin , phys .
88 * , 017904 ( 2002 ) .
g. ivanyos , s. massar , a.b .
nagy , phys .
a * 72 * , 022339 ( 2005 ) .
b. schumacher and r.f .
werner , quant - ph/0405174 ( 2004 ) .
s. bose , phys .
lett * 91 * , 207901 ( 2003 ) . c. albanese , m. christandl , n. datta , and a. ekert , phys .
lett . * 93 * , 230502 ( 2004 ) .
p. karbach and j. stolze , quant - ph/0501007 ( 2005 ) .
d. gottesman , _ stabilizer codes and quantum error correction _
, phd thesis , california institute of technology , quant - ph/9705052 ( 1997 ) .
nielsen and i.l .
chuang , _ quantum computation and quantum information _ , cambridge university press ( 2000 ) .
j. dehaene and b. de moor , phys rev .
a * 68 * , 042318 ( 2003 ) .
et al . _ ,
proc . lett . * 75 * , 101 ( 2000 ) . g.k .
brennen and j.e .
williams , phys .
a * 68 * , 042311 ( 2003 ) .
sufficiently many applications of the elementary transition function @xmath34 reverse the order of qubits within the chain . as stated in ( [ rev ] ) , with @xmath8 the spatial reflection , @xmath166 this phenomenon is related to the transmission of one - qubit states through chains described in @xcite .
relation ( [ rev ] ) is now proved .
what needs to be shown is @xmath167 and @xmath168 , for all @xmath169 . for this purpose , the vector space formulation @xcite-@xcite of the stabilizer formalism
in particular , i use the conventions and results of @xcite .
the evolution of the pauli observables @xmath170 , @xmath171 is followed in terms of the binary quantities @xmath38 and @xmath37 , introduced in section [ construct ] below eq .
( [ exchange ] ) .
first , the recursion relation @xmath172 is translated into a recursion relation for the @xmath173-part of @xmath38 alone and a relation that expresses @xmath174 in terms of @xmath173 . from ( [ c ] ) , taking into account that @xmath44 represents a line graph , the following recursion relations are obtained for @xmath30 and @xmath31 : @xmath175 we seek the solution of these recursion relations for the time @xmath176 , with boundary conditions @xmath177 the translation - invariance of the recursion relations ( [ recur ] ) is broken by the finite extension of the chain .
the problem of solving these recursion relations is now reduced to the translation - invariant case of the infinite chain .
note that if a configuration @xmath178 obeys the recursion relations @xmath179 then @xmath180 obeys the recursion relation ( [ recur ] ) .
the reduction of the case with spatial boundary to the case without boundary is illustrated in fig .
[ foldback ] .
it is easily checked that @xmath181 is the solution of ( [ infi ] ) with the boundary conditions @xmath182 , @xmath183 . if this expression is inserted into the r.h.s . of ( [ reduc ] )
one obtains a solution for @xmath184 with boundary conditions ( [ bc ] ) .
it is observed that @xmath185 next , the sign factor @xmath189 in the above relation is worked out . using the result of @xcite ,
the recursion relation for @xmath37 reads @xmath190 , with @xmath191 the lower triangular part of @xmath44 . with ( [ checkerboard ] ) , @xmath192 such that @xmath168 . finally , with @xmath193 , @xmath194 . | a scheme of universal quantum computation on a chain of qubits is described that does not require local control .
all the required operations , an ising - type interaction and spatially uniform simultaneous one - qubit gates , are translation - invariant . |
extremely sensitive interferometric measurements are implied in several frontier experiments . for instance , large baseline gravitational wave detectors presently surpass the limit of @xmath1 hz@xmath2 in relative displacement spectral noise @xcite .
an improvement of more than one order of magnitude is planned for their advanced versions @xcite , and a sensitivity of @xmath3 m/@xmath4 is envisaged in the khz range for possible future massive , cryogenic detectors @xcite . on the small scale ,
very low - noise background is also obtained in small scale quantum opto - mechanics experiments @xcite . with such extreme requirements ,
critical limits are imposed by several noise sources associated with the interferometer optical components @xcite , and a crucial role is played in particular by the high - reflection coating .
for instance , brownian thermal noise of the dielectric layers is the most stringent foreseen limiting factor in the @xmath5khz range for the next generation of gravitational wave detectors .
another important phenomenon is the photothermal dynamic effect : an ( even small ) fraction of the light impinging on a mirror is absorbed and heats it , causing a local temperature change and , through thermal expansion , a deformation of the mirror surface .
a fundamental limit to the interferometer sensitivity is therefore produced by the shot noise of the absorbed radiation , and such limit can be worse in the case of excess of laser - power fluctuations .
this effect , together with other dissipative phenomena such as photoelastic and brownian noise , has been studied theoretically by braginsky _
_ @xcite and cerdonio _ et al .
_ @xcite in the case of a gaussian beam impinging on a half - infinite homogeneous mirror . in spite of being an annoying noise source ,
photothermal dynamic displacement can be exploited , using controlled laser power variations , for measuring the thermo - mechanical parameters of the mirror .
this possibility is particularly important to investigate the characteristics of the coating .
thin film parameters are generally different from those of the bulk .
moreover , they depend on the film morphology , and finally on the techniques used for deposition .
a photothermal analysis gives therefore useful information for understanding the general behavior of the coating and selecting the optimal fabrication procedure . at this purpose , it is important to analyze in particular the low - losses coating and substrates of the kind used in interferometers .
dynamic photothermal effects were observed in a high - finesse fabry - perot cavity , probably for the first time , by an _ _ et al.__@xcite .
they describe an hysteretic behavior in the cavity transmission when the laser is rapidly scanned , and use it to estimate the mirror absorption .
a more detailed spectral analysis exploiting an amplitude modulated laser source is reported by de rosa _ _ et al.__@xcite , who verify the theoretical prediction of refs .
@xcite and @xcite concerning the photothermal expansion of the mirror substrate .
the study is extended in ref .
@xcite , where on one hand the dependence of the photothermal response on the beam waist , foreseen in ref .
@xcite , is verified ; on the other side , a deviation from the above theoretical expression is observed at low frequencies ( @xmath6 hz ) due to the limited mirror size , and at high frequencies ( @xmath7 hz ) due to the coating .
the effect of the coating is also shown by black _ _ et al.__@xcite , who apply a gold film on the dielectric coated mirror to enhance the absorption at the expenses of the cavity finesse .
their dielectric film is composed of alternate layers of sio@xmath8 and tio@xmath8 , and their results are in agreement with a simple heuristic model approximating the coating as an homogeneous layer . in view of their possible use in the future gravitational wave detectors ,
it is important to check the photothermal effect also on substrates of different materials ( particularly interesting are sapphire and silicon ) and at cryogenic temperature .
such conditions , explored by the pioneering work of the tama group@xcite , would greatly improve the sensitivity ( at the expenses of evident technical complications ) . in the present work we describe a detailed study of photothermal dynamic effect in a high - finesse cavity , where the mirror coating is very close to those conceived for gravitational wave detectors .
our attention is indeed particularly focused on the behavior of the dielectric films , but we also check the effect on a silicon substrate .
our investigation is performed both at room and at cryogenic temperature , and a theoretical model is developed in order to extract the most reliable information from the experimental data .
the photothermal expansion is ruled by the thermal diffusion equation and the equation of elasticity .
the mirror surface displacement is then read with a weighted average that , in the case of a gaussian beam , can be written as @xmath9 where @xmath10 is the surface normal displacement , @xmath11 are the transverse coordinates of the mirror surface , @xmath12 is the beam radius measured in correspondence of a power drop by @xmath13 from its central value .
the solution of the problem in the case of an half - infinite homogeneous mirror , heated and interrogated by a gaussian beam partially absorbed on the mirror surface , is calculated by braginsky , gorodetsky and vyatchanin @xcite ( bgv ) .
a more conclusive form of the photothermal noise spectral density is given by cerdonio _
et al . _
@xcite in the framework of an extended analysis of thermoelastic noise sources .
their expression can be transformed to give the response to a modulated absorbed power @xmath14 , that in the fourier space reads @xmath15 ( @xmath16 is the fourier transformed of @xmath17 ) where @xmath18 , @xmath19 , @xmath20 are respectively the thermal expansion coefficient , the poisson ratio , and the thermal conductivity of the mirror ; @xmath21 , where @xmath22 is the density and @xmath23 the specific heat capacity ; the shape of the response is @xmath24 the requirements for the validity of the above expression ( e.g. , a short mean free - path of the phonons , and a large sound velocity ) are discussed in details in refs .
@xcite . at high frequencies ( @xmath25 , the so - called adiabatic limit )
, @xmath26 tends to @xmath27 .
this regime occurs when the thermal diffusion length in one modulation period is much smaller than the beam radius . at even higher frequencies ,
when this length is comparable with the coating thickness , the approximation of homogeneous mirror fails and a more complete model is necessary . for an immediate understanding of the physics involved , it is useful to consider a one - dimensional model , starting from the thermal conduction equation that reads @xmath28 where @xmath29 is the temperature , the mirror occupies the space with @xmath30 , and @xmath31 is the absorbed power . in the frequency domain , the above equation can be written as @xmath32 where we have introduced a complex wave number @xmath33 the inverse of the absolute value of @xmath34 gives the penetration depth of the thermal wave . in the case of an absorbed intensity @xmath35 on the surface at @xmath36 ,
the boundary condition is @xmath37 and , in a multi - layer medium , at each interface @xmath38 the continuity conditions for temperature and heat flux are @xmath39 the general solution of the homogeneous equation associated with the ( [ thdiff ] ) is @xmath40 in the case of a mono - layer with thickness @xmath41 and associated wave number @xmath42 , lying on a substrate with wave number @xmath43 , the temperature is given by eq .
( [ exp ] ) on the layer and by @xmath44 \label{exp1}\ ] ] on the substrate , where the coefficients @xmath45 , @xmath46 , and @xmath47 are derived from the boundary conditions ( [ bound1 ] ) and ( [ bound2 ] ) .
the temperature on the layer can also be written in a compact form as @xmath48+\mathcal{r}\cosh[k_c ( d - z)]}{\cosh[k_c d]+\mathcal{r}\sinh[k_c d ] } \label{tlayer}\ ] ] where @xmath49 and the subscripts @xmath50 and @xmath51 distinguish respectively the thermal parameters of the substrate and the layer .
the pre - factor determining the temperature in the substrate is @xmath52+\mathcal{r}\sinh[k_c d ] } \,\,\ , .
\label{bs}\ ] ] we remark that eq .
( [ bs ] ) differs from the solution in the absence of the layer ( i.e. , with @xmath53 ) by a filter function @xmath54+\mathcal{r}\sinh[k_c d ] } \label{filtro}\ ] ] which describes the modification of the heat flux caused by its transmission through the mono - layer .
the filter function can be generally defined as @xmath55 where in general the heating power can be absorbed both on the surface and within the medium : @xmath56 as suggested by bgv , the mirror surface deformation can be assumed to be @xmath57 in this way , the one - dimensional solution differs from the adiabatic approximation only for the lacking term @xmath58 , if one uses for @xmath35 the peak intensity @xmath59 .
it is useful to notice that integrating eq .
( [ thdiff ] ) along an homogeneous layer and using eqs .
( [ filtrogen ] ) and ( [ iabsgen ] ) we obtain @xmath60 this expression can be used to write the surface displacement in a compact form , even without performing the direct integration of the temperature along the layer . indeed , calculating the mirror thermal expansion according to eq .
( [ uz ] ) and using eq .
( [ thexpc ] ) and the exponentially decreasing temperature in the substrate , with the pre - factor of eq .
( [ bs ] ) , we obtain @xmath61\tilde{i}_{abs } \qquad .\ ] ] the above equation is correct in the one - dimensional approximation , that can be relaxed assuming a gaussian beam with radius @xmath62 , an absorbed intensity @xmath63 , and correcting the expansion coefficient of the substrate with the factor @xmath58 .
observing that @xmath64 and defining @xmath65 similarly to @xmath66 , we can write @xmath67\delta p_{abs}(\omega ) \ , . \label{x1dim}\ ] ] on the right side of the above expression , the first term is due to the expansion of the layer , while in the second term we see the expansion of the substrate caused by the filtered thermal flux . besides the filter , the second term coincides with the adiabatic limit of eq .
( [ xbrg ] ) , as in the case of an homogeneous mirror .
a smooth transition to the homogeneous , 3-dimensional model can therefore be simply obtained by replacing the factor @xmath68 with the function @xmath69 defined in equation ( [ g ] ) .
a more rigorous expression for @xmath70 is calculated in the appendix a using a 3-dimensional formalism for both the thermal conduction and the elastic equations .
the calculation considers an half - infinite mirror with a superimposed layer , thus neglecting the finite mirror dimensions .
the result is summarized in eqs .
( [ b41]-[b46 ] ) .
one further thermo - optical phenomenon that must be considered is the thermo - refractive effect : heating changes the optical length of the coating layers , thus modifying the phase of the reflected field and therefore the position of the effective reflection plane with respect to the physical position of the mirror surface .
we remark that an increased optical length gives a longer effective cavity ( while the thermal expansion brings to a shorter physical cavity ) , therefore the thermo - refractive effect has in general an opposite sign with respect to photothermal expansion .
the attention to the thermo - refractive effect has been drawn by braginsky _
@xcite , who identify it as source of thermo - elastic noise . a correct expression for the thermo - refractive coefficient in the case of our interest ( i.e.
, a coating starting from the high index layer , without cover layer ) is reported in ref . @xcite .
since the effect is mainly due to the first few layers , we can adopt the simplified expression valid for infinite number of layers and using the temperature at the coating surface : @xmath71 where the thermo - refractive coefficient is @xmath72 with @xmath73 @xcite . in the framework of a one - dimensional model
, the temperature @xmath74 can be taken from the expression ( [ tlayer ] ) using @xmath36 .
as for the thermal expansion , the contribution to the signal @xmath75 is calculated using the spatial average given by eq .
. a more accurate calculation in a three - dimensional framework is reported in appendix a , and the result is summarized in eq .
( [ b48 ] ) .
so far we have only considered surface absorption .
in fact , the light is absorbed in the coating , where the electromagnetic field forms a standing wave , with nodes / antinodes in correspondence of the interfaces between layers and a rapidly decreasing amplitude .
moreover , the absorption is in general different in the two materials composing the coating .
the exact solution of the field within the coating can be calculated using a characteristic matrix formalism @xcite . for a high - reflectivity coating
( with many layers ) , the average intensity in the first layer is reduced with respect to the intracavity value by a factor very close to the refractive index @xmath76 of the material ( in our case , ta@xmath8o@xmath77 ) . in each following layer ,
the intensity is further reduced with respect to the previous one by a factor @xmath78 , where @xmath79 is the index of the second material ( for us , sio@xmath8 ) . if we consider that each layer is long @xmath80 , where @xmath81 is the vacuum optical wavelength for which the coating is designed , we see that the absorption in the first period is @xmath82 , where @xmath83 is the extinction length in the h ( l ) material , and @xmath84 is the intracavity power . in each following period , the absorption is reduced by a factor @xmath85 , therefore the total absorption in one mirror is @xmath86 where @xmath87 for our refractive index values of @xmath88 and @xmath89 , about 74@xmath90 of the absorption occurs in the first two periods . in brief , we observe that a ) in the two mirrors the absorption is the same , since the kind of substrate and the number of layers are not critical parameters to determine the intensity decay within the coating ; b ) the two material composing the coating have the same weight in evaluating the total absorption , i.e. , the absorption in each material is just determined by its extinction length , independently from its thickness ; c ) for our parameters , the approximation of surface absorption is well satisfied since , even at the highest experimental frequencies , the thermal penetration depth corresponds to several periods .
the optical cavity is made with a ofc ( oxygen free copper ) spacer , a corning 7980 fused silica @xcite concave mirror ( 50 mm radius ) as input mirror and a flat , silicon rear mirror .
mirrors diameter and thickness are respectively 12.7 mm and 6 mm .
the cavity optical length is 10.7 mm , the beam waist on the flat mirror is 0.083 mm and the corresponding beam size on the input surface is 0.094 mm .
the input mirror has a coating composed of alternate quarter - wave layers of ta@xmath8o@xmath77 and sio@xmath8 ( their physical thickness is respectively 130 nm and 180 nm ) .
the overall number of layers is 29 , giving a transmission of about 100 ppm and a total coating thickness of about 4.5 @xmath91 m .
the rear mirror has a similar coating , with 34 layers ( 5.3 @xmath91 m ) and a lower transmission ( @xmath92 ppm ) .
the cavity spacer is a thick box , with just two 10 mm diameter holes for the optical access , and few smaller holes for evacuating .
it is fixed in thermal contact with the cold finger inside a continuous flow @xmath93he cryostat which is evacuated down to @xmath94 pa .
temperature sensors are placed on the cold finger and on the cavity spacer . to assure an efficient cooling and temperature stability of the cavity mirrors
, they are mounted inside the ofc box .
a spring keep them in contact with the box through their rear , flat surface .
the standard thermal shield of the cryostat has four , 25 mm diameter access holes .
we have further reduced the apertures to few mm , with aluminum foil . as a consequence , the mirrors ` see ' the room temperature through couples of holes ( on the spacer and the shield ) , under a small solid angle .
the experimental setup for the measurement of the photothermal response is sketched in fig .
( [ setup ] ) .
the light source is a cw nd : yag laser operating at 1064 nm .
after a 40 db optical isolator , the laser radiation is split into two beams . on the first one ( probe beam ) , a resonant electro - optic modulator ( eom1 ) provides phase modulation at 13.3 mhz with a depth of about 1 rad used for the pound - drever - hall @xcite ( pdh ) detection scheme .
the probe beam is then frequency shifted by means of two acousto - optic modulators ( aom ) .
we use respectively the + 1 and -1 diffracted orders of the aoms , so that the total frequency displacement corresponds to the difference of the aom frequencies and can be tuned by several mhz around zero .
the intensity of the second beam ( pump beam ) is controlled by an electro - optic modulator ( eom2 ) followed by a polarizing beam splitter .
both beams are sent to the second part of the apparatus by means of single - mode , polarization maintaining optical fibers .
part of the pump is then detected by a photodiode ( pd2 ) for monitoring its amplitude modulation , while the probe is sent through a second optical isolator ( o.i .
the two beams are overlapped with orthogonal polarizations in a polarizing beam - splitter and sent to the optical cavity . the reflected probe beam , on its back path ,
is deviated by the input polarizer of the o.i.1 and collected by a photodiode ( pd1 ) for the pdh detection .
the pump , slightly detuned with respect to the cavity resonance , is used as excitation beam in the thermo - optical response measurements .
such frequency splitting between pump and probe allows to eliminate any spurious interference and reduce the cross - talk between them in the photo - detection . a quarter - wave plate before the cryostat allows to compensate cavity birefringence observed at low temperature and probably due to stress on the mirrors generated by differential contractions .
the laser is weakly locked on the cavity using the pdh signal while the pump beam is amplitude modulated at different frequencies .
the extraction of the photothermal response is obtained from the same pdh signal , that is acquired by a lock - in amplifier which discriminates the signal component synchronous with the modulation .
the depth and phase of the intracavity power modulation is calculated from the signal of pd2 , which is also acquired and de - modulated by the lock - in amplifier , by taking into account the mode matching and the cavity coupling factor .
the latter is inferred from the depth of the dip in the reflected beam when scanning on the resonance , and it is consistent with the independent measurements of cavity finesse and input mirror transmission .
the pdh signal is calibrated in a separate run , stopping the pump beam and modulating the frequency of the laser through its internal piezoelectric crystal , with a depth smaller than the fabry - perot cavity linewidth .
the usual lock - in acquisition and phase - sensitive detection is applied to the pdh signal .
the frequency modulation rate had been previously measured by observing the sidebands at 13.3 mhz with a second , high - finesse optical cavity .
the calibration procedure , performed varying the modulation frequency , allows also to correct the data for the servo loop .
the locking bandwidth is of the order of 1 khz .
typical values of the optical parameters are : a mode - matching of 80@xmath90 ; a finesse of 25000 ( linewidth @xmath95 khz ) ; an intracavity optical power @xmath84 equal to 7000 times the input pump power @xmath96 mw .
the input probe beam power is approximately 0.2 mw .
the photothermal response to intensity modulation of the intracavity field has been investigated both at room and cryogenic temperatures in the 10 hz @xmath0 100 khz frequency range .
the measured signal is reported in terms of changes of the optical path divided by the amplitude of the modulation in the intracavity power , giving a response in m / w .
previous experiments @xcite have shown that , at low frequencies ( below @xmath5100 hz ) , the photothermal expansion is determined by the substrate contribution . in agreement with the theory of homogeneous mirrors , the curve can be roughly divided into two regions , separated by the characteristic angular frequency @xmath97 which depends on the thermal properties of the mirrors material .
well above @xmath97 , the surface displacement is expected to decrease as the inverse of the frequency while , well below @xmath97 , it tends to a constant value which determines the absorption coefficient . on the other hand , at higher frequencies ( above @xmath5100 hz ) , the coating contribution to the total thermal expansion , in particular through the filter function , becomes relevant and the observed behavior differs substantially from the substrate @xmath98 frequency dependence . at even higher frequencies ,
also the thermo - refractive effect becomes important ( in the following , we will use the term ` photothermal response ' for the overall effect , including the thermo - refractive component ) . an accurate data analysis through eq .
( [ xfinal ] ) and the related expressions in appendix a allows us to to measure the effective thermal properties of the ta@xmath8o@xmath77/sio@xmath8 dielectric layers .
moreover , assuming that the thermal parameters of silica films are not significantly different from those of bulk silica , we will estimate in the next section the properties of ta@xmath8o@xmath77 thin film . on this material , to our knowledge , just few , room temperature measurements have been performed . the amplitude and phase of the photothermal response at room temperature
are shown in fig .
( [ fig2 ] ) . the experimental data ( amplitude and phase ) are globally fitted through the complex response function ( [ xfinal ] ) considering both the silica and silicon substrate contributions and the coating .
the fitting function is shown in fig .
( [ fig2 ] ) ( red traces ) , and fig .
( [ fig2]a ) also shows the amplitudes of some specific contributions . in the fitting procedure ,
the substrates thermal properties ( thermal conductivity , specific heat capacity and expansion coefficient ) are fixed to values taken from the literature and reported in table i. the coating parameters derived from the fit are reported in table ii ( first column ) .
the mechanical unbalance parameters @xmath99 and @xmath100 ( see eq .
( [ b39a ] ) ) are both close to unit and depend on the young modulus ( @xmath101 ) and the poisson factor ( @xmath19 ) of substrate and coating .
we have taken the values of @xmath101 and @xmath19 from the literature ( they are reported in tables i and ii ) , including the experimental uncertainty for the coating , and a spread due to the dependence on the direction with respect to the crystal axes for si , since we have no information on their orientation in the mirror .
the calculated parameters are @xmath102 and @xmath103 for the silica mirror ; @xmath104 and @xmath105 for the silicon mirror .
the uncertainty in the mechanical parameters gives a small spread in the fitted parameters ( below the statistical error given by the fit ) , that we have included in their quoted uncertainty in table ii .
the absorption coefficients of the two mirrors are kept as different parameters even if the two coatings are very similar ( their slightly different thickness is not expected to have a great influence on the absorption ) , since they have been provided in different batches and we have no precise information on the post - coating annealing procedure that can be different for the two mirrors .
the coating parameters @xmath106 , @xmath107 , @xmath108 , and @xmath109 are determined by the fit .
its results are reported in table ii together with their uncertainties derived from the 68 % confidence intervals , slightly increased to account for the spread in the used values of young modulus and poisson factor . at frequencies below 100
hz the silica substrate contribution ( whose characteristic frequency is @xmath110 hz ) is dominant while at higher frequencies the photothermal response is dominated by the silicon substrate ( @xmath111 hz ) and the coating .
the importance of the coating is not restricted to its direct contribution to the photothermal expansion ( dash - dotted curves ) , but it is evident above all in the filter function that strongly affects the total response .
this is put into evidence in fig .
( [ fig2]b ) , where the phase data are also compared with a curve obtained form eq .
( [ xbrg ] ) , thus neglecting the filter function contribution ( dashed curve ) . in this case , the disagreement between theory and experiment is relevant even at low frequencies and would lead to unreliable estimations of both the substrate and coating parameters . the phase of the response is sensitive also to the thermo - refractive effect , which seems negligible if one just consider its amplitude .
this is underlined again in fig .
( [ fig2]b ) , where the dotted curve neglects @xmath112 in eq .
( [ xfinal ] ) . by considering the complete model ( red curve ) , instead
, the agreement is excellent in the whole frequency range .
we observe that the coating on silicon mirror displays a lower absorption .
the absorption coefficients that we have obtained are a bit higher than previously reported @xcite , yet remaining within few ppm .
we remark however than the quoted uncertainty is reliable for comparing the two mirror , but the estimate of the overall absorption is conditioned by the evaluation of the intracavity power that is hardly very accurate .
the measurements have been repeated at low temperature , with the cavity spacer at @xmath113 k. the results are shown in fig .
( [ fig3 ] ) , together with the fitting functions , and the derived parameters are reported in table ii ( second column ) . around this temperature ,
the photothermal response is dominated by the fused silica substrate at all frequencies .
we remark that , at low frequencies , the phase curves reported in fig .
( [ fig2]b ) and fig .
( [ fig3]b ) clearly show that the photothermal response at room and cryogenic temperature are @xmath114 out of phase .
this is due to the sign change in the fused silica expansion coefficient , which becomes negative below @xmath115 k. the effect of the coating filter function on the photothermal response phase is even more important than at room temperature , as evidenced in fig .
( [ fig3]b ) . for what concerns
the silicon substrate , on one hand the silicon thermal diffusion length becomes larger than the mirror size for frequencies below few khz , thus invalidating our theoretical model ( the temperature of the silicon substrate could be considered as homogeneous , instantaneously determined by the heat absorption and the conduction through the copper spacer ) ; on the other hand , its thermal expansion coefficient becomes so low that it can be completely neglected .
some figures justifying this assumption are reported in table i. concerning the thermal parameters of the silica substrate , we remark that expansion coefficient and capacitance have a strong dependence on temperature . since the real temperature of the mirrors
is also determined by their thermal contact with the copper spacer , we have kept it as free parameters in the fit , using the known temperature dependence of @xmath116 and @xmath117 ( reported in the caption of table i ) .
the derived temperature of the mirrors is indeed @xmath118 k , giving for bulk silica @xmath119 k@xmath120 and @xmath121 j / kg k. the mechanical parameters ( @xmath99 , @xmath100 , @xmath19 ) are kept at their room temperature values ( their weak temperature dependence is negligible in our work ) .
we have verified that their uncertainty has negligible effects on all the fitted parameters .
the contribution of the coating on silicon is taken into account , imposing @xmath122 in the relevant equations as justified by the silicon thermal parameters reported in table i. the ratio between the absorption coefficients of the two mirrors is fixed at the value found at room temperature ( i.e. , @xmath123 ) .
the @xmath124 uncertainty in this ratio adds a small error to some fitted parameters ( well below the corresponding statistical errors form the fit ) , that is included in the uncertainties quoted in table ii .
the absorption in the silica mirror results to be lower than at room temperature .
however , we remark again that the quoted uncertainty in the absorption coefficients only reflects the statistical error , while absolute intracavity power calibrations ( giving additional systematic uncertainty ) are different between runs at room and cryogenic temperature due to the necessary re - alignment .
the thermo - refractive effect at cryogenic temperature is too small to allow a reliable independent estimate of both @xmath108 and @xmath109 from our data .
yet , it can not be completely neglected .
therefore , we have left @xmath108 as free parameter and used the expression @xmath109 of eq .
( [ betaeff ] ) as follows : for the @xmath125 parameter ( concerning silica ) we have used the silica bulk values of @xmath18 and @xmath126 reported in table i ; the expansion coefficient of tantala is derived from eq .
( [ avexp ] ) using the free parameter @xmath108 and the bulk expansion coefficient of silica ; we have varied the thermo - optic coefficient @xmath126 of tantala between 0 and twice the silica value .
the spread in this coefficient has negligible effect in all the fitted parameters except for @xmath108 , where the spread is comparable to the statistical uncertainty of the fit . since our assumption on the range of the tantala thermo - optic coefficient is rather arbitrary , we have quoted separately its effect on @xmath108 in table ii to allow a reliable evaluation of our results once an independent measurement of @xmath126 for tantala is provided . as discussed before ,
the fitting procedure allows the estimate of the coating effective thermal conductivity @xmath107 , the product @xmath127 , the averaged thermal expansion coefficient @xmath108 and the thermo - refractive coefficient @xmath109 .
the values of these effective parameters are the main result of our work , and they can be directly used in most evaluations of noise sources and , in general , physical phenomena related to the high reflectivity coating in interferometers . in this section , we further discuss our results with the aim of extracting the parameters of the materials composing the coating , and comparing with existing literature .
the models used in our data analysis consider indeed an homogeneous coating layer . in a complete description of the film behavior
, the real parameters of the layers would enter with several different combinations that should be individually averaged over the coating .
however , to extract some estimates of the layers parameters , we adopt the simplified description suggested , e.g. , in ref .
@xcite , that takes volume averages such as @xmath128 @xmath129 @xmath130 @xmath131 where the subscripts ` @xmath132 ' and ` @xmath133 ' refer to the two consecutive layers composing one period of the coating .
( [ parametrikeff1]-[parametrikeff2 ] ) and ( [ betaeff ] ) allow us to extract the parameters of the materials composing the film from our experimental results .
we start our discussion from the room temperature data . ref .
@xcite presents a review of conductivity measurements in sio@xmath8 and ta@xmath8o@xmath77 films , for different thickness and deposition techniques .
the range of results is very wide , extending between 0.017 and 0.73 w / mk for silica , and between 0.026 and 15 w / mk for ta@xmath8o@xmath77 .
if we assume , in spite of that , for the sio@xmath8 layers the bulk parameter values reported in table i , from eq .
( [ parametrikeff1 ] ) we obtain for ta@xmath8o@xmath77 a thermal conductivity of @xmath134=0.4 w / mk . we stress that such estimate should be taken _
cum grano salis _ , since conductivity in thin films can strongly depend on the film structure ; the most meaningful parameter remains @xmath107 .
concerning the heat capacity , it can be estimated using the bulk parameters for silica , the measured density of 7200 kg / m@xmath135 for the ta@xmath8o@xmath77 film @xcite , and the bulk tantala capacity of 306 j / kg k @xcite .
we calculate from eq .
( [ parametrikeff1a ] ) a value of @xmath136 j / m@xmath135 k. the result of our fit is @xmath137 j / m@xmath135 k , that is compatible ( within two standard errors ) , while a bit lower , with the previous estimate .
the thermal expansion coefficient of ta@xmath8o@xmath77 can be extracted through eq .
( [ avexp ] ) , using again the bulk data for silica .
we obtain @xmath138=4.8@xmath139 @xmath140 to be compared with the values of -4.4 @xmath141 @xmath140 @xcite , 2.42 @xmath142 @xmath140 @xcite and 5 @xmath142 @xmath140 @xcite measured in previous experiments .
@xcite also report a value of 3.6 @xmath142 @xmath140 obtained from the analysis of data in the literature .
our value is closer to those reported in refs .
@xcite than to the one obtained by inci @xcite which , apart the unexpected negative sign , is more than a factor of 10 larger in absolute value . as discussed in ref .
@xcite , the variability between these independent measurements could be due to the fact that the properties of tantalum pentoxide layers might depend on the procedure of the layer deposition .
an accurate measurements of the refractive index of bulk silica at temperature varying from 30 to 310 k is reported in ref .
@xcite . from its data , we infer ( at @xmath143 nm ) @xmath144 k@xmath120 at 300 k , and @xmath145 k@xmath120 at low temperature .
the room temperature value is compatible with @xmath146 k@xmath120 measured at 632.8 nm for a sputtered sio@xmath147 film in ref .
using eq .
( [ betaeff ] ) and the bulk data for silica , we obtain for the tantalum pentoxide @xmath148 k@xmath120 , and from our value of @xmath138 we calculate ( @xmath149 k@xmath120 . previous measurements of ( @xmath150 , performed at 632.8 nm by a group in taiwan , gives 2.3@xmath151 k@xmath120 @xcite and , later , 3.64@xmath151 k@xmath120 and 7.89@xmath151 k@xmath120 @xcite ( depending on the thermal stress ) .
a further value of 14@xmath151 k@xmath120 is mentioned in @xcite as a private communication . at low temperature ,
the literature on the coating properties is very poor . by using , as before , the expressions ( [ parametrikeff1])-([parametrikeff2 ] ) together with the bulk properties of table i for silica ,
we deduce the following data for the film of tantalum pentoxide : the expansion coefficient is @xmath138=0.58@xmath152 @xmath140 ( the first quoted uncertainty is due to the spread in the assumed thermo - optic parameter of tantala between 0 and 2 @xmath153 @xmath154 , with @xmath155 at 10.9 k ; the second quoted uncertainty is the statistical error from the fit ) ; the thermal capacitance is @xmath156 j / k m@xmath135 and , using a density of 7200 kg / m@xmath135 , we calculate @xmath157 j / kg k ; the thermal conductivity is @xmath158 w / mk .
we have measured the variation of the effective reflecting surface in a high reflectivity mirror , caused by a modulation in the residual absorbed power .
such complex response is due to photothermal and thermo - refractive effects in the mirror substrate and coating , and our measurements allow a characterization of these effects both at room and at cryogenic temperature .
an accurate knowledge of such phenomena is particularly relevant in experiments aiming to the highest sensitivity in displacement , position and force measurements , for instance in the fields of gravitational wave detection and in quantum optics .
most of the planned future experiments of this kind will address cryogenics to reduce brownian noise , and part of the present work is indeed the focused on cryogenic temperature , where measurements of the kind here presented were still missing .
our experimental results are well described using a theoretical model that we have expressly developed , and the comparison allows us to extract meaningful effective thermo - mechanical parameters of the high reflectivity coating .
this is the most important aspect of our work , since a ) the overall coating behavior is not obviously deduced from the properties of the composing materials ( in particular concerning thermal expansion and conductivity ) ; b ) the parameters that we have measured can be used to calculate weak but important noise contributions such as thermodynamic fluctuations or , of course , photothermal and thermo - refractive noise .
we underline that , while specific experiments on the different oxides in the bulk or thin - film form can give more accurate results concerning the material properties , our setup has the added value of characterizing the mirrors ( and coating ) as they are used in high sensitivity experiments , thus describing their real overall behavior .
moreover , from the effective coating parameters we could still estimate some properties of the dielectric films composing the coating . in particular , literature on previous measurements were rather poor for tantalum pentoxide at room temperature , and , in our knowledge , completely lacking at cryogenic temperature .
this work was partially funded by the european union ilias project ( no .
rii3-ct-2003 - 506222 ) .
in this appendix we calculate the temperature profile and the elastic deformation of an half - infinite mirror extended in the half - space @xmath30 , composed of a layer of thickness @xmath41 an an homogeneous substrate , with surface absorption of a gaussian beam .
we will consider an isotropic substrate , neglecting the directional variation of the si properties .
the calculation is performed using fourier transforms for the temporal and transverse spatial variables , while keeping the dependence on the longitudinal variable @xmath159 .
we start by the study of the temperature distribution .
we define @xmath160 and @xmath161 where @xmath162 , @xmath163 , and @xmath164 .
the equation for the thermal propagation @xmath165 can be written in the time - transformed fourier space as @xmath166 where @xmath34 is defined in eq .
( [ k ] ) , and in the space- and time - transformed fourier space as @xmath167 the absorbed intensity , in the time - transformed fourier space , is written as @xmath168 and , in the complete fourier description where @xmath169 is the spatial fourier transformed of @xmath170 , it becomes @xmath171 the three - dimensional model is now formally equivalent to the one - dimensional model described in section [ model ] .
the equations ( [ b4 ] ) is equal to the homogeneous equation associated to eq .
( [ thdiff ] ) , and the boundary conditions ( [ bound1 ] ) and ( [ bound2 ] ) are still valid , with the following list of substitutions : @xmath172 @xmath173 @xmath174 the discussion on the temperature distribution in the one - dimensional model remains valid too .
we can therefore use eqs .
( [ exp]-[filtro ] ) , with the replacements ( [ b7]-[b9 ] ) and where @xmath43 and @xmath42 contain the respective thermal parameters . for the sake of clarity , we repeat here explicitly the solutions , in the layer ( from eq .
( [ tlayer ] ) ) : @xmath175+\mathcal{r}\cosh\left[\sqrt{k_c^2+k_{\perp}^2}\ , ( d - z)\right]}{\cosh\left[\sqrt{k_c^2+k_{\perp}^2}\ , d\right]+\mathcal{r}\sinh\left[\sqrt{k_c^2+k_{\perp}^2}\ , d\right]}\,\delta p_{abs}(\omega ) \label{b9a}\ ] ] and in the substrate ( from eqs .
( [ exp1],[bs ] ) ) : @xmath176}{\cosh\left[\sqrt{k_c^2+k_{\perp}^2}\ , d\right]+\mathcal{r}\sinh\left[\sqrt{k_c^2+k_{\perp}^2}\ , d\right]}\,\delta p_{abs}(\omega ) \label{b9b}\ ] ] where the reflection coefficient is now @xmath177 we now consider the problem of the mirror thermal expansion .
the elastic deformation @xmath178 is ruled by the static elastic equation @xcite @xmath179 the relevant components of the stress tensor are defined as @xmath180 \label{b10a}\ ] ] @xmath181 @xmath182 where @xmath101 is the young modulus and @xmath19 is the poisson ratio .
following bgv , we search a particular solution of eq .
( [ b10 ] ) in the form of a gradient of a potential @xmath183 .
inserting @xmath184 in eq .
( [ b10 ] ) , we see that the potential @xmath183 obeys the poisson s equation @xmath185 as in the study of the thermal diffusion , we define the time- and transverse spatial fourier transformed variables @xmath189 of @xmath183 ; @xmath190 and @xmath191 of the cartesian components of @xmath178 and @xmath187 , respectively , with @xmath192 .
the equation ( [ b16 ] ) , in its cartesian components , can be written in the fourier space , giving @xmath193 from the @xmath194 component , @xmath195 from the @xmath196 component , and the equivalent of eq .
( [ b19 ] ) with exchange of @xmath197 and @xmath198 for the @xmath199 component .
the axial symmetry of the problem implies that @xmath200 .
we can therefore define @xmath201 that , replaced in eq .
( [ b18 ] ) , gives @xmath202 and replaced in either eq .
( [ b19 ] ) or its equivalent with @xmath203 ( the two equations transforms into the same expression ) gives @xmath204 we remark that the condition @xmath200 simplifies very much the calculation .
however , one can also avoid to impose it at the beginning , and verify later that it results as a consequence of the boundary conditions .
( [ b21 ] ) and ( [ b22 ] ) form a system of homogeneous , second order differential equations with constant coefficients , in the variable @xmath159 , for the vector of functions @xmath205 .
the solution is a linear combinations of the four vectors @xmath206 ; @xmath207 ; @xmath208 ; @xmath209 .
it is useful to express the solution in different forms for the layer ( in @xmath210 ) and in the substrate ( for @xmath211 ) , where we skip the diverging solutions with positive real part of the exponent .
therefore , we write the solutions as @xmath212 @xmath213 \nonumber\\ + v_{ss}\,\big[(4\sigma_c-3)\,\sinh ( k_{\perp } z)\,-\,k_{\perp}\,z\,\cosh ( k_{\perp } z)\big ] \label{b24}\end{aligned}\ ] ] in the layer and @xmath214 @xmath215\,e^{-\,k_{\perp } ( z - d ) } \label{b26}\ ] ] in the substrate . here
@xmath216 and @xmath217 are the poisson factors respectively in the layer and in the substrate , and @xmath218 are functions of @xmath219 and @xmath220 .
the definitions of the stress tensor given in eqs .
( [ b10a]-[b10c ] ) can also be written in the complete fourier space . naming @xmath222
the space- and time - fourier transformed of @xmath223 , using eqs .
( [ b15 ] ) and ( [ b27 ] ) , we obtain @xmath224 \label{b28}\ ] ] @xmath225 \qquad .
\label{b29}\ ] ] once written a complete solution of the elastic equation , we now consider the boundary conditions . at @xmath36
the null stress condition is @xmath226 @xmath227 the continuity conditions for the stress at the interface between layer and substrate are @xmath228 @xmath229 the continuity of the displacement at the interface is given by the relation @xmath230 , that can be transformed , using eqs .
( [ b15 ] ) and ( [ b27 ] ) , into @xmath231 @xmath232 replacing the solutions ( [ b9a]-[b9b ] ) and ( [ b23]-[b26 ] ) into the eqs . ( [ b30]-[b35 ] ) we obtain a system of 6 linear equations for the 6 variables @xmath218 , that can be solved using annoying but straightforward algebra . as we will see , for our purpose the meaningful function is just @xmath233 .
the average displacement measured by the laser beam is given in eq .
( [ ut ] ) . by using the inverse fourier
transform for @xmath234 we can write @xmath235 and performing the integration over @xmath236 and over the angular coordinate of @xmath237 , we can express the time - fourier transformed of such displacement as @xmath238 using eqs .
( [ b15 ] ) , ( [ b23 ] ) and ( [ b27 ] ) the function @xmath239 in the integrand of eq . ( [ b37 ] )
is transformed into @xmath240 the expression of @xmath241 ( and , in particular , of @xmath233 ) is completely and explicitly determined by the algebra described above .
it can be shown that @xmath241 can be written in the following form @xmath242 \nonumber\\ + \,\frac{2\alpha_s}{k_s^2 } \,&\gamma_3(k_{\perp})&\ , \bigg[\frac{\partial\theta}{\partial z}\bigg|_{z = d^+}+k_{\perp}\,\theta\big|_{z = d}\bigg ] \qquad . \label{b39}\end{aligned}\ ] ] the particular combinations of the temperature functions that appear in eq .
( [ b39 ] ) assure the correct frequency dependence for @xmath243 , compensating the divergence given by @xmath244 .
therefore , they must be kept as they are . on the other hand ,
the functions @xmath245 , @xmath246 and @xmath247 are rather complex for a practical use .
it is useful to give their approximations at the lowest order in @xmath248 , that read @xmath249 \nonumber\\ \gamma_2\,=\,\gamma_2(0 ) & = & \frac{1-\sigma_s^2}{1-\sigma_c}\frac{e_c}{e_s } \nonumber\\ \gamma_3(0 ) & = & 1+\sigma_s \qquad .
\label{b39a}\end{aligned}\ ] ] these three parameters are close to unit , to which they reduce if @xmath250 and @xmath251 . when inserting @xmath241 in eq .
( [ b37 ] ) , the approximated expressions of the @xmath252 s can be used , provided that @xmath253 , thanks to the fast decaying gaussian function of @xmath254 .
replacing @xmath255 from eqs .
( [ b9a]-[b9b ] ) in eq .
( [ b39 ] ) , inserting the resulting @xmath241 in eq .
( [ b37 ] ) , and after some simplifications , we can write the expected photothermal response to the modulated absorbed power as the sum of two terms , equivalent in the simplified model to the expansions of coating and substrate : @xmath256 \label{b41}\end{aligned}\ ] ] and @xmath257\,\xi\,e^{-\frac{\xi^2}{2}}\,\delta p_{abs}(\omega ) \label{b42}\ ] ] where the adimensional integration variable is @xmath258 , and we have defined to simplify the notation @xmath259 the filter function @xmath260 , derived from eq .
( [ filtro ] ) by replacing @xmath261 , can be written as @xmath262+\mathcal{r}\sinh[\xi_c\,d / r_0 ] } \qquad . \label{b46}\ ] ] in the limit @xmath263 ( no coating layer ) the filter @xmath260 obviously assumes a unit value , and @xmath264 coincides with the results of refs @xcite and @xcite ( reported here in eq .
( [ xbrg ] ) ) . in other words , while not evident at a first glance , it can be shown that @xmath265\,\xi\,e^{-\frac{\xi^2}{2}}\,=\,g\left(\omega\right)\ ] ] where @xmath26 is given in eq .
( [ g ] ) .
concerning the expansion of the coating , it can be shown that @xmath266 in the high frequency limit reduces to the one - dimensional expression ( first term on the right side of eq .
( [ x1dim ] ) ) , if one set @xmath267 .
one further effect to be considered is the thermo - refractive displacement . from eq .
( [ uzthr ] ) , using eq .
( [ b37 ] ) for the average displacement , we can write it as @xmath268 and replacing @xmath269 from eq .
( [ b9a ] ) @xmath270 finally , the total response to the modulated absorbed power is @xmath271 for an overview of the present status and foreseen developments , see the special issue with the proceedings of the 8th eduardo amaldi conference on gravitational waves : class .
quantum grav .
* 27 * , 084001 - 084036 ( 2010 ) .
m. cerdonio _
lett . * 87 * , 031101 ( 2001 ) .
t. caniard , p. verlot , t. briant , p .- f .
cohadon , and a. heidmann , phys .
lett . * 99 * , 110801 ( 2007 ) .
v. b. braginsky , m. l. gorodetsky , and s.p .
vyatchanin , phys .
a * 264 * , 1 ( 1999 ) .
m. cerdonio , l. conti , a. heidmann and m. pinard , phys .
d * 63 * , 082003 ( 2001 ) .
k. an , a. sones , c. fang - yen , r. r. dasari , and m. s. feld , opt . lett . * 22 * , 1433 ( 1997 ) .
m. de rosa , l. conti , m. cerdonio , m. pinard , f. marin , phys .
89 * , 237402 ( 2002 ) .
m. de rosa , f. marin , f. marino , o. arcizet , a. heidmann , and m. pinard , class . quantum grav .
* 23 * , s259 ( 2006 ) .
e. d. black , i. s. grudinin , s. r. rao and k. g. libbrecht , j. appl . phys .
* 95 * , 7655 ( 2004 ) .
t. uchiyama _
et al . _ ,
a * 242 * , 211 ( 1998 ) ; _ ibidem _ * 261 * , 5 ( 1999 ) . m. m. fejer _
et al . _ ,
d * 70 * , 082003 ( 2004 ) .
m. born and e. wolf , ` principle of optics ' , pergamon press , oxford - new york - seoul - tokyo 1980 .
v. b. braginsky , m. l. gorodetsky , s.p .
vyatchanin , phys .
a * 271 * , 303 ( 2000 ) .
m. l. gorodetsky , phys .
a * 372 * , 6813 ( 2008 ) .
et al . _ ,
d * 78 * , 102003 ( 2008 ) .
corning inc .
corning 7980 data sheet .
http://www.corning.com/ r.w.p .
drever _ et al .
b * 31 * , 97 ( 1983 ) .
r. flaminio _ et al .
_ , class . quantum grav .
* 27 * , 084030 ( 2010 ) .
k. k. kelley , j. am .
soc . * 62 * , 818 ( 1940 ) .
m. l. grilli , d. ristau , m. dieckmann , u. willamowski , appl .
a * 71 * , 71 ( 2000 ) .
n. inci , _ simultaneous measurements of thermal optical and linear thermal expansion coefficients of @xmath272 films _ , ico xix , august 2002 , firenze , italy , spie proceedings vol . 4829 , p. 755
tien , c .- c .
jaing , c .- c .
lee , and k .-
chuang , j. mod . opt . * 47 * 1681 ( 2000 ) . v.
b. braginsky , a. a. samoilenko , phys .
a * 315 * , 175 ( 2003 ) .
d. b. leviton and b. j. frey , _ temperature - dependent absolute refractive index measurements of synthetic fused silica _ , nasa goddard space flight center , document i d 20070018851 ( 2006 ) .
a. k. chu , h. c. lin , and w. h. cheng , j. electro .
* 26 * , 889 ( 1997 ) .
w. h. cheng , s. f. chi , a. k. chu , thin solid films * 347 * , 233 ( 1999 ) .
r. c. zeller , r. o. pohl , phys .
b * 4 * , 2029 ( 1971 ) .
g. x. mack , a. c. anderson , phys .
b * 31 * , 1102 ( 1985 ) .
g. k. white , phys .
lett . * 34 * , 204 ( 1975 ) .
k. g. lyon , g. l. salinger , c. a. swenson , phys .
b * 19 * , 4231 ( 1979 ) .
d. a. ackerman , a. c. anderson , phys .
49 * , 1176 ( 1982 ) . c. j. glassbrenner , g. a. slack , phys
. rev . * 134 * , a1058 ( 1964 ) .
h. r. shanks , p. d. maycock , p. h. sidles , g. c. danielson , phys . rev . *
130 * , 1743 ( 1963 ) .
y. okada , y. tokumaru , j. appl .
phys . * 56 * , 314 ( 1984 ) .
p. flubacher , a. j. leadbetter , j. a. morrison , phil . mag . * 4 * , 273 ( 1959 ) .
k. g. lyon , g. l. salinger , c. a. swenson , g. k. white , j. appl .
phys . * 48 * , 865 ( 1977 ) .
the authors report the relation @xmath273 taken from their reference : g. k. white , aip conf .
* 17 * , 1 ( 1974 ) .
j. j. wortman and r. a. evans , j. appl
. phys . * 36 * , 153 ( 1965 ) .
d. r. m. crooks _ et al .
_ , class . quantum grav .
* 23 * , 4953 ( 200655 ) .
.substrate parameters used in the fitting procedure .
the expressions for the low - temperature thermal capacitance and expansion coefficient of silica are respectively @xmath274 and @xmath275 .
the poisson factors for si are averaged over the respective planes . [ cols="<,^,^,^,^",options="header " , ] ) for both silica and silicon mirrors , keeping fixed the substrate parameters ( see table i ) .
other curves refer to the fused silica ( green ) and silicon ( blue ) mirrors .
dashed lines : substrate contributions ; dash - dotted lines : coating expansion contributions ; dotted lines : thermo - refractive contributions .
b ) phase of the photothermal response and its corresponding fit .
blue curves are traced neglecting some effects , in order to show their relevance : the dashed line corresponds to a response of just homogeneous substrates ( calculated from eq .
( [ xbrg ] ) , the dotted line considers just @xmath264 and @xmath266 in eq .
( [ xfinal ] ) , thus neglecting the thermo - refractive effect . ] | increasing requirements in the sensitivity of interferometric measurements is a common feature of several research fields , from gravitational wave detection to quantum optics .
this motivates refined studies of high reflectivity mirrors and of noise sources that are tightly related to their structure . in this work we present an experimental characterization of photothermal and thermo - refractive effects in high reflectivity mirrors , i.e. , of the variations in the position of their effective reflection plane due to weak residual power absorption .
the measurements are performed by modulating the impinging power in the range 10 hz @xmath0 100 khz .
the experimental results are compared with an expressly derived theoretical model in order to fully understand the phenomena and exploit them to extract useful effective thermo - mechanical parameters of the coating .
the measurements are extended at cryogenic temperature , where most high sensitivity experiments are performed ( or planned in future versions ) and where characterizations of dielectric film coatings are still poor . |
determining the volume of metric balls in riemannian manifold , in particular unitary group , is the key to understand several coding and information theoretical quantities .
performance analysis of unitary space - time codes @xcite requires the knowledge of volume in the unitary group @xcite . for channel quantizations in precoded multi - antenna systems , the characterization of rate - distortion tradeoff
is directly related to volume calculations in the manifold of interest @xcite .
estimating fundamental coding bounds such as gilbert - varshamov and hamming bounds relies on the volume of the corresponding metric ball @xcite . despite the need to accurately estimate the volume of metric balls in the unitary group , results in this direction are rather limited .
volume estimates have been derived in @xcite when the radius of metric ball is small . in this paper
, we study the volume of a metric ball , valid for any radius , in the unitary group with chordal distance . the same problem was considered in @xcite in the study of diversity sum bounds of unitary space - time codes , where the authors relied on numerical integrations to evaluate the volume .
the starting point of the current paper is that such a numerical step may not be necessary .
specifically , we show that the exact volume boils down to an integral involving a toeplitz determinant with bessel function entries .
this representation gives rise to an explicit formula in the simplest case of a two - by - two unitary matrix .
we then present a simple asymptotic volume formula , which is the main technical contribution of the paper .
two distinct paths that led to this asymptotically exact formula have been identified : one is based on the connection of the toeplitz determinant to a hypergeometric function of matrix argument , the other directly invokes szeg s strong limit theorem on toeplitz determinants .
these powerful tools of random matrix theory , albeit being subjects of intense studies for several decades , have not been fully utilized in the coding theory community .
surprisingly , the limiting volume formula is already quite accurate for dimension as small as three .
the reason behind the rapid convergence is examined , where it is found that the asymptotic formula approaches its limit super - exponentially fast as dimension increases due to a discovered mock - gaussian property . as an application
, we study some basic coding - theoretic questions of unitary codes via the derived asymptotic volume formula . in particular , analytical expressions for the gilbert - varshamov and the hamming bounds on codeword cardinality as well as the corresponding bounds on minimum distance of a code are derived .
we also derive formulas for bounds on the code rate and its scaling law , which are key to establish existence results of unitary codes .
in addition , a closed - form upper bound on diversity sum of unitary space - time codes is obtained .
these simple - to - compute analytical results capture the behavior of the bounds reasonably accurately , which also lead to useful insights and properties of the bounds .
the rest of the paper is organized as follows . in section [ sec : problem ] we formulate the problem considered in this paper , where we first define volume of a metric ball in unitary group before stating the coding bounds of interests . section [ sec : vol ] is devoted to the derivation of exact and asymptotic volume formulas . in section
[ sec : app ] the derived analytical results are utilized in the study of coding - theoretic bounds of unitary codes . we conclude the main findings of this paper in section [ sec : con ] . proofs of some the technical results are provided in the appendices .
consider a metric ball around the identity element @xmath0 in the @xmath1-dimensional unitary group @xmath2 with euclidean distance ( chordal distance ) @xmath3 , @xmath4 where @xmath5 is the frobenius norm topological metric .
we consider the invariant haar measure @xmath6 , defining a uniform distribution on @xmath2 . for any measurable set @xmath7 and any @xmath8 , the haar measure satisfies @xmath9 . due to the homogeneity of @xmath2 ,
the characteristics of the ball ( [ eq : ball ] ) centered at any group element , say @xmath10 , would be the same .
the eigenvalue decomposition of @xmath11 takes the form @xmath12 where @xmath13 and the diagonal entries of @xmath14 are @xmath1 complex numbers @xmath15 on the unit circle .
more precisely , @xmath16 takes values in a flag manifold of equivalence classes of @xmath2 modulo diagonal matrices of the form @xmath14 .
the joint density of the angles @xmath17 is given by @xcite @xmath18 where @xmath19 , @xmath20 and the constant @xmath21 .
the product @xmath22 is the jacobian of the transform ( [ eq : evd ] ) of haar measure @xmath23 to the measure @xmath24 . in the random matrix theory literature ,
the joint density ( [ eq : j ] ) is referred to as circular unitary ensemble @xcite and @xmath25 is the corresponding partition function .
the condition on the distance measure @xmath26 in ( [ eq : ball ] ) is equivalent to @xmath27 .
thus , the ( normalized ) volume of the metric ball ( [ eq : ball ] ) equals the following @xmath1-dimensional integral @xcite @xmath28 where @xmath29 . for the maximal distance @xmath30 , the restriction @xmath27 becomes irrelevant and @xmath31 by the definition ( [ eq : j ] ) . for this reason
, we also refer ( [ eq : v ] ) as a restricted partition function of the circular unitary ensemble .
finally , we note that the volume of metric balls with other distance measures such as the geodesic distance could be similarly obtained by the analytical framework developed in section [ sec : vol ] .
a fundamental problem in coding theory is to study the maximum size of codewords as a function of the minimum distance of a code . a unitary code @xmath32 with cardinality @xmath33 , @xmath34 is a finite subset of unitary group @xmath2 . with a slight abuse of notation
, we define @xmath35 as the minimum distance between distinct codewords in @xmath2 .
bounds on codeword cardinality in unitary group rely on the normalized volume of the metric ball @xmath36 formulated in ( [ eq : v ] ) .
in particular , the gilbert - varshamov lower bound and the hamming upper bound on the cardinality @xmath33 are related to volumes of metric ball as @xcite @xmath37 the gilbert - varshamov bound and the hamming bound are referred to as a sphere covering bound and a sphere packing bound , respectively .
the principles of the two coding bounds are different .
the gilbert - varshamov bound ( [ eq : bounds ] ) is based on a greedy but natural approach to construct a unitary code : start with any codeword and keep on adding codewords that have distance at least @xmath3 from all codewords already included until the whole space is covered .
such an algorithm will terminate at @xmath38 , where @xmath39 .
the argument @xmath40 in the hamming bound is related to the error correcting capability of the code , where errors made within ` hamming sphere ' of radius @xmath40 can be corrected .
this generalizes the concept of hamming distance of q - ary block codes to unitary codes . by taking the union of disjoint metric balls each with @xmath40 packing radius ,
we arrive at the hamming bound ( [ eq : bounds ] ) . another important quantity in coding theory , especially in the limit @xmath41 ,
is rate of the code @xcite @xmath42 the bounds on cardinality ( [ eq : bounds ] ) immediately lead to bounds on code rate @xmath43 namely , for any @xmath1 there exists a code in @xmath2 with a minimum distance @xmath3 and code rate as bounded in ( [ eq : rate ] ) .
the corresponding bounds on minimum distance @xmath3 as a function of code rate @xmath44 is obtained as @xmath45 where @xmath46 denotes inverse function of the volume ( [ eq : v ] ) .
note that besides the frobenius norm other metrics such as the spectral norm can be equally considered in the definition of minimum distance ( [ eq : mddef ] ) , and the corresponding bounds in ( [ eq : bounds ] ) , ( [ eq : rate ] ) and ( [ eq : mindist ] ) are still valid .
finally , we define the diversity sum of a unitary code @xcite , which is closely related to the minimum distance ( [ eq : mddef ] ) , as @xmath47 the diversity sum ( [ eq : diverdef ] ) is an important performance measure for unitary space - time codes , where a code with large diversity sum tends to perform well at the most critical regime of low signal - to - noise ratio @xcite .
therefore , it is interesting to know the largest possible value of diversity sum for a given dimension @xmath1 and cardinality @xmath33 .
a useful hamming - type upper bound to @xmath48 is given by ( * ? ? ?
* eq . ( b1 ) ) @xmath49 which is tight for large @xmath33 .
note that the other diversity sum bounds ( b2 ) and ( b3 ) proposed in @xcite as well as the linear programming bounds @xcite can be also analyzed by the framework developed in section [ sec : vol ] .
finally , from the diversity sum bound ( [ eq : diversity ] ) and the relation ( [ eq : diverdef ] ) , we reach another upper bound on minimum distance @xmath50 as discussed in @xcite , this modified bound is tighter than the minimum distance upper bound ( [ eq : mindist ] ) .
as we have seen , the performance of coding - theoretic bounds ( [ eq : bounds ] ) , ( [ eq : rate ] ) , ( [ eq : mindist ] ) , ( [ eq : diversity ] ) and ( [ eq : mindist2 ] ) all rely on an accurate estimate to volume of the corresponding metric ball . in the following ,
we derive an exact ( [ eq : r1 ] ) as well as an asymptotic ( [ eq : r2 ] ) volume formulas using diverse recipes from random matrix theory .
readers who are more interested in the applications to coding theory may skip the technical details in this section .
we start by rewriting the @xmath1-dimensional integral ( [ eq : v ] ) , with the help of a dirac delta function @xmath51 , as @xmath52 inserting the fourier representation of dirac delta function @xmath53 into the reformulation ( [ eq : vde ] ) and performing the integration over @xmath54 first , we have @xmath55 where @xmath56 comparing ( [ eq : v ] ) with ( [ eq : bir ] ) and ( [ eq : mdi ] ) , we see that the reformulation amounts to eliminating the restriction @xmath27 at the expense of introducing a deformation @xmath57 in the integrand . up to now , the idea of derivation is similar to @xcite . to proceed further
, we notice that the term @xmath58 is a product of two vandermonde determinants , where @xmath59 . invoking andrief s identity ( see , appendix [ a : andreief ] ) , we have @xmath60 applying the euler s formula , the integral inside the above @xmath61 determinant equals the sum of the following four integrals @xmath62 where @xmath63 denotes the bessel function of the first kind @xcite .
we now have @xmath64 and by the definition ( [ eq : bf ] ) the toeplitz determinant ( [ eq : bi ] ) has the property @xmath65 inserting ( [ eq : bi ] ) into ( [ eq : bir ] ) , after some manipulations , we arrive at a one - dimensional integral representation for the volume of metric ball ( [ eq : v ] ) , @xmath66 in the simplest case @xmath67 , the integral ( [ eq : r1 ] ) can be explicitly calculated as @xmath68 where @xmath69 and @xmath70 . here
@xmath71 denotes meijer s g - function @xcite ( see appendix [ a : meijer ] for a brief introduction of this special function ) .
the proof of ( [ eq : n2 ] ) can be found in appendix [ a : n2 ] .
it seems difficult to explicitly calculate the integral ( [ eq : r1 ] ) for any @xmath1 .
in addition , from an information theoretical point of view , asymptotics of the bounds ( [ eq : bounds ] ) , ( [ eq : rate ] ) , ( [ eq : mindist ] ) and ( [ eq : diversity ] ) are more interesting .
these facts motivate us to study the asymptotical volume as @xmath1 goes to infinity . from the representation ( [ eq : bir ] ) , this boils down to the study of the limiting behavior of @xmath72 , which is found to be @xmath73 in the following , we will give two distinct proofs of the above strikingly simple limiting formula . the first proof
is based on matrix - variate hypergeometric functions and the associated zonal polynomials , and the second one utilizes szeg s strong limit theorem on asymptotics of toeplitz determinants .
interestingly , while the first approach seems to be more instructive and general , it appears to have received far less attention in literature . : exact ( [ eq : bi ] ) versus asymptotic ( [ eq : a]).,width=432 ] whilst being simple
, the convergence of @xmath72 to the asymptotic limit ( [ eq : a ] ) is quite fast as observed in figure [ fig : p1 ] .
as @xmath1 increases the exact curves oscillate closer to the asymptotic curve @xmath74 , and the difference is already indistinguishable for @xmath75 . note that due to the symmetry ( [ eq : tdp ] ) , we plot only positive @xmath76 in figure [ fig : p1 ] .
the equality denotes the real part of a complex variable . ]
@xmath77 allows one to rewrite the @xmath1-dimensional integral @xmath72 in ( [ eq : mdi ] ) as @xmath78 where we utilized the fact that @xmath22 is the jacobian of the transform ( [ eq : evd ] ) from @xmath23 to @xmath24 . here ,
the haar measure ( @xmath79 ) is normalized to make the total measure unity .
the integral ( [ eq : mbi ] ) over the unitary group can be represented as a hypergeometric function of matrix argument ( * ? ? ?
* eq . ( 91 ) ) as @xmath80 the function @xmath81 is referred to as bessel function of matrix argument @xcite .
it generalizes the classical univariate bessel function expressed as an integral over the unit circle , cf .
the integral inside the determinant ( [ eq : mdii ] ) .
the general form of matrix - variate hypergeometric function of an @xmath61 hermitian matrix @xmath10 is represented as @xcite @xmath82 where the sum over @xmath83 is a sum over all partitions of integer @xmath84 into no more than @xmath1 parts , i.e. @xmath85 with @xmath86 , and @xmath87 is the multivariate hypergeometric coefficient ( * ? ? ?
* eq . ( 84 ) ) . in ( [ eq : mhf ] ) , @xmath88 is a zonal polynomial @xcite , which is a homogenous symmetric polynomial of degree @xmath84 in the @xmath1 eigenvalues of @xmath10 . denoting the @xmath89-th eigenvalue of @xmath10 by @xmath90
, the zonal polynomial can be represented as ( * ? ? ?
( 85 ) ) @xmath91 where @xmath92 and @xmath93 is a schur polynomial .
schur polynomials form a basis in the space of homogeneous symmetric polynomials in @xmath1 variables of degree @xmath84 for all @xmath94 .
in particular , it holds ( * ? ? ? * eq . ( 17 ) ) @xmath95 in terms of the above notations , we can write @xmath72 as @xmath96 where the last equality is established by ( [ eq : schur ] ) . since the leading order term in @xmath97 equals @xmath98 , by the definition ( [ eq : mhc ] ) , for large @xmath1 we have @xmath99 using ( [ eq : tf ] ) , the sum in ( [ eq : dnhs ] ) for large @xmath1 becomes @xmath100 and we arrive at the claimed result ( [ eq : a ] ) , @xmath101 for a detailed account of the asymptotics of matrix variate hypergeometric functions , we refer to @xcite .
we provide here an interpretation of the obtained result . by the identity ( [ eq : lsr ] ) , @xmath72 in ( [ eq : mbi ] ) can be understood as the characteristic function of the random variable @xmath102 for any function @xmath103 , random variables of the form @xmath104 are called linear ( spectral ) statistics in random matrix theory . comparing the limiting expression of @xmath72 in ( [ eq : a ] ) with the characteristic function of a gaussian random variable @xmath105
, @xmath106 one observes that the linear statistics ( [ eq : x ] ) follows a gaussian distribution @xmath107 with mean @xmath108 and variance @xmath109 as @xmath41 .
thus , the limiting formula ( [ eq : a ] ) can be interpreted as a central limit theorem of the linear statistics ( [ eq : x ] ) .
note that this central limit theorem could be also established by a group - theoretic approach @xcite , which did neither use matrix - variate hypergeometric functions nor zonal polynomials . before going to the second proof ,
we examine the reason behind the fast convergence of @xmath72 to @xmath110 as @xmath41 . for @xmath111 , by repeated use of l hpital s rule , the schur function ( [ eq : schur ] ) is simplified to @xcite @xmath112 as a result , the sum in ( [ eq : dnhs ] ) equals @xmath113 for @xmath94 , by the orthogonality relation , see e.g. ( * ? ? ?
* chap . 1 .
4 ) or @xcite , @xmath114 we obtain @xmath115 inserting the above into ( [ eq : dnhs ] ) , we see that the first @xmath1 terms in the power series of @xmath72 equal the corresponding first @xmath1 terms of @xmath116 the @xmath41 limit of @xmath72 .
this implies that for any @xmath1 the first @xmath117 moments of the random variable ( [ eq : x ] ) are exactly the same as the limiting gaussian random variable .
namely , even for @xmath67 the first five moments are precisely the same as its gaussian limit .
we refer to this fact as finite size mock - gaussianity .
the mock - gaussian behavior may shed some light on the rapid convergence of @xmath72 seen in figure [ fig : p1 ] . for @xmath67 , @xmath118 , and @xmath75 ,
the formal power series of @xmath72 are computed via ( [ eq : bi ] ) as @xmath119 @xmath120 and @xmath121 respectively . comparing the above with the power series of the limit of @xmath72 , @xmath122 one
indeed notices that the first @xmath117 powers of @xmath76 in the series of @xmath72 match their infinite @xmath1 counterpart ( [ eq : psa ] ) .
thus , the ` bulk ' of @xmath72 must be very close to its asymptotic limit with the difference mainly resided in the tail as seen in figure [ fig : p1 ] .
the equality in moments up to such a high order indicates that the rate of convergence is quite fast .
indeed , it was proven in @xcite that for a class of linear statistics including ( [ eq : x ] ) , the convergence rate to a gaussian limit is @xmath123 for some constant @xmath124 . this super - exponential rate of convergence for linear statistics is uncommon in random matrix theory . as will be seen , the asymptotical formula ( [ eq : a ] ) of @xmath72 is a direct consequence of szeg s strong limit theorem @xcite when properly interpreted .
we first rewrite the integral ( [ eq : mdi ] ) as @xmath125 where @xmath126 following the same steps that led ( [ eq : mdi ] ) to ( [ eq : mdii ] ) , we have @xmath127 where @xmath128 can be interpreted as the fourier coefficient of the function ( [ eq : f ] ) .
thus , @xmath72 is understood as the determinant of a toeplitz matrix @xmath129 formed by fourier coefficients of the function @xmath130 . in this form ,
the limiting behavior of the toeplitz determinant ( [ eq : tdfc ] ) is known and is given by szeg s strong limit theorem @xcite , which is stated as follows .
if @xmath130 is integrable over the unit circle and the sum @xmath131 converges , where @xmath132 then @xmath133 the landau symbol @xmath134 is understood as @xmath135 and in particular ( [ eq : sslt ] ) is asymptotically tight . in our case
there are only two non - zero fourier coefficients for the function @xmath136 , @xmath137 inserting the above into ( [ eq : sslt ] ) we establish the claimed limiting formula ( [ eq : a ] ) .
szeg s study of asymptotical toeplitz determinants @xcite was motivated by problems in statistical physics in 1940 s .
in particular , the spin to spin correlation function of two - dimensional ising model can be written as a toeplitz determinant @xcite , whose asymptotics corresponds to the thermodynamics limit of such a system .
szeg s original proof @xcite imposed strong conditions on @xmath138 , which were successively weakened by several mathematicians in subsequent work .
the theorem was proven in the present form , among others , in @xcite . for a concise survey on recent development of asymptotical toeplitz determinants , we refer to @xcite .
we end this subsection by pointing out a by - product of this paper . since @xmath72 also equals the determinant ( [ eq : bi ] ) , a limiting formula for a toeplitz determinant of bessel functions ( [ eq : bf ] )
is read off as @xmath139 which seems new . without using tools from matrix - variate hypergeometric functions or szeg s strong limit theorem , it seems challenging to provide an elementary proof of this interesting asymptotic identity .
we are now in a position to state the main technical result of this paper . inserting the limiting expression ( [ eq : a ] ) into ( [ eq : r1 ] ) , a simple integration yields an asymptotic formula of the volume ( [ eq : v ] ) , @xmath140 where @xmath141 is gauss error function and ( [ eq : r2 ] ) is asymptotically tight as @xmath41 . ) versus asymptotic ( [ eq : r2]).,width=432 ] due to the fast convergence of @xmath72 as observed in figure [ fig : p1 ] and the rapid decreasing nature of the function @xmath142 in ( [ eq : r2i ] )
, we expect that the asymptotic formula ( [ eq : r2 ] ) approaches the exact one quite fast .
indeed , we can see from figure [ fig : p2 ] the approximate volume ( [ eq : r2 ] ) is already very accurate for @xmath1 as small as three . in figure
[ fig : p2 ] , the exact volume for @xmath67 is calculated by ( [ eq : n2 ] ) and the ones for @xmath143 are obtained by numerical integration of ( [ eq : r1 ] ) .
in this section , we apply the derived volume formulas to the study of coding - theoretic quantities defined in section [ subsec : spb ] .
as will be shown , the asymptotic volume expression ( [ eq : r2 ] ) leads to simple yet accurate analytical formulas of the bounds .
this fact may not be always mentioned in the following as the finite - size accuracy is of primary concern in practice . ] on cardinality ( [ eq : boundsa ] ) , code rate ( [ eq : ratea ] ) as well as its scaling law ( [ eq : rateasy ] ) , minimum distance ( [ eq : mindista ] ) , ( [ eq : mindista2 ] ) and diversity sum ( [ eq : diversitya ] ) . inserting the limiting volume formula ( [ eq : r2 ] ) into ( [ eq : bounds ] ) , the gilbert - varshamov lower bound and the hamming upper bound on the cardinality @xmath33 of a unitary code ( [ eq : c ] )
are obtained as @xmath144 the corresponding bounds on code rate ( [ eq : ratedef ] ) are read off as @xmath145 a notably important metric in coding theory is the scaling law of code rate , which establishes existence results of codes as dimension grows to infinity @xcite .
the obtained analytical bounds on code rate ( [ eq : ratea ] ) lead to elementary bounds on limiting code rate . to wit , when dimension @xmath1 approaches infinity with the same speed as the squared minimum distance @xmath146 , i.e. under the regime @xmath147 by using an asymptotic expansion of error function @xmath148 and after some straightforward if tedious manipulations , it can be shown that the following limit exists @xmath149},\ ] ] where @xmath150}=\left\{\begin{array}{l l } 1 & \quad \lambda\in[a , b]\\ 0 & \quad \text{otherwise } \end{array}\right.\ ] ] defines an indicator function . in ( [ eq : ratelim ] ) , the limit is taken over the regime ( [ eq : regime ] ) and @xmath151 denotes the natural logarithm . as a direct consequence , the normalized limiting code rate @xmath152 is bounded by @xmath153}\leq\widetilde{r}\leq\frac{(\lambda-8)^{2}}{64\ln(2)}\mathbbm{1}_{[0,8]}.\ ] ] as a function of minimum distance @xmath3 : convergence of the asymptotic bounds ( [ eq : ratea ] ) to the bounds on scaling law ( [ eq : rateasy]).,width=432 ] thus , in the asymptotic regime ( [ eq : regime ] ) there exist codes in the unitary group with code rate per dimension bounded by ( [ eq : rateasy ] ) .
another interpretation of ( [ eq : rateasy ] ) is as a scaling law of code rate bounds , where clearly the bounds scale linearly with dimension in the regime ( [ eq : regime ] ) .
similar results for codes in grassmann manifold under a different asymptotic regime can be found in ( * ? ? ?
* th . 2 ) .
in figure [ fig : pr ] we plot the asymptotic bounds on code rate ( [ eq : ratea ] ) and bounds on the corresponding scaling law ( [ eq : rateasy ] ) as a function of minimum distance @xmath3 .
we see that as @xmath1 and @xmath3 increase the rate bounds ( [ eq : ratea ] ) approach the bounds on scaling law ( [ eq : rateasy ] ) , as expected .
the convergence is reasonably fast as for @xmath154 the differences of the respective bounds seem indistinguishable for the range of minimum distances @xmath3 considered even though the scaling law bounds ( [ eq : rateasy ] ) are formally valid only when @xmath41 , @xmath155 .
it is also observed from figure [ fig : pr ] that the respective lower and upper bounds merge in the high rate regime , whereas their differences increase monotonically as the code rate decreases .
in fact , from the simple scaling bounds ( [ eq : rateasy ] ) one could easily deduce that the maximum gap ( at @xmath156 ) of minimum distances equals @xmath157 for any @xmath1 . by inverting the bounds on code rate ( [ eq : ratea ] ) , the resulting bounds on minimum distance @xmath3 as a function of code rate @xmath44 is obtained as @xmath158 where @xmath159 is inverse error function .
we now turn to diversity sum ( [ eq : diverdef ] ) , which is relevant to performance of unitary space - time codes . by inverting the volume formula ( [ eq : r2 ] ) and inserting it into ( [ eq : diversity ] ) , after some manipulations we arrive at an analytical expression for the upper bound on diversity sum @xmath160 [ cols="^,^,^,^,^,^,^,^,^",options="header " , ] [ table : md ]
namely , the ratio of bounds tends to a unique constant as @xmath41 for any code rate @xmath44 . though the upper bound @xmath161 is at most @xmath162 times looser that the modified bound @xmath163 , their difference is shrinking in the high rate regime as seen from table [ table : md ] .
we also see that for a sufficient large @xmath1 the upper and lower ) is always half of the corresponding upper bound . ]
bounds on minimum distance increase especially in the high rate regime , see e.g. the case @xmath164 in table [ table : md ] . ] with @xmath1 .
this advocates design of large dimensional space - time codes , where there potentially exists codes with improved error performance , i.e. with a larger minimum distance .
a similar conclusion was drawn based on the tabulated minimum distance bounds in @xcite , where , however , some entries of their tables can not be computed due to the small - ball assumption made in @xcite .
in this paper , we derive a limiting volume formula of metric balls in unitary group based on an exact integral representation .
this simple - to - evaluate asymptotic result is obtained by means of random matrix theory .
the fast convergence of the derived formula is due to the identified mock - gaussian behavior .
the proposed volume formula enables analytical characterization of coding - theoretic bounds of unitary codes .
closed - form expressions for bounds on cardinality , code rate including its scaling law , minimum distance and diversity sum have been derived .
these analytical formulas help gain crucial insights into the behavior of the bounds .
in particular , the largest possible minimum distance gaps between the lower and upper bounds as well as between the upper bound and a modified one are quantified .
it is also found that the code rate bounds obey linear scaling laws in some asymptotic regime of interest .
possible future work includes volume computations in other compact classical groups such as orthogonal , symplectic groups , and in the stiefel manifold .
l. wei is supported by the academy of finland ( grant 251170 ) and nokia foundation .
pitaval is supported by the academy of finland ( grants 276031 , 282938 , 283262 ) and the european science foundation under the cost action ic1104 .
j. corander is supported by the academy of finland ( grant 251170 ) .
o. tirkkonen is supported by the academy of finland ( grant 284725 ) .
for two @xmath61 matrices @xmath165 and @xmath166 , with the respective @xmath167-th entry being functions @xmath168 and @xmath169 , and a function @xmath170 such that the integral @xmath171 exists , then the following multiple integral can be evaluated as @xmath172 where @xmath173 .
the general form of meijer s g - function @xcite is @xmath174 where the contour @xmath175 is chosen in such a way that the poles of @xmath176 , @xmath177 are separated from the poles of @xmath178 , @xmath179 . by definition ,
meijer s g - function is mellin transform pair of @xmath180 in the following , we list some properties and results of meijer s g - function that will be utilized in the proof of ( [ eq : n2 ] ) : * shifting property @xmath181 * inverse argument formula @xmath182 * a relation with hypergeometric functions @xmath183 * an integration identity @xmath184 * a relation between a meijer s g - function and a hypergeometric function @xmath185 * a relation between a meijer s g - function and a hypergeometric function @xmath186
by the definition of bessel function ( [ eq : bf ] ) it holds @xmath187 , thus for @xmath67 , @xmath188 the integral ( [ eq : r1 ] ) can now be split into @xmath189 where @xmath190 since ( * ? ? ? * eq . 8.442 ) , @xmath191 by the relation to meijer s g - function ( [ eq : hf ] ) , we obtain @xmath192 by the shifting property ( [ eq : sp ] ) , the inverse argument formula ( [ eq : ia ] ) and the integration identity ( [ eq : ii ] ) , we have @xmath193 the case @xmath194 is obtained by using the representation ( [ eq : r1 ] ) as @xmath195 the integral @xmath196 is calculated by following the procedure of @xmath197 as @xmath198 where the last equality is established by the definition of meijer s g - function ( [ eq : meijerg ] ) .
the case @xmath199 is obtained by using the representation ( [ eq : r2 ] ) as @xmath200 inserting ( [ eq : ip1 ] ) , ( [ eq : ip2 ] ) , ( [ eq : ip3 ] ) and ( [ eq : ip4 ] ) into ( [ eq : ip ] ) , we obtain the claimed result ( [ eq : n2 ] ) . 99 a. shokrollahi , b. hassibi , b. m. hochwald , and w. sweldens , `` representation theory for high - rate multiple - antenna code design , '' _ ieee trans .
theory , _ vol .
47 , no . 6 ,
2335 - 2367 , sept .
x .- b . liang and x .-
xia , `` unitary signal constellations for differential space - time modulation with two transmit antennas : parametric codes , optimal designs and bounds , '' _ ieee trans .
theory , _ vol .
48 , no . 8 , pp . 2291 - 2322 , aug .
2002 .
r. t. krishnamachari and m. k. varanasi , `` on the geometry and quantization of manifolds of positive semi - definite matrices , '' _ ieee trans .
signal process .
4587 - 4599 , sept .
2013 .
p. deift , a. its , and i. krasovsky , `` toeplitz matrices and toeplitz determinants under the impetus of the ising model : some history and some recent results , '' _ comm .
pure appl . math .
66 , sept . 2013 . | volume estimates of metric balls in manifolds find diverse applications in information and coding theory . in this paper ,
some new results for the volume of a metric ball in unitary group are derived via various tools from random matrix theory .
the first result is an integral representation of the exact volume , which involves a toeplitz determinant of bessel functions .
the connection to matrix - variate hypergeometric functions and szeg s strong limit theorem lead independently from the finite size formula to an asymptotic one .
the convergence of the limiting formula is exceptionally fast due to an underlying mock - gaussian behavior .
the proposed volume estimate enables simple but accurate analytical evaluation of coding - theoretic bounds of unitary codes .
in particular , the gilbert - varshamov lower bound and the hamming upper bound on cardinality as well as the resulting bounds on code rate and minimum distance are derived .
moreover , bounds on the scaling law of code rate are found .
lastly , a closed - form bound on diversity sum relevant to unitary space - time codes is obtained , which was only computed numerically in literature .
coding - theoretic bounds , random matrix theory , unitary group , volume of metric balls . |
given a simple , closed curve in three - space is there a surface with minimal area spanning it ?
solutions to the problem of plateau depend , of course , on the class of spanning surfaces permitted .
douglas @xcite won the first fields medal for his proof of the existence of an area minimizing mapping of the 2-disk whose image spans the jordan curve .
regularity took many years to establish @xcite and some aspects are still unresolved .
federer and fleming s solutions @xcite are area minimizing in the class of integral currents .
two years later fleming @xcite proved any such solution is an embedded , orientable surface , smooth away from its boundary .
none of these solutions consider soap films that arise in nature such as moebius strips or films with triple branching .
almgren @xcite invented varifolds to treat soap films , but the lack of a natural boundary operator slowed progress and regularity was never proved .
plateau observed that soap films have only two possible kinds of branching : ( 1 ) three sheets of surface meeting at @xmath1 angles along a curve and ( 2 ) four such curves meeting at approximately @xmath2 angles at a point @xcite . in @xcite
the author provides models for surfaces called _ flat dipolyhedra _ that model all such films , orientable or nonorientable , as well as the surfaces considered in @xcite and @xcite .
flat dipolyhedra take advantage of the fact that soap films are actually two films essentially occupying the same space , but are not cancelling . there is a natural boundary operator of flat @xmath3-dimensional dipolyhedra into flat @xmath4-dimensional dipolyhedra which relies on cohesion of the soap film structure supported by a geometric version of cartan s magic formula .
the _ energy _ of a flat dipolyhedron is defined to be the length of the singular branched set of @xmath5 plus the surface area of @xmath5 . in this paper
we prove the existence of a surface spanning a lipschitz jordan curve that is area minimizing in the space of flat dipolyhedra @xmath5 with energy bounded by a fixed constant .
complete soap film regularity remains an open question , although we prove here that the solution is a smooth surface away from its branched set which is a union of lipschitz jordan curves of finite total length .
a _ @xmath3-cell _ @xmath6 in @xmath7 is defined to be a finite intersection of @xmath3-dimensional half spaces .
@xmath3-cell in @xmath8 _ is a @xmath3-cell in a @xmath3-dimensional subspace of @xmath9 each @xmath3-cell is assumed to be oriented .
the support of a @xmath3-cell @xmath6 is denoted by @xmath10 an _ integral cellular @xmath3-chain _ is a formal sum @xmath11 where @xmath12 and the @xmath13 are @xmath3-cells .
equate two cellular chains @xmath14 if and only if they have a common cellular subdivision which is nonoverlapping .
a polyhedral @xmath3-chain is an equivalence class of cellular @xmath3-chains .
the _ mass _ of a @xmath3-cell @xmath6 is its @xmath3-dimensional hausdorff measure @xmath15 if @xmath16 is a polyhedral @xmath3-chain represented by a nonoverlapping @xmath3-chain @xmath17 , its mass is defined by @xmath18 let @xmath19 be an abelian group with a translation invariant metric making it a complete metric space .
let @xmath20 denote the distance between @xmath21 and the group identity .
if @xmath22 is a closed subgroup of @xmath19 , we use the quotient metric @xmath23 if @xmath24 then @xmath20 denotes the absolute value .
the group @xmath25 is of special interest and we give it the quotient metric .
let @xmath26 this is the group of polyhedral @xmath3-chains with coefficients in @xmath19 . if @xmath27 then @xmath28 where the @xmath13 are nonoverlapping @xmath3-cells .
if @xmath29 , define @xmath30 whitney s flat norm on polyhedra is defined by @xmath31 the completion of the group @xmath32 with the norm @xmath33 is denoted @xmath34 .
its elements are called _ flat chains _ with coefficients in @xmath19 . the _ support _ of a flat chain @xmath5
is well defined ( see @xcite ) and is denoted @xmath35 define subgroups @xmath36 is a cell and @xmath37 a vector in @xmath8 then @xmath38 denotes the translation of @xmath39 through @xmath37 .
@xmath40-multicell _ is a cellular chain of the form @xmath41 where @xmath42 is a cell and @xmath43 is a vector . given a vector @xmath44 and a @xmath45-multicell @xmath46 , define the _
@xmath47-multicell _ @xmath48 as the cellular chain @xmath49 thus @xmath48 is generated by vectors @xmath50 and a cell @xmath51 an _
integral @xmath47-multicellular chain in @xmath8 _ is a formal sum of @xmath47-multicells , @xmath52 with coefficients @xmath53 .
let @xmath54 given a @xmath47-multicell @xmath48 generated by a cell @xmath42 and vectors @xmath55 , define @xmath56 where @xmath57 denotes the norm of a vector @xmath58 .
for consistency of notation define @xmath59 for @xmath60 define @xmath61 denote polyhedral @xmath3-chains in @xmath8 .
suppose @xmath62 and @xmath63 .
for @xmath64 define @xmath65 for @xmath66 define the _
@xmath67-natural _ norm @xmath68 where the infimum is taken over all decompositions @xmath69 + \partial c $ ] with @xmath70 a @xmath47 multicellular @xmath3-chain and @xmath71 a polyhedral @xmath72-chain .
the group of polyhedral @xmath3-chains @xmath73 completed with the norm @xmath74 is denoted @xmath75 elements of @xmath76 are called _ @xmath3-dimensional chainlets of class @xmath77 . _
the boundary operator @xmath78 is naturally defined and continuous .
let @xmath6 be a @xmath3-cell supported in @xmath8 and set @xmath79 in @xcite _ mass cells _ are defined as limits in the @xmath40-natural norm @xmath80 integral mass chains are finite sums of mass cells with integer coefficients .
although the support of a mass cell is @xmath3-dimensional , as a current it is @xmath72-dimensional .
_ dipole cells _ are defined by @xmath81 dipole chains @xmath82 are finite sums of dipole cells .
the _ support _ of a mass chain @xmath83 is defined by @xmath84 and the _ support _ of a dipole chain @xmath85 is defined to be @xmath86 define the _ mass _ of a mass chain @xmath87 by @xmath88 and the _ weight _ of a dipole chain @xmath82 by @xmath89 where the @xmath90 are nonoverlapping .
spaces of mass and dipole chains are denoted @xmath91 and @xmath92 , resp .
let @xmath19 be an abelian group . by taking the tensor product with @xmath19
, we may define spaces of mass and dipole chains with coefficients in @xmath19 and denote them by @xmath93 and @xmath94 , resp .
define the space of _ dipolyhedra _ as the direct sum @xmath95 if @xmath96 define the _ energy _ of @xmath97 by @xmath98 finally define the @xmath99 norm on the space of @xmath3-dimensional dipolyhedra by @xmath100 the completion of the space of dipolyhedra @xmath101 with the @xmath99 norm is an abelian group denoted @xmath102 the boundary operator is continuous in the @xmath99 norm and satisfies @xmath103 an element @xmath104 is called a _
flat dipolyhedron _ with coefficients in @xmath19 . in @xcite
it is shown that weight , mass and energy are well defined and lower semi - continuous in @xmath105 .
the operators @xmath106 and @xmath107 satisfy @xmath108 for every flat chain @xmath5 ( @xcite 5.2 ) ) .
henceforth we set @xmath109 and @xmath110 for our application to soap films in three - space . in this case
the weight of a @xmath3-dipolyhedron coincides with its hausdorff @xmath3-measure .
thus , for @xmath111 , the quantity @xmath112 can be thought of as the _ area _ of @xmath97 and for @xmath113 it is the _ length _ of @xmath97 .
the orthogonal projection of a dipole @xmath3-chain @xmath82 into @xmath114 is a mod two @xmath3-polyhedron denoted @xmath115 .
then @xmath116 similarly , the orthogonal projection of a mass @xmath3-chain @xmath87 into @xmath117 is a mod two @xmath4-polyhedron denoted @xmath118 and satisfies @xmath119 [ converge ] suppose @xmath120 . then @xmath121 given @xmath122 there exist @xmath123 with @xmath124 and @xmath125 note that @xmath126 since @xmath127 then @xmath128 since the boundary of a dipole chain is also a dipole chain @xmath129 and therefore @xmath130 it follows that @xmath131 for the second inequality , note that @xmath132 is a dipole chain and @xmath133 is a sum of a dipole chain and a mass chain @xmath134 as before , @xmath135 therefore @xmath136 since @xmath137 it follows that @xmath138 since these inequalities hold for all @xmath122 , the proposition follows . [ split ] if @xmath5 is a flat @xmath3-dipolyhedron then there exist a unique flat @xmath3-chain @xmath139 and a unique flat @xmath4-chain @xmath71 such that @xmath140 and @xmath141 if @xmath142 with @xmath143 then @xmath144 , @xmath145 and @xmath146 suppose @xmath147 where @xmath148 . by proposition [ converge ] @xmath149 is a cauchy sequence in @xmath150 let @xmath139 denote its flat chain limit . by 5.2 of @xcite @xmath151
similarly @xmath152 converges to a flat chain @xmath71 in @xmath33 and @xmath153 it follows that @xmath154 we next prove uniqueness : suppose @xmath155 then @xmath156 by propotision [ converge ] @xmath157 and @xmath158 it follows that @xmath159 since @xmath33 is a norm .
uniqueness of @xmath139 and @xmath71 follows .
suppose @xmath160 by lower semicontinuity of mass in the flat norm @xcite , @xmath161 thus @xmath162 since @xmath163 and @xmath164 and all terms are nonnegative the result follows .
if @xmath140 we say that @xmath5 _ splits _ into @xmath165 and @xmath166 . since the splitting is unique we may define @xmath167 and @xmath168 [ boundary ] if @xmath140 is a flat dipolyhedron then @xmath169 acccording to ( @xcite , 3.2 ) @xmath170 and @xmath171 the result follows . [ start ] if @xmath172 is a flat @xmath4-chain and @xmath140 is a flat @xmath3-dipolyhedron satisfying @xmath173 then @xmath174 and @xmath175 it follows from lemma [ boundary ] that @xmath176 since the splitting is unique ( theorem [ split ] ) it follows that @xmath177 and @xmath178 let @xmath139 be a flat chain with finite mass . according to ( @xcite , 4 ) there exists a borel measure @xmath179 and for every borel set @xmath180 there exists a flat chain @xmath181 such that @xmath182 moreover , if @xmath183 with @xmath184 then @xmath185 and @xmath186 for every @xmath187 such that @xmath188 ( @xcite 4 ) define @xmath189 and @xmath190 the _ part _ of @xmath165 in @xmath187 and the _ part _ of @xmath191 in @xmath187 , respectively .
suppose @xmath192 is a flat dipolyhedron with bounded energy @xmath193 . from theorem [ split ] we know
@xmath140 where @xmath139 and @xmath71 are flat chains with @xmath194 .
define @xmath195 we call @xmath196 the _ part of _ @xmath5 in @xmath187 .
define borel measures @xmath197 , @xmath198 and @xmath199 the next proposition follows directly . [ exceptional ] if @xmath140 is a flat dipolyhedron with finite energy and @xmath187 is a borel set there exists a unique flat dipolyhedron @xmath196 such that @xmath200 , @xmath201 @xmath202 , @xmath203 , @xmath204 and @xmath205 if @xmath206 with @xmath143 then @xmath207 and @xmath208 for all @xmath187 such that @xmath209 the _ support _ of a borel measure @xmath210 is the smallest closed set @xmath187 whose complement is @xmath210-null .
say that @xmath210 is a _ measure on @xmath211 _ if @xmath211 contains the support of @xmath210 .
we say a closed set @xmath212 _ supports _ a flat dipolyhedron @xmath5 if for every open set @xmath213 containing @xmath212 there is a sequence @xmath214 of dipolyhedra tending to @xmath5 in @xmath99 such that @xmath215 for each @xmath216 .
if there is a smallest set @xmath212 which supports @xmath5 then @xmath212 is called the _ support _ of @xmath5 and denoted @xmath217 .
the next theorem shows that every flat dipolyhedron with finite energy has a well defined support .
[ support ] if @xmath218 with @xmath219 then @xmath220 by theorem [ split ] we know @xmath140 where @xmath139 and @xmath71 are flat chains with @xmath221 and @xmath222 apply ( @xcite , 4.3 ) to deduce @xmath223
let @xmath6 be a @xmath3-cell in supported in @xmath224 the @xmath225-cube in @xmath117 centered at @xmath226 with side length @xmath67 .
the cone @xmath227 is also a cell found by intersecting the cones over the half - spaces forming @xmath228 its boundary satisfies @xmath229 and @xmath230 define @xmath231 and @xmath232 it follows that @xmath233 is a well defined dipole cell with @xmath234 and @xmath235 is a well defined mass cell with @xmath236 extend the definition by linearity to define dipole chains @xmath237 and mass chains @xmath238 taken over dipole chains @xmath82 and mass chains @xmath87 , respectively . next extend the definition to dipolyhedra @xmath239 observe @xmath240 and @xmath241 .
[ coneboundary ] if @xmath242 then @xmath243 and @xmath244 if @xmath97 is supported in @xmath245 then so is @xmath246 and @xmath247 the first part reduces to showing @xmath248 and @xmath249 these follow directly from the definitions and linear relations .
suppose @xmath250 then @xmath251 where @xmath237 is a dipole chain and @xmath238 is a mass chain .
thus @xmath252 [ est ] if @xmath253 is supported in @xmath245 then @xmath254 let @xmath255 there exist dipolyhedra @xmath256 and @xmath257 such that @xmath124 and @xmath258 by proposition [ coneboundary ] @xmath259 for all dipolyhedra @xmath260 .
then @xmath261 since @xmath262 it follows that @xmath263 the result follows since this holds for all @xmath122 .
it follows that if @xmath5 is a flat dipolyhedron then the cone @xmath264 has unique definition as a flat dipolyhedron as follows : if @xmath206 then @xmath265 forms a cauchy sequence .
denote its limit by @xmath266 [ coneboundarya ] if @xmath267 then @xmath268 if @xmath5 is supported in @xmath245 then @xmath269 and @xmath270 the first two relations follow from proposition [ coneboundary ] , theorem [ est ] and continuity of the boundary operator .
since energy is lower semicontinuous , there exists @xmath271 such that @xmath272 by theorem [ est ] @xmath273 and hence @xmath274 let @xmath275 be a lipschitz mapping .
extend @xmath276 to @xmath277 by @xmath278 if @xmath139 is a flat @xmath3-chain supported in an open set @xmath279 then the pushforward @xmath280 is well defined as a flat @xmath3-chain and satisfies @xmath281 and @xmath282 ( @xcite ) .
it is called a _
lipschitz chain_. a chainlet @xmath283 is called a _ lipschitz dipolyhedron _ if @xmath139 and @xmath71 are lipschitz chains .
if @xmath284 define @xmath285 [ chain ] if @xmath286 is a @xmath3-dipolyhedron and @xmath275 is a lipschitz mapping with @xmath287 then @xmath288 is a @xmath3-dipolyhedron with @xmath289 , @xmath290 and @xmath291 by lemma [ boundary ] we know that @xmath292 since @xmath293 we have @xmath294 .
since @xmath295 is a chain map on flat chains we conclude @xmath296 . by theorem [ split ] and proposition [ chain ]
it follows that @xmath297 let @xmath122 .
there exists @xmath124 with @xmath298 since @xmath299 it follows that @xmath300 finally , @xmath301 let @xmath5 be a flat dipolyhedron . by lower semicontinuity of energy we may choose @xmath302 such that @xmath160 it follows that @xmath303 is a well defined flat dipolyhedron with @xmath304 and @xmath305 for @xmath306 , define @xmath307 for @xmath308 ,
define @xmath309 observe that @xmath276 has lipschitz constant @xmath310 denote @xmath311 , the projection of a dipolyhedron @xmath97 .
since @xmath312 we can write @xmath313 with ambiguity .
let @xmath314 then @xmath315 it follows from proposition [ chain ] projections @xmath316 are uniquely defined for
all flat dipolyhedra @xmath5 with @xmath317 and @xmath318 define @xmath319 @xmath320
the next result is the deformation theorem , first proved for integer coefficients by federer and fleming @xcite .
it was extended to @xmath321 coefficients in @xcite and to abelian groups in ( @xcite , 7.3 ) .
let @xmath322 be an @xmath323-cubical grid of @xmath324 a @xmath3-polyhedron @xmath16 is a _
polyhedron of @xmath322 _ if @xmath16 is supported in the @xmath3-skeleton of @xmath322 and @xmath325 is supported in its @xmath4-skeleton .
a @xmath3-dipolyhedron @xmath326 is a dipolyhedron _ of @xmath322 _ if @xmath139 and @xmath71 are both polyhedra of @xmath322 .
[ deformfleming ] there exists a positive number @xmath327 with the following property .
given @xmath328 and @xmath122 there exist an @xmath323-cubical grid @xmath322 , a polyhedral @xmath3-chain @xmath16 of @xmath322 , @xmath329 and @xmath330 such that : 1 .
@xmath331 2 .
@xmath332 3 .
@xmath333-neighborhood of @xmath334 + @xmath335-neighborhood of @xmath336 4 .
if @xmath5 is a polyhedral ( lipschitz ) chain then @xmath256 and @xmath260 are polyhedral ( lipschitz chains ) .
the proof makes use of successive radial projections from well chosen points in the interior of each @xmath3-cube of @xmath322 onto its lower dimensional skeleton which minimize distortion .
( see @xcite . )
we refer to the _ projection path _ of a flat chain @xmath5 as the rays traced by the projections of @xmath337 .
[ deforml ] let @xmath172 be a lipschitz jordan curve in @xmath117 .
there exists a positive number @xmath327 with the following property .
given @xmath338 with @xmath173 and @xmath122 there exist an @xmath323-cubical grid @xmath322 , a dipolyhedral @xmath3-chain @xmath97 of @xmath322 , @xmath339 and @xmath340 such that 1 .
@xmath341 ; 2 .
@xmath342 , + @xmath343 , + @xmath344 ; 3 .
@xmath345neighborhood of @xmath217 , + @xmath346neighborhood of @xmath347 .
if @xmath5 is a ( lipschitz ) dipolyhedron then @xmath256 and @xmath260 are ( lipschitz ) dipolyhedra .
suppose @xmath140 with @xmath348 by corollary [ start ] @xmath174 and @xmath349 .
apply theorem [ deformfleming ] to @xmath139 and @xmath71 to find @xmath350 and @xmath351 satisfying properties ( 1)-(4 ) of theorem [ deformfleming ] .
therefore @xmath352 and @xmath353 since @xmath354 it follows that @xmath355 now the polyhedra @xmath356 and @xmath260 are found by projecting @xmath139 and @xmath71 onto the @xmath3-skeleton of @xmath322 .
thus @xmath357 is the projection path of @xmath358 , @xmath359 is the projection path of @xmath71 .
hence @xmath360 is the projection path of @xmath361 which is the same as the projection path of @xmath172 .
hence @xmath362 similarly , @xmath363 is the projection of @xmath71 , @xmath364 is the projection of @xmath365 hence @xmath366 is the projection of @xmath367 . hence @xmath368 we say that a flat dipolyhedron @xmath5 _ spans _
@xmath382 if @xmath173 and the following condition holds : if @xmath187 is a @xmath383-dimensional subspace of @xmath117 and @xmath384 is an orthogonal projection which is an immersion of @xmath172 , then @xmath385 choose a projection @xmath390 so that @xmath391 is a jordan curve in @xmath187 .
let @xmath392 denote the chain whose support is the inside of @xmath391 in @xmath187 so that @xmath393 let @xmath394 the condition that @xmath395 implies that @xmath396 .
the result follows since @xmath397 * mappings of the two disk @xmath139 , @xmath403 , @xmath404 and @xmath276 is smooth away from @xmath358 .
* area minimizing integral currents * area minimizing soap films @xmath82 , as observed by plateau : the set @xmath82 is smooth away from its branch set @xmath139 . consider the connected components @xmath405 complementary to @xmath139 .
these are embedded and smooth mod two surfaces .
let @xmath406 and @xmath407 then @xmath408 and @xmath409 suppose @xmath410 where @xmath411 we know @xmath193 by lower semicontinuity of energy , and @xmath412 since each @xmath413 since each @xmath414 spans @xmath382 it follows that @xmath5 spans @xmath382 .
thus @xmath402 is closed .
use corollary [ deforml ] to show it is totally bounded . given
@xmath122 there exists a dipolyhedron @xmath97 such that @xmath341 with @xmath415 it follows that @xmath402 is totally bounded and thus compact . we show that @xmath416 since @xmath417 is a cone over a polyhedron it is a flat chain with @xmath418 ( @xcite 6 ) . by proposition [ coneboundary ]
@xmath419 and we know @xmath420 spans its boundary .
since @xmath421 the result follows let @xmath422 there exist @xmath423 such that @xmath424 by compactness ( theorem [ compact ] ) the sequence @xmath414 has a subsequential limit @xmath425 with @xmath426 then @xmath427 according to lemma [ delta ] it follows that @xmath428 it may be that there is another @xmath5 with smaller area spanning @xmath172 with @xmath429 that is not supported in @xmath430 .
let @xmath431 denote the projection into @xmath430 .
since @xmath5 spans its boundary so does @xmath432 .
from proposition [ chain ] we conclude @xmath433 . by proposition [ chain ] @xmath434 and @xmath435 , so we may replace @xmath5 with @xmath436 . by theorem [ split ] @xmath140 where @xmath139 and @xmath71 are flat chains ( mod two ) with finite mass and finite boundary mass .
decompose b into its indecomposable parts @xmath438 . )
since @xmath439 is area minimizing , we may apply mod two regularity ( @xcite ) to deduce each surface @xmath440 is smoothly embedded away from its boundary . in his thesis ,
ziemer proved that mod two boundaries are integral currents ( @xcite , 6.5 ) .
fleming proved that integral 1-cycles with finite mass are sums of closed curves each of which is lipschitz . since @xmath174 and @xmath441 then @xmath71 is a sum of lipschitz jordan curves with finite total length .
( see @xcite , 4.2.25 ) robert osserman , a proof of the regularity everywhere of the classical solution to plateau s problem , ann . of math .
91 , ( 1970 ) , 550 - 569 .
plateau , statique experimentale et theorique des diquides soumis aux seules forces moleculaires , paris , gauthier - villars , 1873 . | we prove existence and a.e .
regularity of an area minimizing soap film with a bound on energy spanning a given jordan curve in @xmath0 . |
the performance of interferometers at ( sub)millimetre wavelengths is often limited by differential fluctuations in the atmospheric path along the line of sight to each of the constituent antennas .
if uncorrected , these fluctuations lead to a loss in sensitivity , imaging artefacts and a limit on the maximum usable baseline ( e.g. @xcite ; @xcite ) .
alma will correct such path variations using a combination of techniques : 1 .
fast - switching observations of bright calibrator sources ( e.g. quasars ) .
water vapour radiometry using dedicated 183-ghz radiometers ( wvrs ) , installed in every 12-m alma antenna .
the wvrs will allow the retrieval of the amount of water vapour along the line of sight to each antenna and thus fluctuations in the atmospheric path length resulting from this water vapour .
3 . self calibration , for a fraction of the brightest science targets . fast switching interleaves science observations with short ( @xmath2 a few seconds ) calibration observations on a cycle time of 20200s .
these calibrations allow an estimate of the atmospheric and instrumental phase errors _ in the direction of the calibrator_. the applicability of the calibrator phase solutions to the science target depends on their angular separation and the observing frequency
. the radiometric technique will be applied _
continuously _ , correcting for atmospheric path fluctuations on timescales shorter than the fast - switching cycle time , but does require an accurate model to convert the wvr measurements into path delays .
variations in the atmospheric path delay to the antennas predominantly arise from a combination of fluctuations in the water vapour content _ and _ density of air in the troposphere respectively , the so - called wet and dry path components .
furthermore , we can split both components ( the wet and dry ) into two parts : one dispersive ( i.e. dependent on frequency ) and the other non - dispersive ( independent of frequency ) .
conventionally , the non - dispersive part is taken to be the path in the low - frequency limit .
this memo focuses on the dispersive part of the wet path delay introduced by the atmosphere .
previously , atmospheric dispersion in the context of fast switching was studied by @xcite and @xcite .
they found the magnitude of the dispersive phase will become non - negligible in alma s submillimetre bands .
furthermore , @xcite quantified the dispersive phase delay for _ typical _ conditions at the alma site .
we extend their work by investigating how the dispersive path delay depends on the _ variation _ of the physical parameters of the atmosphere and what constraints can be placed on it from the proposed ancillary calibration devices on site .
this memo is structured into six parts .
first , we explain the planned phase - correction strategy for alma and where dispersive effects play a role .
second , in , we describe how dispersive and non - dispersive path delays arise in the atmosphere and how we compute the dispersive path delay . also quantifies the dispersive path delay across the full range of alma observing frequencies , before demonstrates how varying different atmospheric parameters changes the magnitude of the dispersive path delay .
finally , we look at what constraints can be placed on the dispersive path delay with temperature measurements from weather - monitoring devices .
before we consider the magnitude of the dispersive path delay , we will detail the current phase - correction plans for alma and how these are affected by atmospheric dispersion .
numerous previous memos have discussed fast - switching phase correction for alma ( e.g. @xcite ) .
briefly , fast switching removes some of the antenna - based phase fluctuations by regularly making short observations of a bright point - source calibrator , close to the science target in the sky .
this difference in direction between the target and calibrator means the calibration solutions need to be interpolated to estimate the science target s phase errors . typically , at alma s highest observing frequencies , it will be unlikely that the calibrations can be taken at the same frequency as the science observations , for the following reasons : 1 .
calibrators of the necessary strength at the science frequency will probably be too far from the science target ( although this depends on the source counts at high frequencies , which are currently not well known ) .
2 . the combination of high frequencies and long baselines may resolve many potential calibrators .
3 . the phase errors may prove to be so large that phase wrapping makes reliable solutions difficult . to overcome these difficulties ,
one proposal is for alma to observe calibration sources at 90ghz when necessary and then scale the corresponding phase solutions to the science frequency ( the _ phase - transfer _ technique ) .
the frequency above which the phase transfer - approach will be necessary depends on the array configuration , atmospheric conditions and ultimately experience gathered at the site .
it is currently expected to be the routine mode at wavelengths below 2 mm ( frequencies @xmath3ghz ) . in the absence of dispersion ,
the required phase scaling is simply the ratio of the calibration and science frequencies .
however , at high frequencies , as we have already mentioned and will detail below , the numerous nearby water vapour lines ensure that atmospheric dispersion will need to be taken into account .
fluctuations in the atmospheric path on timescales shorter than the fast - switching cycle time will be corrected radiometrically using wvrs installed in every 12-m alma antenna .
the wvrs measure the brightness of the 183-ghz atmospheric water line , which is very sensitive to the total amount of water vapour in the atmosphere @xcite .
the fundamental difficulty in the analysis of wvr data is how to convert fluctuations in the measured sky brightnesses around the 183-ghz water line into variations in the atmospheric path delay .
we have begun to develop a bayesian framework for computing such conversion factors @xcite , which naturally incorporates the prior knowledge of the system and all the observable data . in the first memo @xcite , we begin with the simplest possible model atmosphere , comprising a single , thin layer of water , which is the only cause of path fluctuations .
even so , an analysis of test data from prototype wvrs that were installed on the submillimeter array ( sma ) yields corrections that are within @xmath4per cent of optimal ones .
memo 588 @xcite extends this scheme to include the observed correlation between phase and sky brightness , which alma may implicitly record during fast switching , when the calibrator s phase is measured whilst data are taken with the wvrs .
the inclusion of this empirical relationship should significantly improve the accuracy of the phase correction .
typically , the calculation of phase - correction coefficients ( i.e.the conversion factors above ) from wvr data requires a model of the atmosphere , which in principle can be used to estimate the dispersive as well as non - dispersive path delays . in our work to date , which was based on relatively low - frequency ( 220ghz ) data , we have not modelled the dispersive path contribution .
instead , we simply scale the non - dispersive path delay by a constant factor to estimate the total wet delay .
however , in reality this factor is a function of frequency .
furthermore , if the dispersive contribution is a significant fraction of the total path delay , it must be determined with good accuracy . to date
there has been little work on what atmospheric properties influence the magnitude of the dispersive path delay and by what amount . in
, we therefore examine the variation of the dispersive path delay using multiple realistic models of the atmosphere above the alma site .
fluctuations in the wet path delay , @xmath5 , are separated into the sum of non - dispersive ( @xmath6 ) and dispersive delays ( @xmath7 ) : the _ total _ non - dispersive path delay ( i.e. the sum of the wet and dry components ) can be computed from the smith - weintraub equation for the refractive index , @xmath9 , at a temperature , @xmath10 @xcite : @xmath11\rmn{,}\ ] ] where @xmath12 and @xmath13 are the partial pressures of the dry air and water vapour respectively and @xmath14 , @xmath15 and @xmath16 are constants .
since we are only concerned with the excess path introduced by water vapour ( i.e. d@xmath17 above ) , then we may omit the first term and of the remaining terms the last one dominates .
hence we can transform to the following expression @xcite:@xmath18 where @xmath19 is the water vapour column . in general , atmospheres with spectral lines
are dispersive , with the dispersion related to the absorption according to the kramers - krnig relations .
thus , any atmospheric property that affects the shapes of absorption - line features will probably influence the atmospheric dispersion , summarized in the following functional form ( see for a list of the symbols used ) : we calculate the dispersive and non - dispersive contributions to the wet and dry path delays between 1ghz and 1thz from model atmospheres using the atmsoftware @xcite .
atmaccurately predicts the atmospheric opacity above mauna kea , hawaii up to 1.6thz @xcite .
its dispersive calculations have not been as thoroughly verified due to a lack of high - frequency test data , which should change in the near future once the alma wvrs become operational at chajnantor . in , we plot the wet and dry , dispersive and non - dispersive path delays overlaid on top of the atmospheric transmission , all computed using the basic model parameters listed in , which are suitable for the alma array operations site ( aos ) at chajnantor .
the parameters listed in are the median values measured by site - monitoring equipment where possible and their origin will be detailed in later sections .
the computed dispersive path delay is small except around the atmospheric h@xmath22olines ( for the wet component ) or o@xmath22lines ( for the dry ) .
the largest contribution to the total delay at all frequencies arises from the dry non - dispersive path delay .
however , we expect the dry column to be relatively stable and so contribute very little to the _ differential _ path delay @xcite i.e.the difference in delay between antennas .
we will ignore the dry path delays for the rest of this memo but they may prove to be a significant extra source of error that remain after phase correction using the wvrs . | the success of any alma phase - calibration strategy , which incorporates phase transfer , depends on a good understanding of how the atmospheric path delay changes with frequency ( e.g. @xcite ) .
we explore how the wet dispersive path delay varies for realistic atmospheric conditions at the alma site using the atmtransmission code .
we find the wet dispersive path delay becomes a significant fraction ( @xmath0per cent ) of the non - dispersive delay for the high - frequency alma bands ( @xmath1ghz , bands 5 to 10 ) .
additionally , the variation in dispersive path delay across alma s 4-ghz contiguous bandwidth is not significant except in bands 9 and 10 .
the ratio of dispersive path delay to total column of water vapour does not vary significantly for typical amounts of water vapour , water vapour scale heights and ground pressures above chajnantor .
however , the temperature profile and particularly the ground - level temperature are more important . given the likely constraints from alma s ancillary calibration devices ,
the uncertainty on the dispersive - path scaling will be around 2per cent in the worst case and should contribute about 1per cent overall to the wet path fluctuations at the highest frequencies .
[ firstpage ] |
ss 433 is a binary system which is thought to consist of a compact object ( probably a black hole ) and optical star filling its roche lobe .
there is a supercritical accretion rate in the system ( fabrika 2004 ) , which leads to formation of the geometrically thick accretion disk around compact object and two oppositely directed highly collimated jets .
the main contribution to the observed x - ray emission from the system comes from the optically thin multitemperature plasma in the jets via bremsstrahlung mechanism . the x - ray emission from ss433 is subjected to systematic variations due to the thick accretion disk precession ( @xmath5 days ) , nutation ( @xmath6 days ) and orbital ( @xmath7 days ) motion in the binary system ( fabrika 2004 ) .
eclipses by the thick accretion disk and optical star result in obscuring the innermost the hottest parts of the jets from the observer which leads to the drop of the maximum temperature of the registered x - ray emission ( fig.1 ) . in order to find the spectral contribution from the innermost ( the hottest ) regions of the jet we examine differences between spectra taken during eclipses and off eclipses .
the result is presented in the fig.2 .
a strong photoabsorption is observed near the orbital eclipse ( the left panel of fig.2 ) .
the best - fit value of the absorption column density in this spectrum is @xmath8 @xmath9 , much higher than the conventional value .
the spectrum of the hottest ( innermost ) part of the jet as derived from precessional variations off the primary eclipse ( the right panel of fig.2 ) also suggests some photoabsorption , but of considerably smaller value here the best - fit absorbing column density is @xmath10 @xmath9 .
so our investigation shows the presence of the absorbing material in the system .
its density is much higher near the optical star .
the ordinary of this material can appear from powerful winds from the optical supergiant star and super - eddington accretion disk . in order to estimate the possible increase of the star radius due to the dense stellar wind
we take parameters of the wind ( its mass loss rate and velocity distribution ) from the work of achmad et al .
( 1997 ) . in the case of the plasma with solar abundance
it becomes almost opaque for x - rays when @xmath11 .
our estimations show that maximal value of the a supergiant mass loss rate @xmath12 can increase the size of the star up to 20 % .
we describe jets in ss433 as conical plasma flows with constant velocity @xmath13 along the jet axis and constant half - opening angle @xmath14 ( marshall et al . 2002 ) .
radius of the jet cross section is @xmath15 , where @xmath16 is the radius of the jet near the compact object , @xmath17 is the distance from the compact object along jet axis .
we assume that the dominant cooling process of the jet is adiabatic cooling .
so the plasma temperature @xmath18 changes with distance from the compact object @xmath17 as : @xmath19 , where @xmath20 is the plasma temperature and @xmath21 is the jet cross section radius at some distance @xmath22 from the central source ( see also koval & shakura 1989 ) . for disk radius we adopted the disk truncation radius given by paczynski ( 1977 ) .
comparison of the highest visible temperature profiles derived above with observational data enables us to find the model parameters ( the best - fit model is shown in fig.3 ) .
the best - fit parameters of our model referred to the distance from the black hole @xmath23 ( depending on the assumed value of @xmath24 ) are : the highest temperature of the jet @xmath25 kev , the radius of the jet @xmath26 .
in fig.4(left panel ) we present the profile of the maximum x - ray temperature measured in the orbital eclipse near precessional phases @xmath27 .
the orbital x - ray eclipse is clearly visible and shows an appreciable depth .
the ratio of the maximum jet temperature derived from off - eclipse observations to that of the eclipsed jet is @xmath28 .
dashed lines in the figure show examples of eclipsing profiles obtained in our model .
they are not symmetric due to the different jet inclination with respect to the binary orbital plane at this phases .
the ratio of the maximum jet temperatures in the eclipse and off the eclipse is plotted in fig.4 ( right panel ) as a function of the binary mass ratio @xmath24 .
the solid curve shows the ratio obtained under the assumption that the size of the star equals to the volume averaged radius of the roche lobe .
we have shown in section 2 that the size of the eclipsing region might be large then the roche lobe of the optical star .
the innermost parts of dense stellar wind also can absorb x - rays thus increasing the duration and depth of the x - ray eclipse .
so in fig .
4 ( right panel ) we also plot by the dotted and dashed curves the model ratio of maximum jet temperatures in- and out- of the eclipse assuming the star radius to be @xmath29 and @xmath30 correspondingly . from fig.4 ( right panel )
we can conclude that the binary mass ratio in ss433 should not be significantly larger than @xmath31 , because in that case the optical star would be too small to provide the observed depth of the eclipse or the radius of the star ( i.e. the radius of the x - ray eclipser which can exceed the actual star radius ) should be larger than @xmath32 .
using extensive observational data obtained by rxte , we studied systematic variations of x - ray emission of ss433 caused by precessional and orbital motions in the system .
we compared x - ray spectra of ss433 near and in the eclipse and obtained strong signatures of a significant ( @xmath33 @xmath9 ) photoabsorption of x - rays near the companion star .
we argue that this may be due to the presence of a dense stellar wind from the companion star .
the size of the x - ray eclipsing region can be significantly larger than the size of the roche lobe due to absorption in the inner stellar wind .
we recovered the temperature profile of plasma along the jet and estimated the jet cross section at its basement .
the maximum visible temperature is found to be @xmath34 kev at a distance of @xmath35 from the compact object .
the radius of the jet at this distance is @xmath36 .
we reliably detected orbital eclipses in the rxte / pca observations of ss433 during the `` crossover '' precessional phases when the x - ray jets and the axis of the accretion disk lie exactly in the plane of the sky .
the observed depth of the x - ray eclipse implies that the size of the star is larger than @xmath3 , yielding an upper bound on the mass ratio of the components in ss433 @xmath37 assuming that the radius of the eclipser ( star plus inner wind ) can not be much larger than @xmath38 .
research has made use of data obtained from high energy astrophysics science archive research center online service , provided by the nasa / goddard space flight center .
this work has been supported by rfbr grants n04 - 02 - 16349 , n05 - 02 - 19710 , n 06 - 02 - 16025 , n 04 - 02 - 17276 and n 04 - 02 - 16720 , and by joint rfbr / jsps grant n05 - 02 - 12710 . | we present analysis of extensive monitoring of ss433 by the rxte observatory collected over the period 1996 - 2005 .
the difference between energy spectra taken at different precessional and orbital phases shows the presence of a strong photoabsorption near the optical star , probably due to its powerful dense wind . assuming that a precessing accretion disk is thick
, we recover the temperature profile in the x - ray emitting jet that best fits the observed precessional variations of the x - ray emission temperature .
the hottest visible part of the x - ray jet is located at a distance of @xmath0 , or @xmath1 cm from the central compact object and has a temperature of about @xmath2 kev .
we discovered appreciable orbital x - ray eclipses at the `` crossover '' precessional phases ( jets are in the plane of the sky , disk is edge - on ) which put a lower limit on the size of the optical component @xmath3 and an upper limit on a mass ratio of binary companions @xmath4 .
the size of the eclipsing region can be larger than secondary s roche lobe because of substantial photoabsorption by dense stellar wind .
this must be taken into account when evaluating the mass ratio from analysis of x - ray eclipses .
# 1_#1 _ # 1_#1 _ = # 1 1.25 in .125 in .25 in |
this work is motivated by the formation of lipid rafts in lipid membranes .
lipid bilayer membranes are of utmost importance for the survival of cells .
they separate the interior of cells from the extracellular environment and compartmentalize subcellular organelles so suitable micro - environment can be maintained in the enclosed domains for various vital biochemical and biophysical reactions .
they are material basis for morphological changes such as budding , tubulation , fission and fussion that occur during cell division , biological reproduction , and intracellular membrane trafficking .
they also provide a physical platform to store and transduce energy as electrochemical gradients , to segregate or disperse particular membrane proteins , and to act as messengers in signal transduction and molecular recognition processes @xcite . while most of these functionalities depends on the fluidity of the lipids and thereby the free diffusion of lipids and proteins in the bilayer ,
accumulated evidences show that lipids and proteins on bilayer membranes segregates into discrete domains of distinct composition and various sizes @xcite .
the domain boundaries can appear as the barriers of free lateral diffusion of lipids and proteins , as the measured lateral diffusion coefficients of lipids and proteins _ in vivo _ are less than the measured coefficients in artificial pure bilayer by more than one order of magnitude @xcite . inside the domains and on the domain boundaries
particular proteins may aggregate , to cause various membrane curvature as the consequence of the modification of local membrane composition @xcite or to complete specific signal transduction @xcite .
some of these domains are transient , with a duration ranging from seconds to minutes , some can persist for the entire life of the cell , and the domains themselves can diffuse on the membrane surface as well @xcite .
the composition , location , size , configuration , duration of these domains and the dynamics of these characteristics are of functional and structural significance to the associated biological processes .
efforts integrating direct microscopic measurements , biophysical modeling , and computational simulations have been invested to elucidate the underlying physics of the dynamics of these lipid domains and predict their biological consequences @xcite . before introducing our approach based on the geodesic curvature energy of the lipid domain boundaries we first review four most representative theoretical studies on the dynamics of lipid domains .
lipid domains may appear as a result of lipid phase separation caused by distinct spontaneous curvatures .
when bilayer membranes have multiple lipid species of distinct spontaneous curvatures , individual lipid species may be localized to regions where the local mean curvatures best approximate the corresponding spontaneous curvatures of the residing lipid species @xcite .
wang and du formalized this reasoning by summing up the classical canham - evans - helfrich energy @xcite for each individual lipid species and the line tension energy to generate a multi - component lipid membrane model @xcite . by representing the membrane bending energy using the phase field formulation , they have obtained rich patterns of membrane morphology and the generation of lipid membrane domains of different mean curvatures , where lipid species of the approximate spontaneous curvatures are concentrated .
this model was also extended to simulate the open membrane thanks to the line tension energy , and the closing of membrane pores was simulated corresponding to the vanishing linear tension energy .
these permanent domains have sizes that are determined by local mean curvatures of the membrane necks or bumps
. these sizes in general do not match the measured sizes of mobile lipid rafts @xcite .
the classical phase separation model based on the ginzburg - landau ( gl ) free energy could also be directly applied on a membrane surface to generate surface phase separation , and the results can be related to the lipid domains .
a surface cahn - hilliard equation can be derived for the gradient flow of the gl free energy , and the numerical simulations will produce large separated domains as a result of the coarsening dynamics @xcite . in order to generate small domains at spatial and temporal scales comparable to experimental results , camley and brown couples the gl free energy for quasi two - dimensional binary lipids mixtures to the random hydrodynamics and thermal fluctuations @xcite .
the random in - plane velocity field of the membrane is given by saffman - delbruck hydrodynamic model @xcite .
this velocity field is added to the cahn - hilliard equation for the gradient flow of the gl free energy to produce an advection - diffusion equation , above which a gaussian white noise is added , modeling the thermal fluctuation as a random source to the order parameter .
complete phase separation shall occur as the result of a sequence of coarsening dynamics when the gl free energy is minimized , while the domain boundaries flicker as a result of random hydrodynamic and thermal perturbations , with a flickering magnitude depending on the competing between the random perturbation and the persisting linear tension . under high line tension small domains will merge to form large separated domain , but small domains under a critical size could remain separated for a long time during the course of coarsening if the line tension is not large enough to suppress the random perturbation , giving rise to lasting microdomains .
various dynamical scaling rates were summarized to related the microdomain size and the time when the domain size is far way from the saffmann - delbruck length @xmath0 determined by the relative viscosity of the lipid membrane with respect to the surrounding fluid field .
this approach has been recently extended to model multicomponent membranes with embedded proteins @xcite .
it is also possible to simulate lipid microdomains in the deterministic setting . arguing that lipid rafts are microdomains of lipids compactly organized around embedded protein receptors , witkowski , backofen and voigt proposed to supplement the classical ginzburg - landau free energy with gaussian potentials localized at specified positions where the membrane proteins are supposed to be embedded @xcite . by specifying the center and modulation of these external potentials they were able to produce lipid microdomains of arbitrary size at arbitrary position .
coarsening dynamics were reproduced by solving the cahn - hilliard equation for the gradient flow of the total energy , and a scaling law was deduced for the growth of the microdomains .
in contrast to the above approaches , one is not able to drive the lateral diffusion coefficient of the microdomains as their positions are specified in the construction of the free energy . in general this model lacks a biophysical interpretation of the external potential and the related parameters that could justify the striking generation of lipid microdomains in the absence of line tension .
in addition to the above continuum approaches , particle - based discrete methods have also been developed to simulate the lipid microdomains .
molecular dynamics simulations , fully atomic or coarse grained , as reviewed in @xcite , have been able to generate lipid membrane domains that could interpret some experiments on complex model membranes . two discrete methods that gives particular valuable insight into the structure and dynamics of lipid microdomains are dynamical triangulation monte carlo ( dtmc ) @xcite and dispersive particle dynamics ( dpd ) @xcite .
dtmc neglects the solvent hydrodynamics and approximate a bilayer membrane as a randomly triangulated sheet .
each vertex is described by a three - dimensional position vector , and all vertices are connected by flexible tethers , which flip during the course of dynamical triangulation to simulate phase segregation .
dpd adopts a coarse - grained representation of amphiphilic lipids as connected head ( h ) and tail ( c@xmath1 ) beads , with an variable number of tail beads .
the geometric and molecular mechanics representations of lipids in dpd differ from the coarse - grained molecular dynamics simulations ( those based on the martini force field @xcite , for example ) in that all beads are soft with interaction defined by effective forces that reproduce the hydrodynamic behavior of fluid bilayer membrane rather than the classical intermolecular interactions .
dpd allows asymmetric lipid composition in the two leaflets , where different lateral sizes of the lipid domains can be simulated .
while dpd can simulate the membrane properties to length and time scales that are unattainable by md and coarse - grained md simulations , there still exist gaps for these discrete methods to model large systems containing millions of amphiphiles over biologically relevant time scales , for the handling of which the continuum models have intrinsic advantages . here
we work to develop a lipid micro - organization model by recognizing the intrinsic geodesic curvatures of the boundaries between lipids of different species ( saturated and unsaturated , ordered and disordered lipids , too ) .
experimental observations and molecular dynamics simulations show that lipid microdomains have relatively stable size ranging from below 10 nm to larger than 200 nm , depending on the lipid composition and temperature @xcite .
modulo the perturbations to the boundaries of these microdomains caused by random forces due to the thermal fluctuations , hydrodynamic interaction etc , the size of an individual microdomain can be characterized by the curvature of boundary , which is the geodesic curvature since the microdomain is a patch on the three - dimensional surface .
consequently , one can identify an _ intrinsic geodesic curvature _ for a binary mixture of lipid species , similar to the identification of the spontaneous curvature of an individual lipid species .
a geodesic curvature energy can be defined for a given binary mixture of lipid species on a membrane surface , in a way similar to the definition of canham - evans - helfrich curvature energy .
it is expected that the minimization of this geodesic curvature energy will generate the optimal interfaces of the binary mixture of lipids .
the minimization is conducted on the 2-manifold that represents the membrane surface , and these optimal interfaces define the boundaries of the lipid microdomains . since the optimal interfaces are unknown _ a priori _ , and there will be topological changes as the initial interfaces evolve during the course of curvature energy minimization , we adopt an eulerian formulation of the geodesic curvature energy by relating the geodesic curvature to a phase field function on the membrane surface .
we shall show that the phase field representation of the geodesic curvature energy can be regarded as a generalization of the phase field modeling of the elastic bending curvature energy of bilayer membrane deformation @xcite .
the equivalence between the geodesic curvature energy and the phase field representation can be justified when the phase field function is given by a properly scaled hyperbolic tangent function of the signed geodesic distance to the boundaries of lipid microdomains .
the minimizer of the geodesic curvature energy in the eulerian representation is obtained by evolving the gradient flow of the phase field function . the forth - order evolutionary partial differential equation for the gradient flow
are numerically solved by using a @xmath2 interior penalty discontinuous galerkin method on the triangulated membrane surface .
computational simulations using the geodesic curvature energy model on different surfaces demonstrate that our model can produce microdomains with specified intrinsic geodesic curvatures .
the rest of the article is organized as follows . in section [
sect : model ] a brief review of the structure of lipid bilayer membrane is followed by the introduction of geodesic curvature of the interfaces between saturated and unsaturated phases of lipids .
geodesic curvature energy is then defined in lagrangian and eulerian formulations , respectively .
technical details of the justification of some critical properties of geodesic distance and geodesic curvature are given in appendix [ sect : append ] .
the gradient flow of the geodesic curvature energy is derived using the energetical variational principle in section [ sect : numeric ] .
we split the linear and nonlinear components of the governing equation for the gradient flow so that the linear components can be solved using an implicit time marching method and the convergence at each time step is obtained by efficient iterations over the nonlinear components . the proposed mathematical model and the related numerical methods are applied in section [ sect : simulations ] on a set of static surfaces with varying curvatures to examine their ability in simulating the generation of microdomains of specified geodesic curvatures .
we summarize this work and outline its perspectives in the final section [ sect : summary ] .
we consider a bilayer membrane with binary mixture of lipids
. these can be two species of lipids , or lipids in saturated and unsaturated states , or lipids in ordered or disordered states .
these two species of lipids have their own regular molecular geometry at the equilibrium state for a given temperature , modulo random thermal and hydrodynamic fluctuations . by regular we mean the uniform lateral spacing of lipid molecules , the ordered match of hydrophobic tails of two leaflets , among other features .
this regular molecular geometry dictates the intermolecular interactions among lipids of same species . at the interface between domains of distinct lipid species ,
the regular molecular geometry of either species has to be relaxed in a way such that the intermolecular interactions in the transitional region near the interface will fit the different molecular geometry of the other species .
this relaxation shall create curved interface between two species in a manner similar to the creation of the surface tension at a fluid surface .
[ fig : surface_pattern_model ] illustrates the mismatch of lipid molecular structure at the interface between saturated ( s ) and unsaturated ( u ) lipids .
because the lipid domains are to be modelled as patches on membrane surface , the domain boundaies will be curves on a two - dimensional surface embedded in @xmath3 .
notice that if the interface is a geodesic , then it is a locally straight line that does not curve to either domain it separates .
how far an interface is from being a geodesic is characterized by the geodesic curvature , which by definition is the curvature of the interface projected onto the tangent plane of the surface .
the geodesic curvature will be an intrinsic property of _ a binary lipid mixture _
, similar to the spontaneous mean curvature being an intrinsic property of _ an individual lipid species_. we are then motivated to define the curvature energy of the lipid domain boundary to be @xmath4 where @xmath5 is the one - dimensional domain boundary contour on two - dimensional surface embedded in @xmath3 , @xmath6 is the geodesic curvature of the interface , @xmath7 is the spontaneous geodesic curvature of the lipid mixture to be separated , and @xmath8 is the geodesic curvature energy coefficient .
this form of curvature energy is identical to the classical canham - evans - helfrich energy in which the integration in over the two - dimensional membrane surface and the mean curvature is adopted .
the spontaneous geodesic curvature quantifies the geometry and molecular mismatch at the lipid interface . in the transitional region near the interface
two species of lipids coexist in a hybrid state , and a free energy for the hybrid packing of the lipids ( denoted by the subscripts 1 and 2 below ) was proposed in recent theoretical studies @xcite : @xmath9 where @xmath10 is the length of the lipid chains in the transitional region and @xmath11 is the chain length in the equilibrium bulk .
coefficients @xmath12 are the free energetic costs of mismatch between two species and their respective hybrids at the interface , and @xmath13 is the energetic cost of mismatch hybrid chains of different species .
we choose @xmath14 for the purpose of simplification , but the generalization to distinct @xmath8 s can be made easily .
the domain curvature can be related to the lipid geometrical properties by the following relation @xcite : @xmath15 where @xmath16 is the volume of the lipid chains , @xmath17 is the length that characterizes the molecular spacing of the lipid head groups , and @xmath18 is the head group spacing in the hybrid region near the interface . here
the subtraction sign is chosen if the species is included in the microdomain , otherwise the addition sign shall be chosen . the chain length in the equilibrium bulk state , @xmath11 , can be computed by dividing the molecular volume using the head group area in the bulk state @xmath19 eqs.([eqn : vi]-[eqn : li0 ] ) allow us to represent the bending energy @xmath20 in eq.([eqn : bending_energy ] ) as a quadratic function of the geodesic curvature @xmath6 .
the minimizer @xmath7 can be analytically calculated , and can be approximated to the linear order of @xmath21 and @xmath22 by @xmath23,\ ] ] where @xmath24 and @xmath25 .
this approximate minimizer @xmath7 will appear in a more complicated form if @xmath26 , and calculations will show that @xmath7 is not sensitive to the value of @xmath27 in its biologically relevant range @xcite .
the minimizer @xmath7 is identified as the _ spontaneous geodesic curvature _ of the corresponding binary lipid mixture . for a given binary lipid mixture in bilayer membrane
, one can work to minimize the energy in eq.([eqn : geocurv_eng ] ) to get the optimal boundary contour of surface lipid microdomains .
this could be done , for example , by computing the shape derivative following the general analytical approach outlined in @xcite and using this shape derivative as the boundary force to drive the motion of the initial boundary contour .
practical application of this approach , analytically or numerically , might be hindered by the topological changes of the boundary contour , during which there are moments that the surface is not smooth and its principal curvatures are not well defined .
one can resort to phase field or level set methods with which the interface can be traced by evolving a function in a higher dimensional space .
here we adopt the phase field approach , for which we define a phase field function @xmath28 associated with the boundary contour @xmath5 on the membrane surface @xmath29 as @xmath30 here @xmath31 is the signed geodesic distance from @xmath32 to the contour @xmath5 , @xmath33 in the interested domain enclosed by @xmath5 , @xmath34 on @xmath5 , and @xmath35 outside .
correspondingly , @xmath36 and @xmath37 inside and outside , respectively , with the width of the transitional layer near @xmath5 determined by small @xmath38 .
this function @xmath31 can be regarded as the generalization of the signed distance function used in euclidean space @xmath39 to the surface ; some of the critical critical properties of the latter can be shown true for @xmath31 . in appendix
[ sect : append ] we provide details of relating interface normal and geodesic curvature to this signed geodesic distance . with this definition of @xmath28
we can compute the geodesic curvature @xmath6 of its arbitrary level set , including @xmath5 where @xmath40 . to facilitate the computation we first define @xmath41 which gives @xmath42 , ~ q''(x ) = -\frac{1 } { \epsilon } \tanh \left ( \frac{x}{\sqrt{2}\epsilon } \right ) \operatorname{sech}^2 \left ( \frac{x}{\sqrt{2}\epsilon } \right ) .\ ] ] then @xmath43 where @xmath44 and @xmath45 are surface gradient , surface divergence , and laplace - beltrami operators on @xmath29 , respectively .
the proof of the last equality ( [ eqn : lap_phi ] ) can be found in appendix [ sect : append ] as well , where we also proved that the tangent normal vector field for the level sets of @xmath31 is @xmath46 and the geodesic curvature @xmath47 .
we then compute @xmath48 it follows that the geodesic curvature is @xmath49 replacing @xmath50 with @xmath28 and noticing that @xmath51 ( to be proved in appendix [ sect : append ] ) , we finally get the phase field representation of the geodesic curvature : @xmath52 correspondingly , we define a phase field representation of the geodesic curvature energy on the entire surface @xmath29 , as follows : @xmath53 where @xmath54 . for the phase field function @xmath28
defined in eq.([eqn : phi_def ] ) one can show that this phase field curvature energy formulation is equivalent to eq.([eqn : geocurv_eng ] ) .
this nature is analogous to the equivalence between the canham - helfrich - evans curvature energy and the phase field representation of the membrane elastic energy @xcite .
the proof of the equivalence proceeds rather similarly , and is omitted here .
our geodesic curvature energy model of surface microdomain formation can be regarded as a generalization of the mean - curvature driven membrane bending model from three - dimensional euclidean space to two - dimensional surface . and the other way around , the mean - curvature driven membrane bending model can be viewed as as a special case of the curvature energy driven model on manifolds to euclidean spaces .
we expect our phase field based geodesic curvature model can be applied to other physical systems that occur on manifolds where the formation and the topological change of surface patterns are driven by appropriate curvatures .
these may include the change of spacetime topology curved by matter and energy in general relativity @xcite .
we follow the gradient flow of the curvature energy ( [ eqn : g_phase_field ] ) to find its minimizer @xmath28 whose zero level set shall give the the optimal lipid microdomain boundary . to derive the governing equation for the gradient flow we compute the first variation of @xmath55 with respect to @xmath28 : @xmath56 , \label{eqn : var_g}\ ] ] where @xmath57 the full expansion of this variation reads @xmath58 during the energy minimization the amounts of each species of lipids shall be conserved ; for this reason we consider the following constraint @xmath59 whose first variation with respect to @xmath28 is @xmath60 we then introduce the following governing equation for the gradient flow of @xmath28 @xmath61 where @xmath62 is the pseudo - time , @xmath13 is the diffusion coeffcient , and @xmath63 is a lagrangian multiplier used to ensure the conservation of @xmath28 . the representation of @xmath63 can be derived by integrating eq .
( [ eqn : gradient_flow_surface_pattern ] ) and noting that @xmath64 , thus @xmath65 and consequently , @xmath66 which yields a forth - order nonlinear surface diffusion equation . @xmath67 alternatively , one could derive a cahn - hilliard equation for the surface phase field function @xmath28 as @xmath68 which guarantees the conservation of @xmath28 and thus does not need a lagrangian multiplier .
however , it involves a sixth order surface derivative and thus is more complicated when the equation is to be solved numerically on a discretized surface @xmath29 . to begin the time discretization of eq.([eqn : gradient_flow_surface_pattern_2 ] )
we first average the function @xmath69 over the current and the next steps @xmath70 and @xmath71 , upon which the crank - nicolson scheme is applied : @xmath72 where the averaged function is defined by @xmath73 with @xmath74 to achieve convergence of this nonlinear implicit scheme at each time step , we define an inner iteration for computing @xmath75 . the solution during these inner iterations are denoted by @xmath76 , which is expected to converges to @xmath75 as @xmath77 .
we then replace @xmath75 in eq.([eqn : scheme_1 ] ) with @xmath78 and @xmath79 to get @xmath80 where the new average function is defined by @xmath81 with @xmath82 to efficiently solve the nonlinear eq.([eqn : scheme_2 ] ) we split the average functions into linear and nonlinear components , with the linear component being the function of @xmath79 only and the nonlinear component being the function of @xmath83 and @xmath78 , as follows : @xmath84 the nonlinear implicit scheme is now given by @xmath85 which , after all terms of @xmath79 collected to the left - hand side , turns out to be @xmath86 \delta t + \phi_n \nonumber \\
= & \frac{1}{2 } \big ( \lambda(\psi_{m } ) + \lambda(\phi_n ) \big ) \delta t - \frac{\delta \gamma k \epsilon}{2 } \delta_s^2 \phi_n + \nonumber \\ & \delta t \gamma \delta_s \left ( \frac{k}{4 \epsilon } ( \psi_m^2 + \phi_n^2 -2)(\psi_m + \phi_n + 2 h_c \epsilon ) \right ) + \nonumber \\ & \delta t \gamma \delta_s \left ( \frac{k}{2\epsilon^2 } ( \psi_m^2 + \psi_m \phi_n + \phi_n^2 + h_c \epsilon(\psi_m + \phi_n ) \right ) \cdot \nonumber \\ & \left ( f(\psi_m ) + f(\phi_n )
\right ) + \phi_n.\end{aligned}\ ] ] this last equation is iterated over the inner index @xmath87 until the convergence is reached as signified by @xmath88 for some desired tolerance .
@xmath78 is initialized as @xmath83 for inner iterations at each time step .
the convergent @xmath89 will be passed to @xmath75 to advance the computation to the next time step .
we adopt a @xmath90 interior penalty discontinuous galerkin ( dg ) method to solve the forth - order linear equation ( [ eqn : scheme_4 ] ) for @xmath79 .
this method was initially developed for solve forth - order equations in two - dimensional euclidean space @xcite . when it is applied to solve equations on triangulated two - dimensional surface ,
an additional term arises on all edges , representing the mismatch of fluxes due to the different outer normal directions of two neighbouring triangles at the shared edge .
the order of convergence of the surface dg method with respect to the mesh refinement matches that observed on the original two - dimensional numerical experiments . a large time step is allowed in the computational simulated presented below owning to the implicit time discretization .
the inner iterations in general converge within five steps .
the solutions of a surface cahn - hilliard equation in @xcite is based on the reduction to two second - order equations , one for the order parameter and the other for the chemical potential .
the equations are then solved using standard surface finite element method along with an explicit time discretization , the latter suffers from a strict discrete energy stability condition and only very small time increments are allowed . an even larger time step might be chosen by applying a recently developed exponential time differencing method @xcite . with this method
one needs to deliberately estimate the nonlinear component of @xmath91 at every step so its slowly varying part can be combined with the linear component of @xmath91 , possibly further improving the discrete energy stability of the scheme .
in this section we shall apply the geodesic curvature energy variational model to simulate the lipid microdomain formation on three surfaces with different intrinsic geodesic curvatures .
the dynamics of the microdomain formation in each set of simulations will be compared to those generated by the surface cahn - hilliard equation @xcite , the latter will produce large surface domain as a result of minimizing the arc length of the domain boundaries .
we first consider a unit sphere on which a trigular mesh with @xmath92 approximately uniformly distributed nodes .
we choose @xmath93 , a time increment @xmath94 , and a random field of @xmath28 .
the results are compared side by side with those of the classical surface cahn - hilliard equation in fig .
[ fig:1atm_3963 ] . using a x - means clustering method
@xcite we are able to identify a number of microdomains whose radii are then calculated .
the radius associated with each microdomain is approximately @xmath95 .
this means the curvature is approximately @xmath96 , close to the specified spontaneous geodesic curvature .
nodes show the formation of local microdomains predicted by the geodesic curvature energy ( left column ) and domain separation predicted by the allen - cahn equation associated with the classical ginzburg - landau energy ( right column ) from the same initial random field ( top row ) .
sampling time from top to bottom : @xmath97 and @xmath98.,title="fig:",width=143 ] nodes show the formation of local microdomains predicted by the geodesic curvature energy ( left column ) and domain separation predicted by the allen - cahn equation associated with the classical ginzburg - landau energy ( right column ) from the same initial random field ( top row ) .
sampling time from top to bottom : @xmath97 and @xmath98.,title="fig:",width=143 ] + nodes show the formation of local microdomains predicted by the geodesic curvature energy ( left column ) and domain separation predicted by the allen - cahn equation associated with the classical ginzburg - landau energy ( right column ) from the same initial random field ( top row ) .
sampling time from top to bottom : @xmath97 and @xmath98.,title="fig:",width=143 ] nodes show the formation of local microdomains predicted by the geodesic curvature energy ( left column ) and domain separation predicted by the allen - cahn equation associated with the classical ginzburg - landau energy ( right column ) from the same initial random field ( top row ) .
sampling time from top to bottom : @xmath97 and @xmath98.,title="fig:",width=143 ] + nodes show the formation of local microdomains predicted by the geodesic curvature energy ( left column ) and domain separation predicted by the allen - cahn equation associated with the classical ginzburg - landau energy ( right column ) from the same initial random field ( top row ) .
sampling time from top to bottom : @xmath97 and @xmath98.,title="fig:",width=143 ] nodes show the formation of local microdomains predicted by the geodesic curvature energy ( left column ) and domain separation predicted by the allen - cahn equation associated with the classical ginzburg - landau energy ( right column ) from the same initial random field ( top row ) .
sampling time from top to bottom : @xmath97 and @xmath98.,title="fig:",width=143 ] + nodes show the formation of local microdomains predicted by the geodesic curvature energy ( left column ) and domain separation predicted by the allen - cahn equation associated with the classical ginzburg - landau energy ( right column ) from the same initial random field ( top row ) .
sampling time from top to bottom : @xmath97 and @xmath98.,title="fig:",width=143 ] nodes show the formation of local microdomains predicted by the geodesic curvature energy ( left column ) and domain separation predicted by the allen - cahn equation associated with the classical ginzburg - landau energy ( right column ) from the same initial random field ( top row ) .
sampling time from top to bottom : @xmath97 and @xmath98.,title="fig:",width=143 ] the second set of simulations is conducted on a more complicated surface as shown in fig .
[ fig:3atm ] .
we choose this surface to be the molecular surface of three particles of unit radius respectively centered at @xmath99 and @xmath100 .
such molecular surfaces are obtained by tracing the centers of spherical probe of the radius of the water molecule ( @xmath101 , we are using dimensionless length is this work , though ) as the probe is rolling over the surfaces of the molecules @xcite .
this surface is quasi - uniformly meshed with @xmath102 nodes and we set @xmath103 and @xmath94 . starting with a random initial field we finally identified six microdomains using the x - mean clustering method at the equilibrium state , whose radii are estimated . as seen in fig .
[ fig : domain_radii](left ) , the radii of the microdomains approximate the given spontaneous geodesic curvatures . and @xmath98.,title="fig:",width=143 ] and @xmath98.,title="fig:",width=143 ] + and @xmath98.,title="fig:",width=143 ] and @xmath98.,title="fig:",width=143 ] + and @xmath98.,title="fig:",width=143 ] and @xmath98.,title="fig:",width=143 ] + and @xmath98.,title="fig:",width=143 ] and @xmath98.,title="fig:",width=143 ] in the last set of simulations we choose the molecular surface of six particles of unit radius respectively centered at @xmath104 and @xmath105 .
the quai - uniform surface mesh has @xmath106 nodes and we choose @xmath107 and @xmath94 for the simulation .
the dynamics of the microdomain formation is presented in fig .
[ fig:6atm ] along with that of the domain separation described by the surface cahn - hilliard equation .
one can see from fig .
[ fig : domain_radii](right ) that the largest raft radius obtained by the simulation is about @xmath108 which means the curvature of that raft is about @xmath109 , a value close to given spontaneous geodesic curvature . and @xmath110.,title="fig:",width=143 ] and @xmath110.,title="fig:",width=143 ] + and @xmath110.,title="fig:",width=143 ] and @xmath110.,title="fig:",width=143 ] + and @xmath110.,title="fig:",width=143 ] and @xmath110.,title="fig:",width=143 ] + and @xmath110.,title="fig:",width=143 ] and @xmath110.,title="fig:",width=143 ] the radii of the microdomains generated in our simulations are not exactly the given spontaneous geodesic curvature .
rather they are distributed around the given curvature .
apart from the numerical error in simulation and in x - mean clustering and radii estimate , this non - uniform distribution of domain radii is mostly related to the total quantity of the lipid phases in the initial random field .
the initial quantity may not be exact to cover an integer number of microdomains with the given radius .
however , the overall distribution of radii around the given radius of curvature demonstrates that our geodesic curvature model is capable of predicting the formation of microdomains that are caused by the geometrical and molecular mechanical mismatch of lipid mixtures .
the predicted microdomains can be compared to the observed lipid rafts , and the boundaries of these microdomains can be identified to provide locations where specific proteins can aggregate .
coupling of our model of geodesic curvature driven microdomains formation to the localization of proteins will provide a very useful quantitative technique for studying the crucial roles of these proteins in high - fidelity signal transmission in cells @xcite .
in this work we investigated formation of microdomains on lipid membrane surfaces .
we examine the geometrical and mechanical mismatch of lipids of different species on domain boundaries and introduce the geodesic curvature of these boundaries in the energy characterization of the equilibrium state .
this gives rise to the geodesic curvature energy model of the surface microdomain formation .
a phase field representation of the model is derived to ease the computational simulations of the model .
we develop a @xmath90 interior penalty surface finite element method along with an explicit / implicit time splitting scheme for solving the forth order nonlinear partial differential equation on surfaces .
numerical simulations on different surfaces demonstrate that our model and numerical techniques are able to produce the circular raft - like microdomains we are interested in .
the generated microdomains have the radii corresponding to the given spontaneous geodesic curvature for the binary lipid mixture .
there are accumulated evidences that lipid rafts aid in vesicle budding and the deformation of the membrane .
currently many models are being produced to model vesicle budding and membrane deformation .
this opens up a whole research area of connecting predicted microdomains using the developed geodesic curvature model to other biophysical processes .
this work has been partially supported by the simons foundation and national institutes of health through the grant r01gm117593 as part of the joint dms / nigms initiative to support research at the interface of the biological and mathematical sciences .
this section is devoted to the justification of a number of surface calculus identities used in the phase field formulation in this article .
although the counterparts of these identities in euclidean spaces are well known , a rigorous derivation of these surface identities , to our knowledge , does not exist .
we believe the justification of these identities is useful to scientists working on the various problems involving surface patterns and surface energies . since the phase field function @xmath111 is an algebraic function of the signed geodesic function @xmath31 , it is sufficient to work with @xmath31 in the derivation below
. let @xmath29 be an oriented smooth surface without boundary in three - dimensional euclidean space @xmath3 and let @xmath112 be a curve in @xmath29 , parameterized by the arc length @xmath113 .
the darboux frame of the curve @xmath112 is defined by @xmath114 where @xmath115 are respectively the unit tangent vector , the unit normal vector , and the unit tangent normal vector .
let @xmath31 be the signed geodesic function from the curve @xmath112 , following the convention that @xmath33 in the surface region of interests enclosed by @xmath112 , zero on @xmath112 , and positive outside .
one can define the similar darboux frame on an arbitrary level set of @xmath31 on @xmath29 , giving rise to three vector fields @xmath116 and @xmath117 on @xmath29 .
we shall have the tangent normal vector field @xmath117 on @xmath29 is the normalized surface gradient of the signed geodesic distance function @xmath31 : @xmath118 let @xmath119 be an arbitrary point on @xmath29 and let @xmath120 be the level set of @xmath31 passing through @xmath119 for a constant @xmath121 , parameterized by the arc length @xmath113 .
associated with @xmath112 at @xmath119 are the three vectors , @xmath122 defined by the darboux frame .
let @xmath123 be the two - dimensional orthogonal coordinate of the tangent plane of the surface @xmath29 originated at @xmath119 , with @xmath124 aligned with @xmath125 .
furthermore , we consider a local monge parameterization of @xmath29 around @xmath119 , with the coordinate variables @xmath126 .
the linear map @xmath127 from @xmath128 to @xmath123 is such that @xmath129 the signed geodesic distance function @xmath31 can be given locally by either @xmath130 or @xmath131 .
the specific level set @xmath132 can be locally given by @xmath133 in the @xmath128 plane .
it follows that @xmath134 using the mapping @xmath127 we will have at point @xmath119 @xmath135 suggesting that @xmath136 is orthogonal to the tangent vector @xmath125 of @xmath137 at @xmath119 . since @xmath136 is on the tangent plane it is orthogonal to surface with the normal vector @xmath138 , and hence it must be in the direction of the tangent normal vector @xmath139 .
finally we check @xmath140 .
since @xmath132 is a level set and @xmath124 is aligned with the tangent direction of this level set it follows that @xmath141 . now that @xmath142 is in the direction of the tangent norm , and given that @xmath132 is a level set of the signed geodesic distance function , @xmath142 is indeed in the direction of the geodesic passing through @xmath119 , thus @xmath143 .
this suggests that @xmath144 and the theorem follows .
after the establishment of the tangent normal vectors field @xmath117 on @xmath29 , we can compute the geodesic curvature of any level set of the signed geodesic distance function @xmath31 by using the surface divergence of @xmath117 .
let @xmath112 be a level set @xmath145 , parameterized by the arc length @xmath113 .
one can associate a frenet - serret frame to @xmath112 , defined by @xmath146 where @xmath147 are respectively the unit tangent vector , the frenet normal vector , and the frenet binormal vector .
the frenet normal vector is also called the principal normal vector .
the curvature @xmath8 of the curve @xmath112 is @xmath148 the plane spanned by the orthogonal vectors @xmath149 is called osculating plane .
let @xmath150 be the angle between the osculating plane and the tangent plane .
the geodesic curvature is defined to be the projection of the curvature to the tangent plane , i.e. , @xmath151 it can be shown that the geodesic curvature is indeed the ordinary curvature of the curve obtained by projecting @xmath112 on to the tangent plane ( p.261 , @xcite ) .
this last fact will be used to prove the following theorem : the geodesic curvature @xmath152 of the space curve @xmath112 given by the level set @xmath153 of the signed geodesic distance function equals to the surface divergence of the tangent normal vector field @xmath117 : @xmath154 let @xmath123 be the orthogonal coordinate of a local monge parameterization of the surface @xmath29 near the point @xmath155 such that @xmath156 are respectively in the directions of the tangent and tangent normal vectors @xmath157 at @xmath155 , and @xmath155 itself is mapped to the origin of the @xmath123 plane .
let the height of the surface be locally @xmath158 . under this parameterization the space curve @xmath112 will be mapped to a curve on @xmath123 plane , denoted by @xmath159 .
the space curve @xmath112 can be parameterized by @xmath124 only by setting @xmath160 accordingly , the tangent vector @xmath161 can be given as a function of @xmath124 by @xmath162 since this tangent vector is identical to the element vector @xmath163 at @xmath164 by the choice of the monge parameterization , it follows that at @xmath155 @xmath165 we compute the geodesic curvature of the surface curve @xmath112 at @xmath155 by evaluating the curvature of the plane curve @xmath166 at @xmath164 : @xmath167 where the negative sign is assigned so the orientation of the plane curve gives outer - pointing normal vector that is consistent with the orientation of @xmath123 coordinate .
it remains to show that @xmath168 .
we represent the tangent normal vector field @xmath169 using the local monge parameterization . for any point @xmath170 on the surface curve @xmath112
we have @xmath171 for some @xmath124 and then @xmath172 where @xmath173 to compute surface divergence we notice that @xmath174 where @xmath175 are the components of the vector @xmath117 respectively in @xmath176 directions .
on one hand we shall have @xmath177 at @xmath155 where @xmath178 for that @xmath179 achieves its maximum value as a component of a unit normal vector owning to @xmath180 at @xmath155 .
on the other hand , @xmath181 - \\ & \frac{f_{\xi}}{f^2 } \big [ 2(h_{\xi \xi } + h_{\xi \eta } f'(\xi ) ) h_{\xi } + 2 ( h_{\xi \eta } + h_{\eta \eta } f'(\xi ) ) h_{\eta } \big ] \big [ ( h_{\eta}h_{\xi } + h^2_{\eta } f'(\xi ) ) - f'(\xi ) \big ] \end{aligned}\ ] ] it follows again from @xmath180 that @xmath182 and @xmath183 as requested . the last fact to justify is the identity eq.([eqn : lap_phi ] ) . recalling that for a scalar @xmath184 and a vector field @xmath185 on the surface @xmath29 with outer normal vector @xmath186 it holds true that @xmath187 \\ & = a ~ n \cdot \curl ( n \times v ) + n \cdot \big [ \nabla a \times ( n \times v ) \big ] \\ & = a \nabla_s v + n \cdot \big [ ( \nabla a \cdot v ) n - ( \nabla a \cdot n ) v \big ] , \end{aligned}\ ] ] we shall have @xmath188
99 g. van meer , d. r. voelker , g. w. feigenson , membrane lipids : where they are and how they behave . ,
nat . rev .
cell biol . 9
( 2008 ) 112124 .
[ ] s. l. veatch , s. l. keller , separation of liquid phases in giant vesicles of ternary mixtures of phospholipids and cholesterol , biophys .
j. 85 ( 5 ) ( 2003 ) 3074 3083 .
http://dx.doi.org/http://dx.doi.org/10.1016/s0006-3495(03)74726-2 [ ] .
a. kusumi , c. nakada , k. ritchie , k. murase , k. suzuki , h. murakoshi , r. s. kasai , j. kondo , t. fujiwara , paradigm shift of the plasma membrane concept from the two - dimensional continuum fluid to the partitioned fluid : high - speed single - molecule tracking of membrane molecules , annu .
34 ( 2005 ) 351378 . http://dx.doi.org/10.1146/annurev.biophys.34.040204.144637 [ ] .
f. s. vieira , g. correa , m. einicker - lamas , r. coutinho - silva , host - cell lipid rafts : a safe door for micro - organisms ?
, biology of the cell 102 ( 7 ) ( 2010 ) 391407 . http://dx.doi.org/10.1042/bc20090138 [ ] .
t. wang , s. d. cady , m. hong , nmr determination of protein partitioning into membrane domains with different curvatures and application to the influenza m2 peptide , biophys .
j. 102 ( 4 ) ( 2012 ) 787 794 .
http://dx.doi.org/http://dx.doi.org/10.1016/j.bpj.2012.01.010 [ ] .
w. b. huttner , j. zimmerberg , implications of lipid microdomains for membrane curvature , budding and fission : commentary , current opinion in cell biology 13 ( 2001 ) 478484 .
http://dx.doi.org/10.1016/s0955-0674(00)00239-8 [ ] .
l. d. renner , d. b. weibel , cardiolipin microdomains localize to negatively curved regions of escherichia coli membranes , p. natl .
usa 108 ( 15 ) ( 2011 ) 62646269 .
ryu , d. yoo , n. j. wittenberg , l. r. jordan , s .- d .
lee , a. n. parikh , s .- h .
oh , lipid membrane deformation accompanied by disk - to - ring shape transition of cholesterol - rich domains , j. am .
137 ( 27 ) ( 2015 ) 86928695 . http://dx.doi.org/10.1021/jacs.5b04559 [ ] .
a. hinderliter , p. f. f. almeida , c. e. creutz , r. l. biltonen , domain formation in a fluid mixed lipid bilayer modulated through binding of the c2 protein motif , biochemistry 40 ( 13 ) ( 2001 ) 41814191 , pmid : 11300799 . http://dx.doi.org/10.1021/bi0024299 [ ] .
t. yeung , g. e. gilbert , j. shi , j. silvius , a. kapus , s. grinstein , membrane phosphatidylserine regulates surface charge and protein localization , science 319 ( 5860 ) ( 2008 ) 210213 .
[ ] .
d. v. nicolau , k. burrage , r. g. parton , j. f. hancock , identifying optimal lipid raft characteristics required to promote nanoscale protein - protein interactions on the plasma membrane , mol .
cell biol .
26 ( 2006 ) 313323 . http://dx.doi.org/10.1128/mcb.26.1.313-323.200 [ ] .
x. wang , q. du , modelling and simulations of multi - component lipid membranes and open membranes via diffuse interface approaches , j. math .
( 2008 ) 347371 .
http://dx.doi.org/10.1007/s00285-007-0118-2 [ ] .
c. yuan , j. furlong , p. burgos , l. j. johnston , the size of lipid rafts : an atomic force microscopy study of ganglioside gm1 domains in sphingomyelin / dopc / cholesterol membranes , biophys .
j. 82 ( 2002 ) 25262535 . http://dx.doi.org/10.1016/s0006-3495(02)75596-3 [ ] .
q. du , l. ju , l. tian , finite element approximation of the cahn - hilliard equation on surfaces , comput .
methods appl .
200 ( 29?32 ) ( 2011 ) 2458 2470 . http://dx.doi.org/http://dx.doi.org/10.1016/j.cma.2011.04.018 [ ] .
b. a. camley , f. l. h. brown , dynamic simulations of multicomponent lipid membranes over long length and time scales , phys .
( 2010 ) 148102 .
http://dx.doi.org/10.1103/physrevlett.105.148102 [ ] .
t. witkowski , r. backofen , a. voigt , the influence of membrane bound proteins on phase separation and coarsening in cell membranes , phys .
( 2012 ) 1450914515 .
http://dx.doi.org/10.1039/c2cp41274h [ ] .
j. c. shillcock , r. lipowsky , equilibrium structure and lateral stress distribution of amphiphilic bilayers from dissipative particle dynamics simulations , j. chem .
phys . 117 ( 10 ) ( 2002 ) 50485061 . http://dx.doi.org/http://dx.doi.org/10.1063/1.1498463 [ ] .
h. li , a. a. gorfe , aggregation of lipid - anchored full - length h - ras in lipid bilayers : simulations with the martini force field , plos one 8 ( 7 ) ( 2013 ) e71018 .
q. du , c. liu , r. ryham , x. wang , modeling the spontaneous curvature effects in static cell membrane deformations by a phase field formulation , comm .
anal . 4 ( 2005 )
http://dx.doi.org/10.3934/cpaa.2005.4.537 [ ] .
q. du , c. liu , x. wang , simulating the deformation of vesicle membranes under elastic bending energy in three dimensions , j. comput .
212 ( 2 ) ( 2006 ) 757 777 .
http://dx.doi.org/doi : 10.1016/j.jcp.2005.07.020 [ ] .
r. brewster , s. a. safran , line active hybrid lipids determine domain size in phase separation of saturated and unsaturated lipids , biophys .
j. 98 ( 6 ) ( 2010 ) l21l23 .
http://dx.doi.org/http://dx.doi.org/10.1016/j.bpj.2009.11.027 [ ] .
s. c. brenner , s. gu , t. gudi , l. yeng sung , a quadratic c0 interior penalty method for linear fourth order boundary value problems with boundary conditions of the cahn - hilliard type , siam j. numer .
50 ( 4 ) ( 2012 ) 20882110 .
http://arxiv.org/abs/http://dx.doi.org/10.1137/110847469 [ ] , http://dx.doi.org/10.1137/110847469 [ ] .
x. wang , l. ju , q. du , efficient and stable exponential time differencing runge - kutta methods for phase field elastic bending energy models , j. comput .
( 2016 ) 21 39 .
http://dx.doi.org/http://dx.doi.org/10.1016/j.jcp.2016.04.004 [ ] .
d. pelleg , a. w. moore , x - means : extending k - means with efficient estimation of the number of clusters , in : proceedings of the seventeenth international conference on machine learning , icml 00 , morgan kaufmann publishers inc . , san francisco , ca , usa , 2000 , pp .
727734 .
h. j. hocker , k .- j .
cho , c .- y .
k. chen , n. rambahal , s. r. sagineedu , k. shaari , j. stanslas , j. f. hancock , a. a. gorfe , andrographolide derivatives inhibit guanine nucleotide exchange and abrogate oncogenic ras function , p. natl .
usa 110 ( 25 ) ( 2013 ) 1020110206 . http://arxiv.org/abs/http://www.pnas.org/content/110/25/10201.full.pdf [ ] , http://dx.doi.org/10.1073/pnas.1300016110 [ ] . | lipid bilayer membranes are not uniform and clusters of lipids in a more ordered state exist within the generally disorder lipid milieu of the membrane . these clusters of ordered lipids microdomains are now referred to as lipid rafts .
recent reports attribute the formation of these microdomains to the geometrical and molecular mechanical mismatch of lipids of different species on the boundary . here
we introduce the geodesic curvature to characterize the geometry of the domain boundary , and develop a geodesic curvature energy model to describe the formation of these microdomains as a result of energy minimization .
our model accepts the intrinsic geodesic curvature of any binary lipid mixture as an input , and will produce microdomains of the given geodesic curvature as demonstrated by three sets of numerical simulations .
our results are in contrast to the surface phase separation predicted by the classical surface cahn - hilliard equation , which tends to generate large domains as a result of the minimizing line tension .
our model provides a direct and quantified description of the structure inhomogeneity of lipid bilayer membrane , and can be coupled to the investigations of biological processes on membranes for which such inhomogeneity plays essential roles .
lipid bilayer membrane ; lipid domain ; geodesic curvature ; surface phase separation ; phase field method ; variational principle ; numerical simulations 53b10 , 65d18 , 92c15 |
winds of many early - type stars show cyclic , or even strictly periodic , variability on a rotational timescale .
this behaviour was extensively studied in the uv with the iue satellite .
both pulsations and magnetic fields have been suggested as the cause for this variability .
it is clear that kg magnetic fields as observed in the bp stars can explain the observed , strictly periodic , variability . for other early - type stars where
no such fields have been detected and which can have much denser winds , the origin of the variability is still unknown .
recently , magnetic fields have been discovered in several early b type stars , with @xmath0 cep among them .
these stars have weaker fields ( a few hundred gauss * ? ? ?
* ; * ? ? ?
* ) and stronger winds than the magnetically strong bp stars , which are mostly of a later spectral type .
using numerical simulations , we investigate the effects of the magnetic field on the stellar wind of @xmath0 cep .
our method consists of two steps .
first we compute the wind geometry in the presence of a magnetic field , using a mhd code that starts out from the relaxed non - magnetic wind solution and a dipole field and is evolved in time until quasi - steady behaviour is observed .
second , from the end state of our simulations we calculate line profiles using a sei ( sobolev with exact integration ) method .
stellar rotation is simulated by adjusting the line - of - sight of the observer .
+ ll|ll spectral type & b1ive & polar field & 360(30 ) g + mass & 12 m@xmath1 & rotation period & 12.00 days + radius & 6.5 r@xmath1 & inclination & 60@xmath290@xmath2 + temperature & @xmath326000
k & @xmath0 & 90@xmath2 + the * 2d - mhd * code ( see * ? ? ?
* ) is based on zeus .
it uses the standard mhd equations , the energy equation with a cooling term appropriate for stellar winds , and radial cak - based radiative driving . since @xmath0 cep is a slow rotator ,
rotation can be ignored . for our models of @xmath0 cep
, we use basic stellar parameters as listed in table [ bcep_pars ] .
we treat mass loss rate , @xmath4 as a parameter that ranges from @xmath3@xmath5@xmath6 .
the line profiles are calculated using the * sei * method .
opacities are determined using solar abundances and a ( recently included ) lte ionisation balance . at low temperatures ( @xmath7 ) the civ density
is increased with a few percent due to the auger effect .
the highest relative civ densities are found in shock heated sheets of @xmath3@xmath8 k. our calculations do not account for the combined effect of the doublet members .
+ m@xmath1 yr@xmath9 .
one can see the outflowing disk " , in the magnetic equator , which has higher temperatures , higher densities and lower velocities than its surroundings.,title="fig:",scaledwidth=32.0% ] m@xmath1 yr@xmath9 .
one can see the outflowing disk " , in the magnetic equator , which has higher temperatures , higher densities and lower velocities than its surroundings.,title="fig:",scaledwidth=32.0% ] + m@xmath1 yr@xmath9 .
one can see the outflowing disk " , in the magnetic equator , which has higher temperatures , higher densities and lower velocities than its surroundings.,title="fig:",scaledwidth=32.0% ] m@xmath1 yr@xmath9 .
one can see the outflowing disk " , in the magnetic equator , which has higher temperatures , higher densities and lower velocities than its surroundings.,title="fig:",scaledwidth=32.0% ] cep .
the top panel shows the significance of the variability . *
( middle ) * results for a model with b@xmath10=190 g and a fixed civ abundance .
the basic features are reproduced , but fine - tuning of the parameters is needed . *
( right ) * results for b@xmath10=360 g , including a temperature dependent civ abundance .
note the 0.25 offset in phase relative to the observations.,title="fig:",scaledwidth=32.5% ] cep .
the top panel shows the significance of the variability . *
( middle ) * results for a model with b@xmath10=190 g and a fixed civ abundance .
the basic features are reproduced , but fine - tuning of the parameters is needed . *
( right ) * results for b@xmath10=360 g , including a temperature dependent civ abundance .
note the 0.25 offset in phase relative to the observations.,title="fig:",scaledwidth=32.5% ] cep .
the top panel shows the significance of the variability . *
( middle ) * results for a model with b@xmath10=190 g and a fixed civ abundance .
the basic features are reproduced , but fine - tuning of the parameters is needed . *
( right ) * results for b@xmath10=360 g , including a temperature dependent civ abundance .
note the 0.25 offset in phase relative to the observations.,title="fig:",scaledwidth=32.5% ]
for a model with a polar field ( @xmath11 ) of 360 g and mass loss rate of @xmath310@xmath12 m@xmath1 yr@xmath9 , fig.[snapshots ] shows the density , temperature , and velocity at a typical time snapshot long after any influence of the initial state has faded .
the material is guided by the magnetic field towards the magnetic equator , where it is heated by shocks . due to the higher densities in this `` disk '' ,
the radiation force is reduced and radial velocities are relatively small . in fig.[lineprofiles ] we show the observed civ variability as a function of the rotational phase and calculations for the blue line of the doublet .
both in the observations and the simulations with a fixed abundance , most absorption is seen along the magnetic equatorial plane . for the simulations including the ionisation balance
, the pattern appears to be offset by 0.25 in phase .
this is caused by the high temperature in the disk which reduces the civ abundance .
a more detailed treatment of the auger effect is likely required to account for this offset , since most x - rays will come from the shock heated disk .
recent results @xcite show that magnetic fields measured with the lsd method could be overestimated .
if @xmath0 cep would indeed have a weaker field , this could help to better explain the observed wind terminal speed .
although more work needs to be done , in principle the effect of the magnetic field on the wind of @xmath0 cep can explain the observed uv line - profile variability .
further modelling of other magnetic early b stars and wind lines like nv and siiv , will allow a detailed view on how stellar winds are affected by magnetic fields , and the importance of the auger effect .
henrichs , h. f. , de jong , j. a. , et al .
2000 , in aspcs vol214 , the be phenomenon in early - type stars , ed .
smith , h.f .
henrichs and j. fabregat , p.324 leone , f. , schnerr , r. s. , stift , m. j. , & henrichs , h. f. 2005 , , in prep .
neiner , c. , geers , v. c. , henrichs , h. f. , et al .
2003a , , 406 , 1019 neiner , c. , hubert , a .- m . ,
fr ' emat , y. , et al . 2003b , , 409 , 275 neiner , c. , henrichs , h. f. , floquet , m. , et al . 2003c , , 411 , 565 ud - doula , a. , & owocki , s. p. 2002
, , 576 , 413 | we present results of numerical simulations of wind variability in the magnetic b1 i ve star @xmath0 cephei .
2d - mhd simulations are used to determine the structure of the wind . from these wind models
we calculate line profiles for different aspect angles to simulate rotation .
the results are compared with the observed uv wind line profiles . |
electrons are identified at mid - rapidity using the information provided by the time projection chamber ( tpc ) , the transition radiation detector ( trd ) , and the time - of - flight ( tof ) detector .
the time - of - flight of the individual particles is required to be consistent with the electron hypothesis . using the trd information ,
an electron likelihood is computed from the energy deposited in the 6 detector layers by charged particles . in order to provide a good e/@xmath1 separation ,
tracks are required to have at least 5 out of maximum 6 chambers providing information on the energy deposit .
a cut providing an electron selection efficiency of 80% is applied to the likelihood value .
finally , we consider the tpc energy loss d@xmath2/d@xmath3 expressed as deviation from the bethe - bloch electron curve and select electron candidates from the top half of the d@xmath2/d@xmath3 distribution , to guarantee very high electron purity .
inclusive electron spectra are based on high - quality tracks passing the particle identification criteria over a momentum range where the hadron contamination does not exceed 10% . in the pp analysis
all three detectors were used to identify electrons and the spectra extend up to 10 gev/@xmath4 in transverse momentum @xmath5 . in the pb
pb analysis only the tof and tpc information were used and the spectra extend up to 6 gev/@xmath4 .
the remaining hadron contamination is estimated by fitting the electron and the hadron tpc d@xmath2/d@xmath3 gaussian distributions in narrow momentum slices and counting the hadrons passing the pid cut .
this contribution is subtracted .
the electron yield is corrected for acceptance and for the efficiency of the selection criteria , using monte - carlo simulations .
an unfolding procedure is applied to correct for the distortion of the @xmath5 distribution due to bremsstrahlung .
the inclusive electron spectrum contains contributions from many sources other than heavy - flavour hadron decays .
the most important are : dalitz decays of light - flavour neutral mesons ( @xmath6 , @xmath7 , @xmath8 , @xmath7 , @xmath9 ) , photon conversions in the beam pipe and in the detector material , di - electron decays of vector mesons ( @xmath10 , @xmath8 , @xmath9 ) , heavy quarkonia , real and virtual qcd photons . the spectrum composed of these background electrons
is described by the so - called electron cocktail , calculated on the basis of meson spectra measured with the alice detector , as described in more detail elsewhere @xcite .
further components are parametrized according to other measurements at the lhc ( e.g. quarkonia ) or theoretical predictions ( e.g. qcd photons ) . in this analysis
, the statistical subtraction of the cocktail from the inclusive electron spectrum provides the transverse momentum distribution of electrons from heavy - flavour hadron decays .
a sample of minimum bias pp collisions ( 2.6 nb@xmath11 ) recorded by alice in summer 2010 at @xmath0=7 tev was used to measure the production cross section of electrons from heavy - flavour decays according to the method discussed above .
the measured spectrum , including both the charm and the beauty contributions , is shown in fig .
[ fig1 ] on the left .
the present conservative estimate of the systematic uncertainty is @xmath1220% on the measured spectrum and @xmath1210% on the electron cocktail .
the measured cross section is compared to a prediction at fixed - order plus next - to - leading logarithms ( fonll ) @xcite which provides a good description .
= 7 tev.,title="fig : " ] = 7 tev.,title="fig : " ] the silicon inner tracking system ( its ) provides excellent impact parameter resolution allowing for isolation of decay products of longer living particles .
electrons from the decay of hadrons with beauty content are selected by requiring large displacement from the interaction vertex .
the impact parameter resolution is well described in our simulations .
a selection is applied by requiring the electron to miss the interaction vertex by at least three times the error on the separation of the track from the vertex ( of the order of 80 @xmath13 m at 2 gev/@xmath4 ) .
the remaining contribution from charm decays is estimated from the d meson cross section at mid - rapidity , and subtracted .
the resulting spectrum of electrons from beauty decays is shown in fig.[fig1 ] on the right , also compared to the fonll prediction for the beauty component .
this is the first beauty production cross section measurement in alice .
a similar analysis is performed with pb
pb collisions recorded by alice in november and december 2010 .
electrons are identified with the tof and tpc detectors and inclusive electron spectra are obtained in 6 different centrality bins .
the hadron contamination in the electron sample remains below 10% in the momentum range between 1.5 and 6 gev/@xmath4 , to which this first measurement is restricted .
the inclusive electron spectra are corrected for acceptance and efficiencies using a hijing monte - carlo sample enriched with heavy - flavour signals .
an electron cocktail for each centrality bin is calculated based on the charged pion spectra measured by alice . in the pb
pb analysis the systematic uncertainties currently are large : of the order of 35% on the inclusive spectra and 25% on the cocktails . when comparing the inclusive electron spectra with those generated using the cocktail prescription ( see fig . 2 and 3 in @xcite , where more details are discussed ) ,
a hint for an excess is observed in the electron yield in the transverse momentum region between 1.5 and @xmath123.5 gev/@xmath4 .
the excess ( indicated by the ratio in fig .
3 in @xcite , significantly larger than in pp collisions , despite the large systematic uncertainties ) exhibits a clear centrality dependence .
cross checks with the alice measurement of d mesons at mid - rapidity seem to exclude an additional charm source .
the possibility of a contribution from thermal radiation , already measured at rhic @xcite , will be further investigated .
we compute the nuclear modification factor @xmath15 defined as @xmath16 , where is the average nuclear overlap function for a given centrality bin , and @xmath17 and @xmath18 describe the electron yield in pb pb and in pp collisions . 0.5 the latter is obtained by extrapolating the measured spectrum at @xmath0=7 tev down to the energy of 2.76 tev according to fonll prescriptions @xcite .
the resulting @xmath15 for the cocktail subtracted electrons in the 0 - 10% most central events is shown in fig .
[ fig2 ] . in the transverse momentum region between 3.5 and 6 gev/@xmath4 ,
in which we assume that the cocktail - subtracted electron spectra are dominated by electrons from heavy - flavour decays , a suppression by a factor 1.5 - 4 is observed .
this is similar to the suppression measured with single muons at forward rapidity @xcite .
this measurement is compatible with in - medium energy loss of the heavy - flavour quarks .
the measurement of electron spectra at mid - rapidity in alice allows to determine the production cross section of electrons from heavy - flavour decays in pp collisions at @xmath0=7 tev .
the first measurement of beauty hadron decays was presented .
fonll predictions provide a good description of the measured spectra . together with the electron spectra measured in pb
pb collisions , a nuclear modification factor is extracted , showing a strong suppression of electrons from heavy - flavour decays in central collisions .
the next goal is to measure the suppression factor for electrons from charm and beauty hadrons separately , to provide constraints to in - medium energy loss models .
10 aamodt k et al .
[ alice collaboration ] , _ j. instrum . _ * 3 * , ( 2008 ) s08002 dainese a for the alice collaboration , these proceedings masciocchi s for the alice collaboration , _ nucl .
phys . _ a * 855 * ( 2011 ) 432435 cacciari m et al . ,
_ j. high energy phys .
_ jhep9805(1998)007 pachmayer y for the alice collaboration , these proceedings adare a et al . [ phenix collaboration ] _ phys .
lett . _ * 104 * 132301 ( 2010 ) averbeck r et al .
, in preparation zhang x for the alice collaboration , these proceedings | electron spectra measured with alice at mid - rapidity are used to study the production of hadrons carrying a charm or a beauty quark .
the production cross section of electrons from heavy - flavour hadron decays is measured in pp collisions at @xmath0=7 tev .
electrons from the beauty decays are identified via the displacement from the interaction vertex . from the electron spectra measured in pb
pb collisions , we determine the nuclear modification factor , which is sensitive to the heavy - quark energy loss in a hot strongly interacting medium .
the measurement of heavy - flavour production cross sections allows an important test for perturbative qcd descriptions of hard processes in hadronic interactions
. moreover , charm and beauty quarks are a sensitive tool to study the flavour dependence of parton energy loss in high energy nucleus - nucleus collisions , the quark - gluon plasma ( qgp ) .
precision measurements of heavy - flavour production in proton - proton collisions provide the necessary reference for the interpretation of the pb
pb collision results within a qgp framework . one channel to measure the heavy - flavour production cross section is semi - leptonic decays of hadrons carrying a charm or a beauty quark .
such decays have a relatively large branching ratio ( of about 10% ) and offer a complementary approach to that of the reconstruction in exclusive hadronic decays .
alice ( a large ion collider experiment ) @xcite at the lhc is well equipped to address these physics topics : it has very good particle identification and spatial resolution for the separation of secondary vertices .
the latest alice heavy - flavour physics results are summarised in @xcite .
this contribution presents the results from the measurement of electron spectra in proton - proton and pb pb collisions at mid - rapidity . |
magnetic flux ropes play an important role in solar eruptions manifested as coronal mass ejections ( cmes ) and/or flares .
recently , a new line of observational structures , namely euv hot - blobs and/or channels , have been proposed to be the most direct manifestation of flux ropes ( cheng et al .
2011 ; zhang et al . 2012
; cheng et al . 2013 ; patsourakos et al . 2013 ) with the atmospheric image assembly ( aia ; lemen et al . 2012 ) on board the _ solar dynamics observatory ( sdo ) _ ( pesnell et al . 2012 ) . in particular , zhang et al . (
2012 ) revealed the flux rope as a conspicuous hot channel structure prior to and during a solar eruption , which initially appeared as a writhed sigmoid with a temperature as high as @xmath010 mk , then continuously transformed itself toward a semi - circular shape and acted as the essential driver of the resulting cme .
while previous studies have been focusing on the morphological and kinematic evolution of magnetic flux ropes , little is known about their thermal evolution .
the scenario of heating solar flare loops is well accepted in classical flare models ( carmichael 1964 ; sturrock 1966 ; hirayama 1974 ; kopp & pneuman 1976 ) , in which flare loops are mainly heated by the thermal plasma from the chromospheric evaporation ( doschek et al .
1980 ; feldman et al .
now the question is whether flux ropes , seen as expanding hot channels in a much higher corona , are heated by the same process . in this letter , we intend to use differential - emission - measure ( dem ) based temperature analysis method on an event of failed flux rope eruption to address this issue .
recently , dem analyses have been applied to diagnose the physical properties of cmes ( zhukov & auchre 2004 ; landi et al .
2010 ; tian et al . 2012 ; cheng et al . 2012
; tripathi et al .
2013 ) , the quiet sun ( vsquez et al . 2010 ) and the post - flare loop systems ( reeves & weber 2009 ; warren et al .
the multi - passband broad - temperature capability of aia makes it ideal for constructing dem models .
we find that this method is particularly useful in studying the failed solar eruption , in which the thermal structure is conspicuous and long lasting in aia field of view ( fov ) .
there have been several studies on failed eruptions , but the physical cause of the failure remains elusive .
it is found that long duration flares tend to be more eruptive ( kahler et al .
1989 ) , but many other factors could be invovled , such as the interaction between filaments and their associated magnetic environment ( ji et al .
2003 ; williams et al . 2005 ; gibson & fan 2006 ; jiang et al .
2009 ; kuridze et al .
2013 ) , the degree of the helical twist in the filament ( rust & labonte 2005 ) , the overlying arcade field ( wang & zhang 2007 ) , and the position of the reconnection site ( gilbert et al .
numerical simulations showed that the kink instability could trigger a failed filament eruption if the overlying magnetic field decreases slowly with height ( e.g. , t " or " ok & kliem , 2005 ) .
for the failed eruption in this paper , we can clearly observe and track the flux rope structure , as a hot euv blob , before , during and after the eruption .
using advanced dem - based temperature map method , we reveal , for the first time , the detailed thermal evolution of the flux rope , and conclude that the thermal structure is caused by the direct heating from the energy release of the magnetic reconnection in the flux rope .
further , we find convincing evidences that the failure of the eruption is caused by the strapping effect of the overlying magnetic loop arcade . in section 2 , we present the observations and results , which are followed by a summary and discussions in section 3 .
the aia images the multi - layered solar atmosphere through 10 narrow uv and euv passbands almost simultaneously with high cadence ( 12 seconds ) , high spatial resolution ( 1.2 arcseconds ) and large fov ( 1.3 r@xmath1 ) .
the temperature response functions of their passbands indicate an effective temperature coverage from 0.6 to 20 mk ( odwyer et al .
2010 ; del zanna et al .
2011 ; lemen et al . 2012 ; ) . during eruptions ,
the 131 and 94 passbands are more sensitive to the hot plasma from flux ropes and flare loops , while the other passbands are better at viewing the cooler leading front and dimming regions ( e.g. , cheng et al . 2011 ; zhang et al . 2012 ) .
the observed flux @xmath2 for each passband can be determined by @xmath3 where the @xmath4 and @xmath5 are the temperature response function of passband @xmath6 and the plasma dem in the corona , respectively .
cheng et al ( 2012 ) and we use the xrt_dem_iterative2.pro " routine in ssw package to compute the dem . this code was originally designed for _ _ hinode__@xmath7x - ray telescope data ( golub et al .
2004 ; weber et al . 2004 ) and was modified slightly to work with aia data ( schmelz et al . 2010 , 2011a , 2011b ; winebarger et al .
2011 ; cheng et al .
more details and tests of this method were discussed in the appendix of cheng et al .
( 2012 ) . here
, we use the dem - weighted average temperature per pixel defined in the following formula ( cheng et al .
2012 ) to construct the temperature map in spatial domain and study their temperature evolution in time .
@xmath8 errors in dem inversion arise from the uncertainties in the response function @xmath4 ( judge 2010 ) and the background determination ( e.g. , aschwanden & boerner 2011 ) . according to cheng et al .
( 2012 ) , the error of dem - weighted temperature at the flux rope center could be @xmath015% .
we should point out that the plasma , integrated along the line of sight , contains multiple temperatures .
the weighted " temperature we are obtaining here , is an indicator of the overall thermal trend of the plasma in the temperature range the instrument is sensitive to . on 2013
january 5 , an m1.7 class soft x - ray flare occurred at the northeast limb of the sun , which started at 09:28 ut and peaked at 09:31 ut .
the flare was located at @xmath0n20e88 ( noaa 11652 ) from the perspective of the earth .
no associated cme was observed by the large angle spectroscopic coronagraph ( lasco ; brueckner et al .
1995 ) and the sun earth connection coronal and heliospheric investigation ( secchi ; howard et al .
but in aia 131 and 94 passbands , an obvious structure with high temperature was observed to erupt , but stopped to rise further . in figure 1 , we present the observations of this failed eruption process .
panels ( a)-(c ) are observations from 131 , ( d ) is a composite image of 94 ( blue channel ) and 171 ( green channel ) and ( e)-(f ) are composite images of 131 ( red channel ) and 171 .
( see supplementary movies 1 and 2 for the entire and continuous eruption process in six aia euv passbands ) .
the left panels show the coronal images immediately before the eruption .
the upper - pointing red arrows indicate the position of the flux rope structure that was about to rise and erupt .
the flux rope is better seen in 94 in the early time of the evolution , which means that its initial temperature should be around 6 mk ; this is consistent with the dem analysis result shown in figure 2 ( a ) .
this temperature effect explains why the flux rope is better seen in 94 but less visible in 131 ( figure 1 ( a ) ) and completely invisible in 171 ( figure 1(d ) ) in the early time .
this flux rope can be observed for its favorable orientation .
it seems that the axis of the rope lied along the line of sight at the limb of the sun , making it a bright blob - like structure . at 09:28:56 , a hot and fast narrow jet started to appear in 131 images , and became obvious at 09:29:20 as depicted with blue arrows in figures 1(b ) and ( e ) . with dem analysis
, we learned that the hot jet temperature is around 8 mk ( figures 2(b ) and ( c ) ) .
therefore , the jet is better seen in 131 but could not be observed in 171 passband , see figure 1(e ) . the flux rope started to rise at about 09:29:20 ut , stopped at 09:32:35 ut , and then stayed in the position for several hours before faded away .
the flux rope could not be observed in 171 during the entire process because of its high temperature .
but the 171 observations show the movement of the overlying loop arcade clearly ( bottom panels in figure 1 ) .
the asterisks in the panels show the position of the inner edge of the overlying loop arcade . with a horizontal white line passing through the two asterisks in the frames of earlier times when
the loop arcade had not begun to rise , it is easy to notice that the loop arcade was shifted and raised to a higher position following the eruption .
the 211 and 193 observations also show the overlying loops stressed by the rope ( see supplementary movie 1 ) .
the red plus symbols shown in the right panels point to the apex position of the flux rope .
figure 2 shows a sequence of temperature maps throughout the entire thermal evolution process of the event . before the eruption ,
the temperature of the flux rope ( depicted with a white arrow ) is over 5 mk as shown in figure 2(a ) .
figures 2(b ) and ( c ) show the hot jet ( depicted with the white arrows ) underneath the flux rope .
the appearance of the hot jet seemed to signal the onset of the flare and also the onset of the flux rope rising .
we find that the temperature over the eruption region increased quickly ( see supplementary movie 3 ) .
two sub - regions , as depicted with a large square in the high corona and a small rectangle close to the solar surface in figures 2(b ) , ( e ) and ( h ) , are selected to be the regions of the flux rope and flare loops respectively , for the purpose of tracking their temperature evolutions . to obtain the characteristic average temperature of the entire flux rope , all the pixels with temperatures greater than 5.0 mk in the selected square
are regarded as the flux rope pixels , and their average temperature is regarded as the characteristic temperature . the same way is used to get the average temperature of the flare loop region as selected with the rectangle .
the temperature evolution of the flux rope and the flare loops are shown in figure 3 with red and blue solid line , respectively , along with the _ geostationary operational environmental satellite ( goes ) _ soft x - ray 1 - 8 profile in the black line .
the insert panel in figure 3 is an enlarged portion from 09:28:08 to 09:39:08 ut to better show the relations at the beginning of the eruption .
apparently , the temperature profile of the flare loops is almost the same as the soft x - ray profile , i.e. , the same peak time and the similar rise phase .
this is consistent with the classic thick - target model that the flare region is heated by the precipitation of energetic electrons accelerated in flare reconnection ; note that we might under - estimate the temperature of the flare loops for the following two reasons : first , a part of the hot near - footpoint sources might be occulted by the limb ; second , a small fraction of the flare region were saturated in aia images .
nevertheless , we do nt believe that these effects will change the temporal profile of the temperature evolution .
however , the temperature evolution of the flux rope is remarkably different from that of the flare loops , i.e. , the duration of the temperature increase lasted much longer than that in the flare loops and the rise phase of the soft x - ray flare .
for clarity of discussion , we may divide this duration into two phases : impulsive - heating phase and post - impulsive - heating phase . in the impulsive - heating phase ,
the temperature increased from @xmath05.2mk ( @xmath009:29 ut ) to @xmath08.5 mk ( @xmath009:31 ut ) , while the temperature increased from @xmath08.5 mk to @xmath010 mk ( @xmath009:38 ut ) during the post - impulsive - heating phase .
then the temperature began to slowly decrease to @xmath05.6 mk until 12:00 ut .
it is obvious that the impulsive and post - impulsive heating phases correspond well with the rise phase and decay phase of the associated flare .
the continuing rise of temperature in the flux rope after the flare rise phase is an interesting discovery in this study .
it seems that the heating of the flux rope should be different from that of the flare loops .
we suggest that the direct thermal heating from the magnetic reconnection region , without via the chromospheric evaporation , should be responsible for the heating of the flux rope .
the continuing strong heating of the flux rope after the flare soft x - ray peak time is related with the continuing energy release during the decay phase of the soft x - ray flare . in this long duration flare , there should be continuing magnetic reconnection in the corona .
further , the failure of the eruption resulted in less energy converted to the kinetic energy of the bulk plasma from the released magnetic energy ; this extra energy should further heat the flux rope trapped in the corona .
consequently , the eruption produced a conspicuous large - size high - lying ( @xmath0100 mm above the surface ) high - temperature ( @xmath0610 mk ) fire ball " sitting in the corona for more than two hours . in the end ,
the hot structure faded away and became indistinguishable from the ambient as shown in figure 2(i ) . to further elucidate the relation between the flux rope heating and the magnetic reconnection , we investigated the location of the coronal hard x - ray sources using data from _ reuven ramaty high energy solar spectroscopic imager ( rhessi ) _
( lin et al 2002 ) .
unfortunately , rhessi was in night until @xmath009:46 ut . rhessi observations around 09:47 ut are thus selected and shown as black contours in the 1220 kev band at 50% , 70% and 90% of the maximum in figures 1(c ) and 2(f ) .
the contours apparently show a double x - ray source .
it has been suggested that the region of magnetic reconnection should be between the two x - ray sources ( e.g. , liu et al .
therefore , it seems that the magnetic reconnection separates the flux rope structure into two parts : the upper part and the lower part .
such scenario of magnetic reconnection leading to partial eruption have been reported in several studies ( gilbert et al . 2000 ; gibson & fan 2006 ; tripathi et al .
2007 ; sterling et al . 2011 ; tripathi et al . 2013 ) .
we also study the kinematic evolution process of the flux rope and its overlying loops .
the heights of flux rope apexes ( red plus symbols shown in the right panels of figure 1 ) and the inner edge of the overlying loop arcades ( blue asterisk symbols shown in the bottom panels of figure 1 ) are tracked in the aia images and shown in figure 4(a ) .
their velocities are shown in figure 4(b ) , along with the _ goes _ soft x - ray 1 - 8 profile shown in the black line .
the velocities are calculated from the numerical differentiation using the 3-point lagrangian interpolation of the height - time data .
note that the uncertainties of the velocities come mainly from the uncertainties in the height measurement .
the measurement errors are estimated to be 2 pixels .
these errors are propagated in the standard way to estimate the errors of velocity .
the red , blue and black filled circles show the times of the maximum velocities of the flux rope and overlying loops , and the peak of the soft x - ray flux , respectively . from figure 4(a )
we find that the flux rope started to show noticeable movement at 09:29:20 ut , and the loop arcade , started to move at a later time at 09:30:11 ut .
apparently , the ascending motion of the overlying loop arcade is induced by the rising motion of the erupting flux rope ; the early interaction between the flux rope and the loop arcade should take place at a time between 09:30:11 ut and 09:30:23 ut as indicated by the two vertical dashed lines in figure 4 ; at this time , the flux rope had already obtained a speed of about 400 km / s .
the interaction apparently prevented the further eruption of the flux rope and led its velocity to decrease quickly , until came to a full stop two minutes later .
before the contact of the interaction , the velocity of the flux rope increased from @xmath095 km s@xmath9 at 09:39:20 ut to @xmath0431 km s@xmath9 at 09:30:08 ut ; during this 48 s of free " acceleration , the average acceleration rate is estimated to be @xmath07000 m s@xmath10 .
this is an extremely strong acceleration , when compared with the rates of most cmes which are typically lower than 1000 m s@xmath10 ( zhang & dere 2006 ) .
the interaction made the velocity of the rope just increased from 431 km s@xmath9 at 09:30:08 ut to 438 km s@xmath9 at 09:30:20 ut with an average acceleration 580 m s@xmath10 .
then the velocities of the flux rope began to decrease quickly from 438 km s@xmath9 to almost zero at 09:32:35 ut with the average decceleration rate 3240 m s@xmath10 . on the other hand ,
the velocity of the loop arcade started at about 70 km s@xmath9 when the movement became noticeable at 09:30:11 ut , and increased to 98 km s@xmath9 at 09:30:23 ut as its peak value .
the early acceleration of the loop arcade was not observed , probably because the early interaction was short and prompt , even the 12-second cadence of aia was not sufficient to capture the evolution .
the instantaneously accelerated loop arcade immediately began its deceleration .
the deceleration phase of the loop arcade coincided well with that of the flux rope , coming to a full stop about two minutes later .
we conclude that the magnetic strapping effect of the overlying loop arcade , likely due to the magnetic tension force along the loops , made the eruption a failed one , instead of forming an escaping cme .
a failed flux rope eruption associated with an m1.7 class flare was observed by the aia at the northeast limb of the sun on 2013 january 5 , which provided us an unprecedented opportunity for studying the detailed temperature evolution of the flux rope .
we find that the flux rope existed in the corona with a high temperature ( @xmath05 mk ) before its eruption .
the temperature evolution of the flux rope can be divided into two phases : impulsive - heating phase and post - impulsive - heating phase , which correspond well with the rise phase and decay phase of the associated flare . on the contrary ,
the temperature of the low - lying flare loops only increased during the flare rise phase , and quickly decreased during the decay phase .
furthermore , there was a gap of low temperature between the flux rope and the flare loops .
all of these observations indicate that the flux rope should be heated by a different process from that of the flare loops .
the heating process of flare loops have been well understood and widely accepted as summarized below .
energetic electrons and ions produced by flare magnetic reconnection precipitate from the coronal acceleration site and lose their energy in the dense underlying chromosphere via coulomb collisions . the temperature in the chromosphere increases and the resulting pressure exceeds the ambient chromospheric pressure , which leads the heated plasma to expand along the magnetic field , forming flare loops ( doschek et al .
1980 ; feldman et al .
1980 ) . while for the flux rope
, we suggest that it should be heated directly by the thermal energy generated at the reconnection site through the thermal conduction .
it is unlikely that chromospheric evaporation plays a strong role in heating the flux rope .
the velocity analysis of the flux rope and the overlying loop arcades strongly suggests that the failure of the eruption be caused by the strapping effect of the overlying magnetic loop arcade .
the strapping effect , or the tension force of the line - tying field , is the determining factor that contained the flux rope from further eruption .
we thank the anonymous referee for his / her valuable comments that helped to improve the paper .
sdo is a mission of nasa s living with a star program .
this work is supported by the 973 program 2012cb825601 , nnsfc grants 41104113 , 41274177 , 41274175 , and 41331068 .
h. q. song is also supported by the natural science foundation of shandong province zr2010dq016 .
j. zhang is supported by nsf grant atm-0748003 , ags-1156120 and ags-1249270 .
b. li is supported by nnsfc grant 41274176 .
reeves , k. k. , & weber , m. a. 2009 , in asp conf .
415 , the second hinode science meeting : beyond discovery - toward understanding , ed .
b. lites , m. cheung , t. magara , j. mariska , & k. reeves ( san francisco , ca : asp ) , 443 weber , m. a. , deluca , e. e. , golub , l. , & sette , a. l. 2004 , in iau symp .
223 , multi - wavelength investigations of solar activity , ed .
a. v. stepanov , e. e. benevolenskaya , & a. g. kosovichev ( cambridge : cambridge univ .
press ) , 223 , 321 | in this paper , we report for the first time the detailed temperature evolution process of the magnetic flux rope in a failed solar eruption . occurred on january 05 , 2013 , the flux rope was impulsively accelerated to a speed of @xmath0400 km / s in the first minute , then decelerated and came to a complete stop in two minutes .
the failed eruption resulted in a large - size high - lying ( @xmath0100 mm above the surface ) high - temperature fire ball " sitting in the corona for more than two hours .
the time evolution of the thermal structure of the flux rope was revealed through the differential emission measure analysis technique , which produced temperature maps using observations of the atmospheric imaging assembly on board _ solar dynamic observatory_. the average temperature of the flux rope steadily increased from @xmath05 mk to @xmath010 mk during the first nine minutes of the evolution , which was much longer than the rise time ( about three minutes ) of the associated soft x - ray flare .
we suggest that the flux rope be heated by the energy release of the continuing magnetic reconnection , different from the heating of the low - lying flare loops , which is mainly produced by the chromospheric plasma evaporation .
the loop arcade overlying the flux rope was pushed up by @xmath010 mm during the attempted eruption .
the pattern of the velocity variation of the loop arcade strongly suggests that the failure of the eruption be caused by the strapping effect of the overlying loop arcade . |
with the discovery of neutrino masses and mixing in neutrino oscillation experiments , leptogenesis @xcite has become the most attractive model of baryogenesis to explain the observed matter - antimatter asymmetry of the universe .
this can be expressed for example in terms of the baryon - to - photon number ratio and is very well measured by cmb observations @xcite to be [ etabobs ] _
b^cmb = ( 6.2 0.15)10 ^ -10 .
leptogenesis originates from the see - saw mechanism @xcite that is based on a simple extension of the standard model where right - handed ( rh ) neutrinos with a majorana mass matrix @xmath15 and yukawa couplings @xmath16 to leptons and higgs are added . within @xmath0 models , three rh neutrinos @xmath17
( @xmath18 ) are nicely predicted and for this reason they are traditionally regarded as the most appealing theoretical framework to embed the seesaw mechanism .
however , within the simplest set of assumptions inspired by @xmath0 models @xcite , barring strong fine - tuned degeneracies in the rh neutrino mass spectrum and using the experimental information from neutrino oscillation experiments , the traditional @xmath19-dominated leptogenesis scenario predicts an asymmetry that falls many orders of magnitudes below the observed one @xcite .
this is because , within @xmath19-dominated leptogenesis , where the spectrum of rh neutrinos is hierarchical and the asymmetry is produced from the decays of the lightest ones , successful leptogenesis implies a stringent lower bound on their mass @xcite , @xmath20 . on the other hand , @xmath0 grand - unified theories typically yield , in their simplest version and for the measured values of the neutrino mixing parameters , a hierarchical spectrum with the rh neutrino masses proportional to the squares of the up - quark masses , leading to @xmath21 and therefore to a final asymmetry much below the observed one .
however , it has been shown @xcite that , when the production from the next - to - lightest rh neutrinos @xcite and lepton flavour effects @xcite are simultaneously taken into account @xcite , the final asymmetry can be generated by the decays of the next - to - lightest rh neutrinos and allowed regions in the low energy neutrino parameter space open up . in this paper
we proceed with the analysis of @xcite and present the resulting constraints on all low energy neutrino parameters .
the paper is organized as follows . in section 2
we discuss the current experimental status on low energy neutrino parameters , we set up the notation and describe the general procedure to calculate the the asymmetry and find the constraints . in section 3 we first consider the case already studied in @xcite , when the dirac basis and the charged lepton basis coincide and then , in section 4 , we allow for a misalignment between the two bases not larger than that one described by the ckm matrix in the quark sector .
finally , in section 5 we present a global scan in the space of parameters where all possible cases between the case of no misalignment and the case of a misalignment at the level of the ckm matrix are taken into account .
we also discuss two scenarios , one at small @xmath22 and one at large @xmath22 , and show how , within @xmath0-inspired models , minimal leptogenesis could be tested in future low energy neutrino experiments .
notice that our discussion is made within a non - supersymmetric framework .
recently a study of @xmath0-inspired models within a supersymmetric framework has also enlightened interesting potential connections with lepton flavour violating decays and dark matter @xcite .
an analysis of leptogenesis within left - right symmetric models , where a type ii seesaw contribution to the neutrino mass matrix is also present , has been performed in @xcite . within these models ,
the minimal type i scenario considered here represents a particular case recovered under specific conditions .
after spontaneous symmetry breaking , a dirac mass term @xmath23 , is generated by the vacuum expectation value ( vev ) @xmath24 gev of the higgs boson . in the see - saw limit , @xmath25 , the spectrum of neutrino mass eigenstates splits in two sets : three very heavy neutrinos , @xmath26 and @xmath27 respectively with masses @xmath28 almost coinciding with the eigenvalues of @xmath15 , and three light neutrinos with masses @xmath29 , the eigenvalues of the light neutrino mass matrix given by the see - saw formula @xcite , m_= - m_d1d_mm_d^t , that we wrote in a basis where the majorana mass matrix is diagonal defining @xmath30 .
the symmetric light neutrino mass matrix @xmath31 is diagonalized by a unitary matrix @xmath32 , @xmath33 with @xmath34 , that , in the basis where the charged lepton mass matrix is diagonal , can be identified with the lepton mixing matrix .
neutrino oscillation experiments measure two neutrino mass - squared differences . for no one
has @xmath35 and @xmath36 .
the two heavier neutrino masses can therefore be expressed in terms of the lightest neutrino mass @xmath22 as [ m2m3nor ] m_2 = , m_3 = , where we defined @xmath37 and @xmath38 @xcite .
recently , a conservative upper bound on the sum of neutrino masses , @xmath39 , has been obtained by the wmap collaboration @xcite combining wmap 7 years data plus baryon acoustic oscillations observations and the latest hst measurement of @xmath40 . considering that it falls in the quasi - degenerate regime
, it straightforwardly translates into [ upperbound ] m_1 < 0.19ev ( 95% cl ) .
we will adopt the following parametrization for the matrix @xmath32 in terms of the mixing angles , the dirac phase @xmath41 and the majorana phases @xmath42 and @xmath43 @xcite @xmath44 and the following @xmath45 ranges for the three mixing angles @xcite [ twosigma ] _ 12= ( 31.3 ^ -36.3^ ) , _ 23= ( 38.5 ^ -52.5^ ) , _ 13= ( 0 ^ -11.5^ ) . in the case of io the expression of @xmath46 in terms of @xmath22 becomes [ m2m3inv ] m_2 = , while the expression for @xmath47 does not change . with the adopted convention for the light neutrino masses , @xmath48 , the case of io corresponds to relabel the column of the leptonic mixing matrix performing a column cyclic permutation ,
explicitly @xmath49 the predicted baryon - to - photon ratio @xmath50 is related to the value of the final @xmath51 asymmetry @xmath52 by @xcite [ etab ] _ b 0.9610 ^ -2 n_b - l^f , where @xmath53 is the @xmath54 number in a co - moving volume that contains on average one rh neutrino @xmath55 in thermal ultra - relativistic equilibrium abundance ( @xmath56 ) .
the dirac mass matrix can be diagonalized by a bi - unitary transformation m_d = v_l^d_m_du_r , where @xmath57 .
the matrix @xmath58 can be obtained from @xmath59 , @xmath32 and @xmath60 , considering that it provides a takagi factorization @xcite of @xcite m^-1 d^-1_m_dv_lud_mu^tv_l^td^-1_m_d , explicitly [ takagi ] m^-1 = u_rd_m^-1u_r^t . for non degenerate @xmath61
, the matrix @xmath58 can be determined noticing that it diagonalizes @xmath62 , i.e. m^-1(m^-1)^ = u_rd_m^-2u_r^ .
this relation determines @xmath58 unless a diagonal unitary transformation , since any @xmath63 is also a solution .
however , given a @xmath64 , one can fix @xmath65 from the eq .
( [ takagi ] ) , [ dphi ] d _ = and in doing so @xmath58 is unambiguously determined .
inspired by @xmath0 relations , we can parameterize the eigenvalues of @xmath66 in terms of the up quark masses as [ so(10 ) ] _1= _ 1m_u , _2= _ 2 m_c , _3= _ 3m_t . within @xmath0 models
one can expect @xmath67 and we will refer to this case . the reader is invited to read ref .
@xcite for a more comprehensive discussion about these @xmath0-inspired relations .
notice however that our results will be valid for a much broader range of values , since , quite importantly , it turns out that they are independent of @xmath68 and @xmath69 provided @xmath70 and @xmath71 . with the parametrization eq .
( [ so(10 ) ] ) and barring very special choices of parameters where the rh neutrino masses can become degenerate @xcite , and @xmath72 .
this is clearly a conservative condition , since the asymmetry gets enhanced when @xmath73 or @xmath74 .
however , in this way , we only neglect very special points in the parameter space yielding @xmath75 and @xmath76 .
we will comment again later on this point . ]
the rh neutrino mass spectrum is hierarchical and of the form ( for generic expressions in terms of the low energy parameters , see ref .
@xcite ) [ alpha ] m_1:m_2:m_3=(_1m_u)^2:(_2m_c)^2:(_3m_t)^2 . as we said ,
the values of @xmath77 and @xmath78 are actually irrelevant for the determination of the final asymmetry ( unless @xmath77 is unrealistically large to push @xmath79 from @xmath80 above the lower bound @xmath81 gev to achieve successful @xmath19 leptogenesis ) . on the other hand ,
the value of @xmath82 is relevant to set the scale of the mass @xmath83 ( valid for @xmath84 ) of the next - to - lightest rh neutrino mass , but it does not alter other quantities crucial for thermal leptogenesis , such as the amount of wash - out from the lightest rh neutrinos . defining the flavoured @xmath85 asymmetries as _ 2-_2-_2 _
2+_2 , these can be calculated using @xcite [ eps2a ] _ 2 \ { im+ } , where ( x)=x and @xmath86 is the decay rate of the rh neutrino @xmath1 into the flavor @xmath87 with couplings given by the yukawa s matrix @xmath16 .
we will assume an initial vanishing @xmath1-abundance instead of an initial thermal abundance as in @xcite .
in this way , a comparison of the results in the two analyses gives a useful information about the dependence of the final asymmetry on the initial @xmath1 abundance when successful leptogenesis is imposed .
let us now define the flavored decay parameters as k_i = = , where @xmath88 is the hubble rate , m_= 1.0810 ^ -3ev , @xmath89 is the number of the effective relativistic degrees of freedom and @xmath90 is the planck mass .
the total decay parameters are then just simply given by @xmath91 .
it is also convenient to introduce the quantities @xmath92 . from the decay parameters
one can then calculate the efficiency factors that are the second needed ingredient , together with the @xmath85 asymmetries , for the calculation of the final asymmetry .
these can be well approximated by the following analytical expression @xcite at the peak for @xmath93 .
for @xmath94 , the difference is below @xmath95 . ] ( k_2,k_2 ) = _ -^f(k_2,k_2)+ _ + ^f(k_2,k_2 ) , where the negative and the positive contributions are respectively approximately given by [ k- ] _
-^f(k_2,k_2)-2p_2 ^ 0 e^-3k_2 8 ( e^p_2 ^ 02n_n_2(z_eq ) - 1 ) , @xmath96 and [ k+ ] _
+ ^f(k_2,k_2 ) ( 1-e^-k_2z_b(k_2)n_n_2(z_eq)2 ) , where z_b(k_2 ) 2 + 4k_2 ^ 0.13e^-2.5k_2=o(110 ) .
the @xmath0-inspired conditions @xmath97 , yield a rh neutrino mass spectrum with @xmath98 , though , as we already noticed , this spectrum is obtained for a broader range of @xmath99 values . in this situation ,
the asymmetry is dominantly produced from @xmath1 decays at @xmath100 in a two flavour regime , i.e. when final lepton states can be described as an incoherent mixture of a tauon component and of coherent superposition of a an electron and a muon component .
therefore , at the freeze - out of the @xmath1 wash - out processes , the produced asymmetry can be calculated as the sum of two contributions , [ nbmltm2 ] n_b - l^t~m_2 _ 2(k_2,k_2)+ _
2e+(k_2,k_2e+ ) , where @xmath101 stands for @xmath102 and @xmath103 .
more precisely , notice that each flavour contribution to the asymmetry is produced in an interval of temperatures between @xmath104 $ ] and @xmath105 $ ] , with @xmath106 .
at @xmath107 the coherence of the @xmath108 quantum states breaks down and a three flavour regime holds , with the lepton quantum states given by an incoherent mixture of @xmath109 , @xmath110 and @xmath111 flavours .
the asymmetry has then to be calculated at the @xmath19 wash - out stage as a sum of three flavoured contributions .
the assumption of an initial vanishing @xmath1-abundance allows to neglect the phantom terms in the muon and in the electron components @xcite so that the final asymmetry can be calculated using the expression n_b - l^f
_ 2e+(k_2 e+ ) e^-38k_1 e+ p^0_2p^0_2e+_2e+(k_2 e+ ) e^-38k_1 + _ 2 ( k_2 ) e^-38k_1 .
notice that successful leptogenesis relies on points in the parameter space where one out of the three @xmath112 . from this point of view the constraints on low energy neutrino experiments that we will obtain should be quite stable against effects that could enhance the asymmetry such as a resonant enhancement for special points where @xmath113 .
such effects are however still able to relax the lower bound on @xmath114 and on the @xmath115 , since the @xmath116 s do not depend on @xmath114 .
we start from the case @xmath117 that has been studied already in @xcite deriving constraints in the plane @xmath118 for no . here
we show constraints on all low energy neutrino parameters , including the case of io .
let us first discuss the case of no . in fig . 1 we plotted the final asymmetry @xmath50 for the same three sets of values of the involved parameters as in the fig . 4 of ref .
@xcite , where these three choices were corresponding to three different kinds of solutions for successful leptogenesis .
+ + + + this time the third solution ( right panel ) , is suppressed and successful leptogenesis is not attained . in @xcite ,
this was the only solution corresponding to a final asymmetry dominantly in the muon flavour instead than in the tauon flavour ( as for the first two ) .
the suppression that we find now is explained partly because we are adopting a correct determination of the phases in the @xmath58 matrix ( cf .
( [ dphi ] ) ) and partly because we are now assuming an initial vanishing @xmath1-abundance instead than an initial thermal one .
we will see however that , allowing for @xmath119 , this kind of solution will again yield successful leptogenesis in some allowed regions of the parameter space , characterized in particular by large values of @xmath120 .
the solution in the central panel is also partly suppressed and successful leptogenesis is not attained .
however , in a parameter scan , we find that this kind of solution can still give successful leptogenesis for slightly different values of the parameters than those indicated in the figure caption . in this case
the difference with respect to the results in @xcite is explained just in terms of the different assumption on the initial abundance .
this dependence on the initial conditions is due to the fact that @xmath121 , i.e. the solution falls in the weak wash - out regime at the production . finally , the first solution ( left panel ) is fully unchanged .
it therefore exhibits a full independence of the initial conditions and this is in agreement with the fact that it respects all the necessary conditions for the independence on the initial conditions found in @xcite .
notice that these conditions also enforce an efficient wash - out of a possible pre - existing asymmetry .
a scan in the space of parameters confirms that these three solutions obtained for special sets of values are actually representative of the three general kinds of solutions that come out and , therefore , the drawn conclusions apply in general . in the panels of figure 2
we show the results of such a scan that highlight the allowed regions in the parameter space projected on different two - parameter planes .
+ + the scatter plots have been obtained scanning the three mixing angles @xmath122 and @xmath6 over the @xmath123 ranges eqs.([twosigma ] ) , the three phases @xmath124 over the ranges @xmath125 $ ] and the absolute neutrino mass scale for @xmath126 .
these ranges also coincide with those shown in the plots , except for @xmath22 where the plots are for @xmath127 simply because no allowed solutions have been found for lower values .
the shown results have been obtained in two steps .
a first scan of @xmath128 points has yielded a first determination of the allowed regions . with a second scan of additional @xmath129 points , restricted to the excluded regions
, we have then more robustly and sharply determined the contours of the allowed regions .
notice that these regions have no statistical significance and the random values of the parameters have been generated uniformly . in the top left panel the three rh neutrino masses are plotted versus @xmath22 .
we have also plotted the lower bound on the reheating temperature calculated as [ trhmin ] t_rh^min .
this calculation relies on the fact that in the case @xmath117 the solutions , as we commented , always fall in a tauon @xmath1-dominated scenario .
it can be seen that the lowest bound is given by @xmath130 that in a supersymmetric version , if unchanged , would be marginally reconcilable with the upper bound from the gravitino problem @xcite .
this is another reason to extend our investigation to cases with @xmath119 in next sections . in the top central panel
we have then plotted the allowed region in the @xmath118 plane that can be compared with an analogous figure in @xcite . here
, however , we show only those points that respect the condition @xmath131 but for 2 different values of @xmath132 ( in @xcite we were only showing points for @xmath133 ) .
the ( red ) star represents a point found for a minimum value @xmath134 .
this point basically roughly indicates where the maximum of the asymmetry occurs in the parameter space for a fixed value of @xmath135
. we will continue to use this convention ( yellow circles for @xmath133 , green squares for @xmath136 and red stars for minimum found @xmath135 value ) throughout the next figures .
the structure of the allowed region in the @xmath118 plane can be understood as follows . since @xmath137 and @xmath138
, we immediately deduce that a large lepton asymmetry in the tau flavor may be produced only for sufficiently large values of @xmath22 .
this is rather easy to understand .
if @xmath22 tends to zero , we go into the so - called decoupling limit , @xmath139 .
as the @xmath85 asymmetry needs ( at least ) two heavy states to be generated at the one - loop level , and disregarding the contribution from the @xmath19 , @xmath140 must vanish .
the wash - out parameter @xmath141 is @xmath142 @xcite and therefore the final baryon asymmetry may be estimated to be _
b510 ^ -3_2 5 ( ) 10 ^ -10 , which requires m_1()^210 ^ -3ev , for no .
this estimate holds if the wash - out from the interaction with @xmath19 is negligible , _
i.e. _ @xmath143 .
of course , the smaller is @xmath22 , the smaller @xmath144 needs to be . for @xmath145 ev , the only possibility is that @xmath144 is significantly below unity .
extending the analysis of ref .
@xcite , one finds [ pp ] s_13(-2 ) > 0.04 . to get the feeling of the figures involved , we may set @xmath146 and find that the wash - out mediated by the @xmath19 s vanishes for an experimentally allowed value of the mixing between the first and the third generation of lh neutrinos , @xmath147 in agreement with our numerical results .
if @xmath22 is larger than @xmath148 ev , then @xmath149 is allowed and @xmath6 can be taken to be vanishing .
notice also that the lower bound eq .
( [ pp ] ) on @xmath6 increases with @xmath150 .
this nicely reproduces the linear dependence emerging from the numerical results in the _ left column middle panel _ for the plane @xmath151 and that is described , roughly for @xmath133 and more accurately for @xmath136 , by [ linear ] _ 2344^+4(_13 - 7^ ) , represented with a dashed line in the panel . in the _ top right panel _ we show the allowed region in the plane @xmath152 .
the @xmath85 non conserving terms in neutrino oscillation probabilities can be expressed in terms of the _ jarlskog invariant _ @xmath153 given by @xcite j_cp & = & im[u_3u_e 2u_2^u_e3^ ] + & = & c_12s_12c_23s_23c^2_13s_13 , such that p___-p_|_|_= 4j_cp_k > js_;kj(m^2_kjl2e ) , where @xmath154 . in the bottom left panel we show the allowed points in the plane @xmath155 .
it can be noticed that a non zero value of @xmath153 is not crucial .
looking at the bottom - central panel , it is interesting to notice that the allowed regions for the majorana phases are centered approximately around @xmath156 and @xmath157 .
these play a role in the determination of the _ effective majorana mass _ of @xmath158 in @xmath159 decays that is given by m_ee & = & |_im_iu_ei^2 | + & = & | m_1c_12 ^ 2c_13 ^ 2e^2i+m_2s_12 ^ 2c_13 ^ 2 + m_3s^2_13e^2i(- ) | . in the bottom - right panel
one can see how there is a precise relation between @xmath160 and @xmath22 , given approximately by @xmath12 .
it can be also noticed that there is quite a strict lower bound @xmath161 .
lowest values @xmath162 are the most favoured ones in this case .
though current planned experiments will not be able to test the full allowed range , it is still interesting that they will test it partially , tightening the constraints on the other parameters as well .
we have also made an interesting exercise .
we determined the constraints without making use of any experimental information on the mixing angles and letting them just simply variate between @xmath163 and @xmath164 .
the results are shown in fig .
[ thijarb ] .
first , notice that the lower bound on @xmath22 relaxes of a few orders of magnitude ( see left panel ) .
then notice quite interestingly that small values @xmath165 are well allowed for @xmath166 but values @xmath167 would have been very marginally consistent .
therefore , the current bound @xmath165 seems to match quite well with successful @xmath0-inspired leptogenesis on the other hand , values @xmath168 would have been more optimal for @xmath165 than the current experimental large atmospheric values ( see the central panel in the figure ) . however , they are still allowed thanks to the observed range of values of the solar neutrino mixing angle ( see the right panel ) . for the solar neutrino mixing angle
there is no real favourite range of values for @xmath165 .
let us now discuss the results for io .
it has been shown @xcite that in grand unified models with conventional type i seesaw mechanism one can always find , for any no model satisfying the low energy neutrino experimental constraints , a corresponding io model . therefore ,
though they exhibit some unattractive features that quite strongly disfavour them ( e.g. instability under radiative corrections ) , io models within grand unified theories are not unequivocally excluded .
it is therefore legitimate to check whether the requirement of successful leptogenesis can somehow provide some completely independent information .
we repeated the same scan performed in the case of no and the results are shown in figure [ constrio ] , the analogous of the figure [ constrno ] for the no case .
one can see that io is only very marginally allowed . for @xmath133 , there is only a small region at large values of @xmath169 . extending the analysis in ref .
@xcite , this is explained by the fact that the wash - out parameter @xmath144 turns out to be k_110 ^ 3evm_atm , while in the no case @xmath144 was proportional to @xmath170 .
this constrains @xmath171 to be as large as possible , thus ruling out small values of @xmath22 .
it is interesting to notice that in this case the allowed values for @xmath7 lie in the second octant and correspond to the largest ones compatible with the current experimental limits .
the allowed values of the effective neutrino mass fall in a narrow range , @xmath172 .
+ + therefore , io will be in any case fully tested from cosmology and @xmath159 experiments during next years .
we will see that this conclusion will hold also allowing @xmath119 . as usual , in the plots the red star corresponds to the minimum value of @xmath135 for which we have found a solution , @xmath173 .
the corresponding set of values indicates approximately where the asymmetry has a maximum for a fixed @xmath135 value . for this choice of values , in figure [ io ] , we show the plots of the rh neutrino masses , of the asymmetry @xmath50 , of the @xmath85 asymmetries @xmath174 , @xmath175 , of @xmath176 and of @xmath177 versus @xmath22 .
+ one can see how the heaviest rh neutrino mass @xmath178 decreases with @xmath22 much faster and at @xmath179 one has @xmath180 .
therefore , as one can see from the central top panel , the @xmath85 asymmetries are this time strongly suppressed at @xmath181 . on the other hand , in the range @xmath182
the @xmath85 asymmetries are large enough that successful leptogenesis is still possible .
notice , that this kind of solutions are a sort of modification of the solution obtained at large @xmath22 values for no , simply shifted at somehow larger values .
the asymmetry is therefore strongly depending on the initial conditions ( @xmath183 ) .
the first kind of solution , at small @xmath22 values , is completely absent .
therefore , though io is strongly disfavoured , it is not completely ruled out , a conclusion somehow very similar to that one obtained from completely independent arguments @xcite . in this case
, however , leptogenesis provides quite a precise quantitative test .
we can conclude this section saying that these results confirm and complete those shown in @xcite .
in particular it is confirmed that there are viable solutions corresponding to the different points shown in the figures falling in the currently experimentally allowed ranges of the parameters , .
the model is therefore not ruled out .
a further step is now to understand whether the model is predictive , excluding regions of the parameter space that future experiments can test . from the figures , as we have discussed ,
it is clear that assuming @xmath117 such excluded regions exist and therefore one obtains interesting constraints .
however , it is important to go beyond the simple condition @xmath117 in order to test the stability of the constraints for variations of @xmath59 .
this is the main objective of the next sections .
we now study how the constraints change when a misalignment between the physical basis where @xmath66 is diagonal and the flavour basis , where the charged lepton mass matrix is diagonal , is considered , corresponding to @xmath119 .
since @xmath59 is unitary , we can parameterize it similarly to the leptonic mixing matrix introducing three mixing angles , one dirac - like phase and two majorana - like phases , @xmath185 where we defined @xmath186 and @xmath187 .
therefore , we have now six additional parameters that give much more freedom .
we will not explore the full parameter space but , in the spirit of @xmath0-inspired models , we will allow only small mixing angles @xmath188 at the level of the mixing angles in the ckm matrix . as a first definite example we repeat the analysis performed for the case @xmath117 for a definite case where the @xmath189 are exactly equal to the mixing angles in the ckm matrix and therefore we set @xmath190 , @xmath191 , @xmath192 , where the latter is the measured value of the cabibbo angle . for no the results are shown in figure 6 .
+ + there is a first result to highlight : @xmath135 values as low as @xmath193 are now allowed .
this is an interesting result in connection with the study of realistic @xmath0 models . at the same time
this result also implies slightly lower values of @xmath114 and consequentially of the minimum value of @xmath194 that can be now as low as @xmath195 , as it can be noticed in the top - left panel in fig .
6 . in this case
we have more generally calculated the minimum reheat temperature as [ trhmin2 ] t_rh^min , considering that in the case the asymmetry at the production can be either tauon dominated or @xmath108 dominated .
this is because the third kind of solution that was highly suppressed in the case @xmath117 , the right panel in fig . 1 , becomes now viable and is @xmath108 dominated , as we will discuss soon in more detail .
notice that if we compare the allowed points for @xmath136 with those found for @xmath117 , the constraints on the low energy neutrino parameters are now less stringent .
in particular an allowed region for values @xmath196 is also found for very small values of @xmath6 . indeed , in the case @xmath117 , and for small values of @xmath22 , the suppression of the wash out value @xmath144 imposed a lower bound on @xmath6 . by choosing @xmath197 introduces the possibility of getting vanishing @xmath144 even for zero @xmath6 angles . extending the analysis of ref .
@xcite , one finds indeed that one configuration where @xmath144 is smaller than unity is attained if @xmath198 ( mod @xmath199 ) and @xmath200\,[m_{\rm sol}^4/(m_2 ^ 3\ , m_3)]\sim -(5/12)(m_{\rm sol}/m_{\rm atm})\sim -10^{-1}$ ] .
this implies @xmath201 ( mod @xmath199 ) , as confirmed by our numerical results .
including in the analysis the atmospheric neutrino mixing angle , as one can see from the panel with the constraints in the @xmath151 plane , only values @xmath202 for @xmath165 are allowed for @xmath203 .
notice that , for @xmath204 , the allowed region in @xmath205 only marginally overlaps , at small values of @xmath7 , with the region for the case @xmath117 .
this means that a measurement of these three quantities can distinguish between the two cases , @xmath117 and @xmath197 , and not all combinations of these three quantities seem to be possible
. we will be back on this point in the next section .
in figure 7 we plotted the relevant quantities for three particular choices of the parameters , as indicated in the figure caption , corresponding to the three kinds of solutions found for @xmath197 .
+ + + + these three sets of values correspond to the three kinds of solutions that are found in the scan plots .
the first two sets , corresponding to the left and central panels , give a tauon dominated asymmetry , while the third set , corresponding to the right panels , yields a muon dominated asymmetry .
notice that these three kinds of solutions are the same three kinds , with slight modifications , found for the case @xmath117 .
however , one can see that this time the third kind of solution , where the final asymmetry is muon dominated , also yields successful leptogenesis .
the major difference that explains this result , is that for @xmath197 the flavoured @xmath85 asymmetries @xmath175 are not as hierarchical as in the case @xmath117 , as it can be clearly seen in the three panels showing the @xmath85 asymmetries in figure 7 .
we have also repeated , as for the case @xmath117 , the exercise to leave the mixing angles completely free , without imposing any experimental constraint finding the results shown in fig .
[ thijarb2 ] .
one can see that in this case the points found when the current experimental constraints are imposed ( the green points ) fall in more marginally allowed regions , also for @xmath6 .
this might suggest that @xmath117 seems to be a more attractive case than @xmath197 .
finally , we also present in figure 9 the constraints obtained for io . even though there is again a remarkable suppression of the allowed regions compared to no , they are somehow less restrictive than for @xmath117
. in particular now a broader range of values for @xmath22 is allowed and @xmath7 can be as low as @xmath206 for @xmath204 .
+ + this is also confirmed by the fact that lowest allowed value is now @xmath207 , much lower than in the case @xmath117 ( it was @xmath208 .
however , it is still fair to say that the io case is only marginally allowed and certainly disfavoured compared to the no case .
the two specific cases that we discussed , @xmath117 and @xmath197 , suggest an interesting sensitivity of @xmath0-inspired leptogenesis to slight deviations of @xmath59 from the identity .
this sensitivity was absent in the results found in @xmath19-dominated leptogenesis @xcite . in this way
it seems that one could even gain some information on @xmath59 from low energy neutrino experiments .
however , there is a potentially dangerous aspect of such a sensitivity : if for a slight variation of @xmath59 the entire space of low energy neutrino parameters becomes accessible , then any chance to test @xmath0-inspired leptogenesis is lost . on the other hand , from a comparison of the results obtained for the two definite cases , @xmath117 and @xmath197 , one can understand that this does not happen .
one can still suspect that for a continuous variation of the parameters in @xmath59 , such that @xmath59 changes from @xmath117 to @xmath197 , new solutions appear so that any point in the space of the low energy neutrino parameters can be obtained for a proper choice of @xmath59 .
in this section we study this issue .
we perform a global continuous scan of the parameters for @xmath59 between @xmath117 and @xmath197 .
obviously a precise limit @xmath197 for such a global scan is somehow arbitrary .
it should be therefore taken as a working assumption defining @xmath0-inspired leptogenesis , even more than the condition @xmath97 that , as we stressed many times , should not be regarded as a very restrictive assumption .
clearly within well defined realistic @xmath0 models , more specific conditions on @xmath59 should be obtained . in any case one expects that if the @xmath59 satisfies the condition @xmath209 , then the allowed values for the low energy parameters should fall in the allowed regions for @xmath0-inspired leptogenesis .
therefore , in this section we present the constraints on the low energy neutrino parameters for a continuous variation of the values of the mixing angles @xmath188 in the range @xmath210 ( i.e. for @xmath209 ) .
more explicitly the shown scatter plots are obtained for the low energy neutrino parameters scanned over exactly the same ranges as for the case @xmath117 .
the three angles in @xmath59 are scanned over the ranges @xmath211 , @xmath212 , @xmath213 , while the three phases are scanned over @xmath125 $ ] . in order to determine the allowed regions ,
we have followed the same strategy as in the case @xmath117 , with a similar total number of scanned points , @xmath214 .
the results for no are shown in figure 10 .
one can see how the allowed regions are approximately given by a super - position of those found for @xmath117 and @xmath197 plus all intermediate solutions .
the result is that now the correlations among the parameters found in the two special cases seem to disappear .
there are however still interesting non trivial constraints .
what clearly survives is that the allowed points still cluster within two distinguished ranges of values for @xmath22 , one range at small values , @xmath215 , and one range at high values , @xmath216 , a distinction that is sharp for @xmath217 ( green squares ) while it is softer for @xmath133 ( yellow circles ) .
+ + + at the same time one can see that a global scan actually shows a slight correlation between @xmath22 and @xmath6 in the low @xmath22 range while the interesting linear dependence between @xmath6 and @xmath7 found for @xmath117 seems now to be lost .
however , it should be considered that these plots are projections on two - parameters planes of an allowed region in a seven - parameter space .
therefore , only a full multi - parameters analysis would be able to unreveal correlations involving more than two parameters .
nevertheless , thanks to the distinct analysis that we carried out for the two special cases @xmath117 and @xmath197 , one can catch sight of an interesting correlation among @xmath22 , @xmath218 , @xmath6 and @xmath160 . to this extent , this time we have also plotted the constraints in the plane @xmath219 , showing how the lower bound on @xmath160 increases with @xmath6 . in order to find out whether the linear dependence between @xmath6 and @xmath7 found for @xmath117 ( cf .
( [ linear ] ) ) still holds for a global scan , we show in fig . 11 the same constraints as in fig .
10 imposing the condition @xmath5 , since the linear dependence was found in that range of values .
we only show the constraints on the relevant parameters , therefore only those in the plane @xmath151 , in the plane @xmath220 and in the plane @xmath219 .
this time we could also easily find points for @xmath221 ( red stars ) , showing again how allowing for a @xmath119 the allowed regions get larger . one can see that the quite clear linear dependence eq .
( [ linear ] ) between @xmath6 and @xmath7 holding for @xmath117 , now turns more , for the red star points at @xmath221 , into an allowed region below the dashed line showed in the figure and corresponding approximately to [ th13th23 ] _
23 49^+ 0.65(_13 - 5^ ) .
this result should be also understood in terms of the condition @xmath222 ( cf .
( [ pp ] ) ) when a very small @xmath59 is allowed clearly yielding a dispersion around the linear dependence eq .
( [ linear ] ) .
notice that inside this region there are still sort of sub - regions that seem to be excluded .
we can summarize these results saying that , at low values of @xmath5 , there is an interesting testable constraints in the plane @xmath151 given by the relation eq .
( [ th13th23 ] ) .
in particular experiments that are already taking data such as the nuclear reactor experiment double chooz @xcite and the long baseline experiment t2k @xcite have the capability of a @xmath223 discovery of values @xmath224 .
our results seem to suggest that if such high @xmath6 values will not be found , then a restricted range of values for @xmath7 is predicted . for example , if @xmath225 then @xmath226 , and if @xmath227 then @xmath202 .
such a constraint on @xmath7 should be also tested during next years with quite a good accuracy by the t2k experiment @xcite .
these constraints in the plane @xmath151 should be considered at this level indicative , and should also consider that they are quite sensitive to the value of @xmath135 .
notice that at the same time , cosmological observations and/or neutrinoless double beta decay experiments should also be able to test the condition @xmath228 .
it should be therefore appreciated that this scenario will be tested during next years .
it is also interesting to notice ( see right panel in fig .
11 ) that there is a linear dependence between @xmath160 and @xmath6 as well . in particular , for @xmath229 , at large values @xmath230 one has @xmath231 and even for @xmath224 one has @xmath232 .
these values for @xmath160 are below the sensitivity of future planned experiments ( @xmath233 ) such as exo @xcite . however
, at least , @xmath160 can not be arbitrary small but has a lower bound that , for sufficiently large @xmath6 values , is @xmath234 times below the currently planned reachable experimental sensitivity , a very small value but maybe not completely hopeless . within the two - parameter analysis we are presenting , we can not draw sharper predictions but is seems quite plausible that from a more involved multi - parameter analysis precise correlations could emerge , maybe also involving the solar neutrino angle @xmath218 . in this respect
the central panel in figure 11 suggests that the solar mixing angle could indeed play also a role and that maybe sharper predictions in the 3 parameter space @xmath235 exist .
we can also study how the allowed regions would reduce requiring large values @xmath236 .
the results are shown in figure 12 .
one can see that in this case one obtains very clear constraints that will allow to test this scenario during next years in a quite unambiguous way .
first of all from the fig . 10 , thanks to the very precise values of the majorana phases , one can notice that there is a very clear relation between @xmath22 and @xmath160 .
second , one can see from the left panel of fig .
12 how there is an upper bound @xmath13 for @xmath229 . for values of @xmath237
, one has even @xmath238 .
it should be said however that at these large @xmath22 values , one typically obtains a final asymmetry that depends on the initial conditions .
since we are assuming vanishing initial @xmath1 abundance and vanishing initial asymmetry , these constraints should be regarded as the most stringent ones , but likely also the best motivated ones .
finally , we repeated the global scan for io as well and the results are shown in fig . 13 .
+ + + one can see how the allowed regions somehow merge those found for the two extreme cases @xmath117 and @xmath197 .
there is therefore nothing really new .
io is quite strongly constrained and it will be fully tested in next years .
in particular we can notice again how there is a clear lower bound on @xmath7 rather than an upper bound as in no .
more particularly , one can notice that the allowed region in the plane @xmath152 is approximately described by [ m1th23io ] _
2343^+12^(0.2ev / m_1 ) ( the dashed line in the upper right panel ) .
it is then quite interesting that @xmath0 inspired leptogenesis is able to distinguish no and io even at @xmath11 , when the same values of @xmath160 and of @xmath239 ( the quantity tested by cosmological observations ) are found both for io and for no . from this point of view
@xmath0 inspired leptogenesis provides a way to solve this ambiguity .
we have derived constraints on the low energy neutrino parameters from @xmath0-inspired leptogenesis .
our investigation shows that even minimal leptogenesis , based on a type i seesaw mechanism and assuming a thermal production of the rh neutrinos and with a traditional high mass scale rh neutrino spectrum , can be testable within a well motivated framework , where the see - saw parameter space is restricted by the @xmath0-inspired conditions .
the role played by the @xmath1 decays is crucial in this respect , not only in re - opening the viability of these models .
the presence in the @xmath1-dominated regime of a double stage , a production stage and a lightest rh neutrino wash - out stage , seems to introduce , as shown simultaneously both by the numerical and by the analytical results , a strong direct dependence on neutrino mixing angles as well , in addition to the dependence on the absolute neutrino mass scale , already found in usual @xmath19-dominated leptogenesis @xcite . in the significant case of no with low @xmath22 values
, the neutrinoless double beta decay effective mass seems to be too small to be measured but not arbitrary small and in any case future experimental results can be anyway useful to restrict the allowed regions for the other parameters and sharpening the predictions .
the results for @xmath119 seems also to be sensitive to @xmath59 itself and they therefore suggest that there is an opportunity to gain information on it , an interesting point within studies of specific @xmath0 models .
it is quite interesting that there is an allowed region in the parameter space that allows large values of @xmath6 testable with on - going reactor neutrino experiments and that for these large values the models favours either large or small @xmath7 values depending whether @xmath5 or @xmath240 . in the small @xmath22
range it is also interesting that the constraints are completely independent of any assumption on the initial conditions , a point that maybe makes this option more attractive .
it is actually quite interesting that this conclusion is also supported by completely independent and general considerations based on the possibility to reproduce , without a particularly fine tuned @xmath58 matrix , the observed atmospheric to solar neutrino mass ratio , @xmath241 , starting from hierarchical neutrino yukawa couplings .
it is found @xcite that this experimental observation is far more natural if the lightest neutrino presents a much stronger hierarchy than the the two heavy ones , as it occurs in the region that we have found at small @xmath22 .
it should be also stressed again , that since our results are independent of @xmath78 and @xmath77 , as far as @xmath242 and @xmath243 , they hold even for a yukawa couplings hierarchy milder than in the case of up quark masses .
this can help to make even more natural to reproduce the result @xmath241 without a fine tuned @xmath58 .
a more precise measurement of @xmath218 could also play a relevant role in testing these models , a point that should be addressed by a more involved multi - parameter analysis .
a future accurate determination of the neutrino mixing angles will be therefore crucial to test @xmath0-inspired leptogenesis and could even yield some interesting information on the matrix @xmath59 . in conclusion
, it seems that @xmath0-inspired leptogenesis provides an interesting well justified example that gives some hopes about the possibility of testing minimal leptogenesis even only with low energy neutrino experiments
. it will be then quite interesting in next years to compare the experimental results with the constraints and the predictions from @xmath0-inspired models that we discussed .
we wish to thank s. blanchet , m. chun chen , f. feruglio , carlo giunti , s. king , r. mohapatra , m. schmidt for useful discussions . p.d.b . acknowledges financial support from the next institute and sepnet . for a recent review ,
see s. davidson , e. nardi and y. nir , arxiv:0802.2962 [ hep - ph ] .
see also , g. f. giudice , a. notari , m. raidal , a. riotto and a. strumia nucl .
b * 685 * , 89 ( 2004 ) ; w. buchmuller , p. di bari and m. plumacher , annals phys . *
315 * , 305 ( 2005 ) .
e. komatsu _ et al .
_ , arxiv:1001.4538 [ unknown ] .
p. minkowski , phys .
b * 67 * , 421 ( 1977 ) ; m. gell - mann , p. ramond and r. slansky , _ proceedings of the supergravity stony brook workshop _ , new york 1979 , eds .
p. van nieuwenhuizen and d. freedman ; t. yanagida , _ proceedings of the workshop on unified theories and baryon number in the universe _ ,
tsukuba , japan 1979 , ed.s a. sawada and a. sugamoto ; r. n. mohapatra , g. senjanovic , phys .
. lett . * 44 * , 912 ( 1980 ) .
a. abada , s. davidson , f. x. josse - michaux , m. losada and a. riotto , jcap * 0604 * , 004 ( 2006 ) ; e. nardi , y. nir , e. roulet and j. racker , jhep * 0601 * , 164 ( 2006 ) ; a. abada , s. davidson , a. ibarra , f. x. josse - michaux , m. losada and a. riotto , jhep * 0609 * , 010 ( 2006 ) .
s. blanchet , d. marfatia and a. mustafayev , jhep * 1011 * ( 2010 ) 038 [ arxiv:1006.2857 [ hep - ph ] ] .
their analysis is restricted to normal hierarchical models with @xmath244 . though we find agreement on some general features , for example on the existence of a lower bound on @xmath22 , in some cases our results seem to indicate different conclusions , like for example on the lower bound on the reheating temperature and on many features of the low energy neutrino constraints when a comparison is possible , for example we do not find a lower bound on @xmath6 when we allow @xmath119 .
however , it should be said that a precise understanding of the different results is made difficult because of the different framework , non supersymmetric in our case and supersymmetric in @xcite , and because the exact assumptions on @xmath59 and the procedure for the calculation of @xmath58 and of the rh neutrino mass spectrum are not specified in @xcite .
s. antusch , s. f. king and a. riotto , jcap * 0611 * ( 2006 ) 011 .
s. antusch , p. di bari , d. a. jones and s. f. king , arxiv:1003.5132 .
e. bertuzzo , p. di bari and l. marzola , arxiv:1007.1641 [ hep - ph ] .
m. y. khlopov and a. d. linde , phys .
b * 138 * ( 1984 ) 265 .
j. r. ellis , j. e. kim and d. v. nanopoulos , phys .
b * 145 * ( 1984 ) 181 ; k. kohri , t. moroi and a. yotsuyanagi , phys .
d * 73 * ( 2006 ) 123511 . | we extend the results of a previous analysis of ours showing that , when both heavy and light flavour effects are taken into account , successful minimal ( type i + thermal ) leptogenesis with @xmath0-inspired relations is possible . barring fine tuned choices of the parameters , these relations enforce a hierarchical rh neutrino mass spectrum that results into a final asymmetry dominantly produced by the next - to - lightest rh neutrino decays ( @xmath1 dominated leptogenesis ) .
we present the constraints on the whole set of low energy neutrino parameters . allowing a small misalignment between the dirac basis and the charged lepton basis as in the quark sector
, the allowed regions enlarge and the lower bound on the reheating temperature gets relaxed to values as low as @xmath2 .
it is confirmed that for normal ordering ( no ) there are two allowed ranges of values for the lightest neutrino mass : @xmath3 and @xmath4 . for @xmath5 the allowed region in the plane
@xmath6-@xmath7 is approximately given by @xmath8 , while the neutrinoless double beta decay effective neutrino mass falls in the range @xmath9 for @xmath10 .
for @xmath11 , one has quite sharply @xmath12 and an upper bound @xmath13 .
these constraints will be tested by low energy neutrino experiments during next years .
we also find that inverted ordering ( io ) , though quite strongly constrained , is not completely ruled out .
in particular , we find approximately @xmath14 , that will be fully tested by future experiments .
c i u
# 1 # 1#1 # 1 # 1#2#3phys
. lett .
* b # 1 * ( # 2 ) # 3 # 1#2#3nucl .
phys .
* b # 1 * ( # 2 ) # 3 # 1#2#3phys .
rev .
lett . *
# 1 * ( # 2 ) # 3 # 1#2#3phys .
rev .
* d # 1 * ( # 2 ) # 3 # 1#2#3z .
phys .
* c # 1 * ( # 2 ) # 3 # 1#2#3class . and quantum grav .
* # 1 * ( # 2 ) # 3 # 1#2#3commun . math .
phys . *
# 1 * ( # 2 ) # 3 # 1#2#3j . math .
phys . *
# 1 * ( # 2 ) # 3 # 1#2#3ann . of phys .
* # 1 * ( # 2 ) # 3 # 1#2#3phys . rep . * # 1c * ( # 2 ) # 3 # 1#2#3progr .
theor .
phys . *
# 1 * ( # 2 ) # 3 # 1#2#3int .
j. mod .
phys . * a # 1 * ( # 2 ) # 3 # 1#2#3mod .
phys .
lett . * a # 1 * ( # 2 ) # 3 # 1#2#3nuovo cim . *
# 1 * ( # 2 ) # 3 # 1#2#3_ibid .
_ * # 1 * ( # 2 ) # 3 |
the university of central florida ( ucf ) is a large metropolitan university located in orlando , florida . shortly before this study was started , the u.s .
department of education national center for education statistics ranked ucf as the sixth largest university in the country based on a fall 2006 enrollment of 46,646 students @xcite . since that time
, the university s enrollment has grown to over 55,000 , making ucf the third largest university in the united states going into the fall 2009 semester @xcite .
the calculus - based introductory physics courses , physics i ( newtonian mechanics ) and physics ii ( electricity and magnetism with some optics ) are required for all students of the college of engineering and computer science .
this college alone had an enrollment of over 6000 in the spring of 2009 @xcite .
the large number of students going through the introductory classes each semester , along with the students from the college of sciences , which includes the physics department , and a large contingent of students from other colleges taking the department s physical science class make ucf s physics department the second - largest in the country based on the number of credit - hours taught per semester @xcite .
the teaching of these introductory classes is a large - scale operation , with ten to twelve lecture sections of eighty to over one - hundred students taught each semester , along with twenty to twenty - two lab sections of thirty to thirty - five students each .
retention of the students within the majors that make up the science , technology , engineering and mathematics fields ( stem ) is a challenge for ucf as well as universities nationally @xcite .
the introductory physics courses are one of the first large hurdles that students must clear on their way to a stem degree .
there are many innovative approaches to teaching introductory physics ( see ref . @xcite and references therein ) .
one that has been demonstrated to effectively improve students conceptual learning is curriculum presented in _ the physics suite _ @xcite .
some elements of _ the physics suite _ apply readily to the large - scale approach to the teaching of the introductory classes at ucf , with little or no change to the scheduling format or the infrastructure used by the department .
_ interactive lecture demonstrations _ ( ild ) @xcite can be used effectively in a large lecture hall setting , so they can be implemented in any of the sections scheduled at ucf .
similarly , _ realtime physics _ ( rtp ) @xcite is designed for use in a lab setting with groups of two to four students performing experiments using computers and probeware , much of which is already used in the ucf introductory physics labs .
these elements of _ the physics suite _ bring physics education research - based curriculum into an existing large - scale physics instruction program while maintaining the department s existing method of scheduling and class structure and making use of the department s existing equipment .
the evaluations outlined in this paper provide evidence of improved student learning due to the effectiveness of including _ the physics suite _ methods as a major part of the instruction in the introductory physics classes .
the large scale of ucf s introductory physics program has allowed a comparison of conceptual learning between significant numbers of students taking part in the new method of instruction with those in traditional instruction , as well as those with partial involvement in both methods , over a short span of time .
the number of sections of each class available for this survey has also provided the opportunity to study the relative benefits of _ interactive lecture demonstrations _ and _ realtime physics _ when implemented seperately as well as in concert .
the lecture sections of the introductory physics classes at ucf are taught in an auditorium - style lecture hall seating 80 - 120 students .
traditional format lectures are taught in a visual method determined by the instructor that is , either projected powerpoint slides or flashing of transparencies or blackboard / whiteboard presentation .
a few instructors have made occasional use of electronic student response systems ( ers ) to incorporate some active student participation into the format and/or have used a combination of the visual methods for the delivery of the lectures .
the sections chosen for participation in _ the physics suite _ project used _ interactive lecture demonstrations _ during one lecture class each week .
a demonstration relating to the current topic being addressed in the class is set up .
students are asked to make a prediction of what results will be seen during the demonstration .
the predictions are recorded as a poll , using the ers .
students are then encouraged to interact in small groups , discussing their predictions and explaining their reasoning .
after the discussion interval , students are polled again and the results are displayed .
the demonstration is then conducted and students can see the actual results . finally , the physical principles involved are discussed , with participation both from students who had reasoned correctly and from some who had based their answers on incorrect preconceptions .
this approach of direct confrontation of preconceptions with actual demonstrations , requiring the students to thoroughly reason through the problem both before and after the demonstration has been shown to be an effective teaching method .
it develops deeper understanding by connecting students intuition to the concepts addressed , and it is effective in changing flawed preconceptions @xcite .
the lab component of the introductory classes is a separate , one credit - hour class .
the classrooms are arranged for eight stations of four students each , and each station equipped with a computer and pasco ci-6560 interface .
the experiments used in the traditional format lab are mostly``cookbook''-style .
the students present their results in a report composed outside of class and turned in the following meeting .
the format of the report is determined by the instructor .
part of the focus of the manual is on experimental set - up and method , and there is some introduction to error analysis .
the lab sections using _ realtime physics _ as their lab manual are set in the same rooms and make use of much of the same equipment as the traditional format labs .
the rtp manual makes use of a guided - inquiry approach to the execution of the experiments . after a brief outline of the experimental set - up ,
the students are asked to make predictions of the results they expect to see . like the ilds in the lecture sections ,
this prediction step will target preconceptions and lead to discussion of the underlying principles among members of the lab group . for most activities ,
the data is plotted in real time , as the experiment takes place .
the students are encouraged to make multiple runs of each experiment , watching the relation between the physical activity and the mathematical plot .
the actual results are compared with the predictions made earlier .
questions are asked to get the students to think about the physical principles determining the experimental results .
some mathematical analysis is done with the graphs and data from the experiment to show relations between physical quantities .
an example of an activity performed during one of the one - dimensional motion labs from physics i follows : a student holds a ball above a motion detector that is hooked to a computer .
the students in the group predict what will happen to velocity and acceleration for the ascent , turning point , and descent of the ball thrown vertically .
then the ball is thrown in the air so the motion detector can plot velocity versus time and acceleration versus time .
now students discuss the results and can see that even though velocity changes , acceleration is constant .
the plots of velocity and acceleration are compared , illustrating the mathematical relation between the quantities .
careful attention to the graphs also helps illustrate the vector nature of the quantities @xcite .
the lab report is completed within the lab class time , consisting of answering the questions posed during the activities .
there is a set of homework questions following each lab , designed to reinforce concepts developed through the activities in the lab . through the fall 2007 semester ,
the first semester of the project , the homework was optional .
since then , the students have been required to turn this in at the start of the following lab .
the emphasis of the rtp labs is on understanding the physical concepts addressed and interpreting the graphs of mathematical relations followed by the physical quantities @xcite .
to evaluate the effectiveness of these portions of _ the physics suite _ method at ucf , all the students in the introductory physics classes were given a pre - test during the first meeting of the lab section , and a post - test on the last day of the lab .
the tests used are concept inventory - style tests and the same tests were given to all students both times .
the physics i test is the _ force and motion concept evaluation _ and is included as an appendix to the thornton / sokoloff article cited @xcite .
the physics ii test is the _ electric circuits concept evaluation _ , also created by sokoloff and thornton .
the difference between each student s score between the pre- and post - tests was not the score evaluated . instead , students pre- and post - test scores were compared based on the percentage of material that they learned during the semester .
the improvement score , normalized gain as introduced by hake , calculates what percentage of the questions missed on the pre - test are answered correctly on the post - test .
the formula used to calculate the normalized gain is @xmath0 .
@xcite the students in each semester registered for their lecture and lab classes without knowing if either would include a component of _ the physics suite _ , so their placement was random .
a small number of students each semester take either the lab or lecture component alone .
the students taking only the lecture component did not take the concept evaluation tests .
the scores from the pre- and post - tests were grouped into four categories : students taking an ild lecture section and an rtp lab , students taking an ild lecture section and a traditional format ( tf ) lab , students taking a traditional format ( tf ) lecture and an rtp lab , students taking a tf lecture and a tf lab . students who missed either test due to add / drop of a class or an absence on either test day were not counted . also , for any student who had taken a pre - test and switched sections and took it again the lower score was kept .
these statistics were used to evaluate the effect of _ the physics suite _
elements on conceptual learning .
the students successful completion of the course , defined as attaining a `` c '' or better , is compared among the four categories .
the percentage of each category that successfully completed the course during the study is presented for both physics i and physics ii .
finally , student opinions were solicited .
interviews were conducted with student volunteers after two semesters , the spring 2008 semester and the spring 2009 semester .
the students did not receive any monetary or grade compensation for the interviews . in the spring of 2008 ,
twenty - eight students were interviewed , eight students from physics i and twenty from physics ii .
six of the students were physics ii rtp students who had taken physics i during the fall with the rtp method .
a survey was given to all sections of the lab classes at the end of the spring 2009 semester .
excerpts from the interviews and results from the survey are presented .
the students conceptual learning as measured by the performance on the _ force and motion concept evaluation_(fmce ) and _ electric circuits concept evaluation _ ( ecce ) pre- and post - tests is presented first .
the comparison is made between students taking sections using either _ interactive lecture demonstrations_(ild ) in their lecture , _
realtime physics_(rtp ) in their lab , or both components of _ the physics suite_(ild & rtp ) and those taking only traditional format classes(tf ) in both their lecture and lab component .
figures [ fig:2048cumulative ] and [ fig:2049cumulative ] show the two - year cumulative performance in physics i ( phy2048 ) and physics ii ( phy2049 ) separately .
the gains for all formats are higher in physics i , with students taking both ild and rtp components gaining over 50% compared to 19% for students having both components in the traditional format . in physics
ii , the gains are less for all formats , but students taking any component of _ the physics suite _ more than doubled their gain over the tf - only group . the remaining figures for this section show the performance of students in the introductory classes by semester for the first two years of the project , from the most recent results back to the start of project .
note that in phy2049 there are no students in the tf lab categories after the first semester of implementation .
after positive results in the initial semester of the project , all sections of the physics ii lab component were switched to the rtp format .
the latest results from the spring semester of 2009 are shown in figure [ fig : sp2009 ] .
figure [ fig : f2008 ] shows the results from the fall semester of 2008 .
figure [ fig : sp2008 ] shows the performance in the spring of 2008 .
finally , figure [ fig : f2007 ] shows the fall of 2007 , the first semester of implementation . for physics i ,
the number of students in each category and for each semester is given in table [ physics1_numbers ] .
the physics ii numbers are in table [ physics2_numbers ] .
fall 2007 was the only semester with traditional format labs in physics ii .
.the number of students taking the fmce , by semester . [ cols="<,^,^,^,^ " , ]
the fmce and ecce test conceptual understanding . another important measure of the success of a course is given by the percentage of students that successfully complete the course .
in the case of the introductory physics courses , the majority of the students are from the college of engineering and computer science , which requires a c or better in the core courses .
figure [ fig : abc ] shows the percentage of students attaining at least a c grade in physics i from the fall 2007 semester through the spring 2009 semester .
again , the performance is divided into the categories based on the combination of lecture and lab instruction method .
this is the one area where taking both ild and rtp components has not shown to be an advantage .
the effect may be masked by less grade inflation in the ild classes taught by the project s pi(c.e . ) , and by the stricter rules of student evaluation he has established in the rtp labs . since the combination of the two elements of _ the physics suite _ created the largest gain measured in conceptual understanding , it may be that it leads to a bigger split between the successful students and those that are still struggling .
this may lead to a higher withdrawal rate . at this point
the investigation has not differentiated between withdrawals and failures .
the retention issue is an important one and the other apparent successes of the project make it seem that it should help with this also .
the evaluation also involved student input .
a survey was given to all students taking the post - test at the end of the spring 2009 semester .
the survey consisted of fifteen statements relating to the benefits of the lab component of the course , the correlation between the lab and lecture components , the recitation sessions conducted during the lab meeting time , and possible ways to maximize the benefit of the lab component .
students were asked to rank their agreement with each statement on a five - point scale : strongly agree , agree , neutral , disagree , or strongly disagree . here
the responses of physics i students are shown for two statements that are key to both the motivation for , and the success of the implementation of _ the physics suite_. they refer to the goal of challenging common preconceptions held by students of the introductory physics classes , which _ the physics suite _
curriculum is written specifically to address , and the goal of better coordinating the lab component topics with the lecture component topics chronologically during the semester .
figure [ fig : survey q8 2048 ] shows the responses to the statement , `` the experiments performed in the lab have caused me to change my mind about how some physical process works . ''
the students who had taken a _ realtime physics _ lab ( rtp ) had a higher percentage of positive responses than those taking the traditional format lab ( tf ) .
the strongest positive response came from students taking the rtp lab , with the traditional lecture component .
these students seemed to have a strong sense that the lab exercises contributed to their understanding of the physical principles involved .
the responses of the students taking an rtp lab component with an ild lecture , although still positive , were not as strong as the rtp lab with traditional lecture students .
this is the group that performed the highest on both the conceptual test and in the lecture component grade .
the lower positive response may be due to the wording of the statement .
these students may feel that their preconceptions were challenged during the ild lecture as well as during the rtp lab experiments .
figure [ fig : survey q2 2048 ] shows the percentage of the responses to the statement , `` there is a connection between the concepts addressed by the lab experiments and the theory covered in the lecture class . ''
again , students have responded more positively if they had the rtp lab than if they had a tf lab .
student volunteers were interviewed following both the spring 2008 and spring 2009 semesters .
the earlier interviews have been transcribed and the student feedback is summarized here . on major items , such as `
do you like the format ? ' , ` have the experiments changed misconceptions ? ' , and ` do the experiments help you learn physics ? ' , the majority of students gave positive feedback .
the interactive nature of the rtp labs was praised by a number of students .
many students claimed to have had preconceptions modified by conducting the experiments in the rtp lab .
a question in the interview addressed the homework and most students had a positive response .
it was an indication the homework was fulfilling its intended purpose .
the interviews were conducted in a focus group setting .
examples of two questions and typical student responses follow .
the responses have been paraphrased to accommodate grouping of like responses . 1 .
* do you like this ( overall ) format as compared to a traditional lab format where you take all the data and later write a report ?
* * much more interactive * more effective because you forget when you try and write the report at home * somewhat , explains to a point that makes you find out a purpose * good to answer questions as going through * does more to reinforce 2 . * do the lab report questions help understand the material ? *
* helps understand * makes you stop and think * helps if you take your time to do it * observing then writing reinforces * seeing the result really helps with concept
although the students are encouraged to do their best on the pre- and post - tests , and most do , there are some who just quickly mark down an answer .
there needs to be an incentive for all to try their best .
it can not be directly related to their grade because , in the case of the pre - test , it would be unfair to grade students on material they have not seen .
the graduate teaching assistants have been asked to offer some form of extra credit to students as an incentive to give a full effort on both tests .
some methods that have been tried are rewarding the top - scoring students on the pre - test , rewarding both the top scoring and highest normalized gain students on the post - test , and counting the post - test as a full - score replacement quiz grade based on effort ( as judged by the instructor ) .
another issue is that as time goes on students from the past rtp labs will have corrected homework that they can share with current rtp students .
although this would not show directly in the pre / post test results , some students may not put full effort into learning .
a possibility may be to have different homework sets appended to those in the lab manual .
attaining a good correlation between the lab and the current topic being covered in the lecture component is always a challenge and was part of the motivation for undertaking the project .
much of the problem involves the different pace of the various lecture instructors . the rtp labs for physics i have been in fairly good agreement with the lectures .
physics ii has been more difficult .
the labs start immediately with circuits , while the lecture deals with static charges , electric force and electric fields .
there is also a point where the lecture starts addressing magnetism and there are no corresponding labs .
the geometric optics labs have been inserted into the lab curriculum at this point , and then circuits are resumed once the lecture begins with some introduction of induction .
although students have had some complaints about the disconnect between the two components , most have been glad for the experience with circuits when that topic is reached in lecture .
there is strong evidence , based on the results from the pre- to post - test gains on the fmce and ecce , that _ the physics suite _ elements implemented into the introductory physics classes at ucf have a positive effect on students conceptual learning .
this is especially true when both the _ interactive lecture demonstrations _ and _ realtime physics _ labs are used in conjunction .
however , the investigation shows evidence that either component benefits student learning when used alone .
this makes _ the physics suite _ a good option for a large - scale physics program with many different instructors . _
realtime physics _ as a stand - alone method for lab instruction pairs well with the lecture component of the course , whether or not the lecture instructor chooses to include _ interactive lecture demonstrations _ as part of the course .
this improved performance does not seem to have affected the number of students who successfully complete the courses , although those that do succeed are doing so at a higher level .
the students that withdraw or fail often do not attend class , or do not put any effort into their studies of the subject .
even if the presentation of the material is improved , it does not have an effect on this group . the effort to stimulate learning in the lowest performing students
is an area that will require further focus .
there is also evidence that the students themselves feel that the approach of _ the physics suite _ has positively impacted their learning .
responses to the survey questions and statements made during the interviews indicate a perception among the students that the inquiry - based activities of _ realtime physics _ and the active participation in lecture required in _ interactive lecture demonstrations _
play a large role in helping them grasp the difficult material of the introductory physics classes .
the success of the project through the first two years , based on the evaluations done and also on positive feedback from the students themselves indicates that there is good reason to continue with , and expand , the implementation of _ the physics suite _ at the university of central florida .
the authors would like to extend thanks for the help provided by professors david sokoloff , ronald thornton and priscilla laws , who have been the consultants in the implementation of the _ the physics suite _ at ucf , and professor ken krane , who has been the evaluator .
the lab manuals and concept inventory tests used for physics i and ii have been created by david sokoloff , ronald thornton and priscilla laws . @xcite .
this work was supported by the department of undergraduate education in the national science foundation under grant no .
due 0633157 .
s. mccole was supported by an reu grant of the national science foundation ( dms 0649159 ) .
department of education national center for education statistics , http://nces.ed.gov/fastfacts/ wftv , orlando , fl , http://www.wftv.com/education/ university of central florida , ucf today , http://today.ucf.edu/ucf-moves-up-now-nations-3rd-largest/ university of central florida college of engineering and computer science , http://www.cecs.ucf.edu/about/ american institute of physics , aip pub .
r-394.14 , http://www.aip.org/statistics/ national science foundation , http://www.nsf.gov/news/ e. f. redish , _ teaching physics with the physics suite _ , wiley 2003 .
karen cummings , priscilla laws , edward f. redish , patrick cooney , david sokoloff , ronald thornton , _ the physics suite _ , john wiley and sons , inc .
david sokoloff , ronald thornton .
_ interactive lecture demonstrations , active learning in introductory physics _ , john wiley and sons , inc . 2006 .
ronald thornton , david sokoloff .
_ assessing student learning of newton s laws : the force and motion conceptual evaluation and the evaluation of active learning laboratory and lecture curricula _ , am .
april 1998 .
david sokoloff , ronald thornton , priscilla laws . _ real time physics active learning laboratories : module 1 mechanics _ , john wiley and sons , inc .
david sokoloff , ronald thornton , priscilla laws . _ real time physics active learning laboratories : module 3 electric circuits _ , john wiley and sons , inc . 2004 .
david sokoloff , ronald thornton , priscilla laws . _ real time physics active learning laboratories : module 4 light and optics _ , john wiley and sons , inc
j. clement , d. brown , a. zeitsman , _ not all preconceptions are misconceptions : finding `` anchoring conceptions '' for grounding instruction on students intuition _ ,
11 ( spec .
issue ) , 554 - 565 ( 1989 ) .
r. r. hake , _ socratic pedagogy in the introductory physics laboratory _
33 , 1 - 7 ( 1992 ) .
r. r. hake , _ interactive engagement versus traditional methods : a six - thousand student survey of mechanics test data for introductory physics courses _ , am . j. phys .
january 1998 .
l. c. mcdermott , p. s. shaffer , _ research as a guide for curriculum development : an example from introductory electricity .
part i. investigation of student understanding _ , am .
60 , 994 - 1003 ( 1992 ) ; erratum , 61 , 81 ( 1993 ) .
david sokoloff , ronald thornton , _ using interactive lecture demonstrations to create an active learning environment _ , phys .
35 , 340 - 347 ( 1997 ) .
david sokoloff , ronald thornton , _ learning motion concepts using real - time microcomputer - based laboratory tools _ , am .
58 , 858 - 867 ( 1990 ) .
e. f. redish , j. m. saul , r. n. steinberg , _ on the effectiveness of active - engagement microcomputer - based laboratories _ , am .
65 , 45 - 54 ( 1997 ) .
l. c. mcdermott , m. l. rosenquist , e. h. van zee , _ student difficulties in connecting graphs and physics : examples from kinematics _ , am .
55 , 503 - 513 ( 1987 ) .
f. m. goldberg , j. h. anderson , _ student difficulties with graphical representations of negative values of velocity _ , phys .
27 , 254- 260 ( 1989 ) .
r. j. beichner , _ testing student interpretation of kinematics graphs _ , am .
62 , 750 - 762 ( 1994 ) . | a key question in physics education is the effectiveness of the teaching methods .
a curriculum that has been investigated at the university of central florida ( ucf ) over the last two years is the use of particular elements of _ the physics suite_.
select sections of the introductory physics classes at ucf have made use of _ interactive lecture demonstrations _ as part of the lecture component of the class .
the lab component of the class has implemented the _ realtime physics _
curriculum , again in select sections .
the remaining sections have continued with the teaching methods traditionally used . using pre- and post - semester concept inventory tests , a student survey , student interviews , and a standard for successful completion of the course , the preliminary data indicates improved student learning . |
the detection of spectral signatures associated with the environment of a @xmath0-ray burst ( grb ) would provide important clues about the triggering mechanism and the progenitor ( mszros & rees 1998 ; lazzati , campana & ghisellini 1999 ; bttcher 2000 ) .
observations with _
chandra _ , _ asca _ and _
beppo_sax have provided tentative evidence for fe k@xmath1 line and edge features in at least five bursts ; grb 970508 ( piro et al . 1999 ) , grb 970828 ( yoshida et al . 1999 ) , grb 991216 ( piro et al .
2000 ) and grb 000214 ( antonelli et al .
2000 ) all show an emission feature during the x - ray afterglow a few hours to a day after the burst event ; while grb 990705 ( amati et al .
2000 ) displays a prominent x - ray absorption feature during the burst itself .
although most of the line detections are only marginally significant and fail to distinguish between the various line excitation mechanisms , grb 991216 shows a [email protected] kev line at a moderate confidence level ( @xmath3 ) .
this is consistent with emission from h - like fe ( fe xxvi ) at the redshift of the most distant absorption system along the line of sight at @xmath4=1.02 ( vreeswijk et al .
2000 ) . a similar observation , but with higher signal - to - noise , may be able to distinguish between the various line emission mechanisms and lead to the correct determination of the line energy .
the large equivalent widths ( @xmath5 a few kev ) inferred from the x - ray features favor models in which the line is produced by reflection ( vietri et al . 2000 ; rees & mszros 2000 ) , rather than transmission .
any detection of emission features in the afterglow spectra some hours after the burst ( such as in grb 991216 ) therefore imposes strong constraints on the location and geometry of the optically thick reflecting material .
observing an x - ray line at a time @xmath6 after the burst implies that the emitting material must be located within a distance @xmath7 from the explosion site , thus strongly limiting the size of the remnant .
however , this gas can not be optically thick along the line of sight because this would smear out the short - time variability of the burst radiation .
these conditions point towards a strongly anisotropic environment from which a grb is seen only if we happen to observe the system through a line of sight with low optical depth ( bttcher 2000 ; vietri et al .
2000 ; lazzati et al .
2001 ) .
two types of reflection models have been developed to explain the origin of the x - ray emission features .
the first invokes the interaction of the primary x - ray emission from the afterglow with an fe - enriched , thomson - thick , asymmetric remnant , which compton scatters x - rays into our line of sight ( vietri et al .
2000 ; bttcher & fryer 2000 ) .
this scenario requires a mass @xmath8 of fe at a distance about 1.5 light days , possibly due to a remnant of an explosive event or a supernova that occurred days or weeks prior to the grb ( see vietri & stella 1998 ; in contrast with macfadyen & woosley 1999 which favors a simultaneous supernova explosion ) .
the other type of model involves a long - lived ( @xmath9 ) magnetar or accreting black hole with a continuing but decaying outflow that interacts with the stellar envelope at distances less than a light - hour ( rees & mszros 2000 ; mszros & rees 2001 ) . in this case
only a small mass of fe is required , and can be readily produced by the star itself . under both interpretations
, it is likely that reflection takes place in highly ionized surfaces .
this can lead to strong comptonization of the emergent fe line , and other absorption and emission features . in this paper , we present and discuss detailed , self - consistent computations of the temperature and ionization structure of a uniform slab of gas ionized by the incident radiation of a grb and of the resulting reflection spectra .
our analysis applies to an optically thick , homogeneous medium , significantly extending previous analysis in the optically thin regime ( e.g. weth et al .
2000 ; bttcher 2000 ) , which is hard - pressed to explain the observed fe - line feature in grb 991216 ( vietri et al .
we estimate the strength of the fe k@xmath1 line that each model produces .
finally , we discuss the implications of these results for current and future x - ray observations .
we employ the reflection code developed by ross , weaver & mccray ( 1978 ) and updated by ross & fabian ( 1993 ) .
we consider the illumination of the first 12 thomson depths of a infinite , uniform slab of gas by radiation with a power - law spectrum of photon index @xmath10 incident at an angle @xmath11 to the normal ( ross , fabian & young 1999 ) .
the incident radiation is treated analytically in a ` one - stream ' approximation , while the diffuse radiation that results from both the compton scattering of the incident radiation and emission from the gas itself , is treated using the fokker
planck / diffusion method of ross et al .
once thermal and ionization equilibrium in the slab is found , the reflection spectrum is computed .
we assume that hydrogen and helium are completely ionized , but include the partially ionized species c v vii , o v ix , mg ix xiii , si xi
xv , and fe xvi xxvii . for a given @xmath10 , the temperature and ionization state of the surface of the slab is expected to depend mainly on the value of the ionization parameter , @xmath12 where @xmath13 is the total illuminating flux ( from 0.01 - 100 kev ) and @xmath14 is the hydrogen number density ( ross et al .
if the grb explodes within a young supernova ( sn ) remnant , the x - rays from the afterglow will illuminate fe - rich ( @xmath5 ten times solar fe abundance ) material , leading to recombination line emission by reflection ( vietri et al .
reasonable values of @xmath15 expected in this scenario are @xmath16 where @xmath17 is the source luminosity in units of erg s@xmath18 , @xmath14 is the hydrogen density in units of @xmath19 , @xmath20 is the distance from the burst explosion to the material in units of cm .
we adopt the convention @xmath21 for expressing the physical parameters .
the time delay of a few days observed in grb 991216 yields an incident angle for the radiation of @xmath22 ( vietri et al .
2000 ) .
alternatively , line emission can be produced when a post - burst outflow , possibly magnetically dominated , impacts on the walls of a funnel excavated in the stellar envelope ( se ) at distances less than a light - hour ( rees & mszros 2000 ) .
luminosities as high as @xmath23 erg s@xmath18 are expected 1 day after the burst , if they are due either to the spinning down millisecond pulsar or to a highly magnetized torus around a black - hole ( rees & mszros 2000 ) .
the ionization parameter in the stellar envelope case is @xmath24 where @xmath25 is the ratio of ionizing to mhd luminosity ( rees & mszros 2000 ) .
the incident flux would be deflected along the funnel walls with different incident angles before escaping the funnel : for simplicity we assume @xmath22 .
for these uniform - density slabs we vary @xmath15 by changing the total illuminating flux while keeping the hydrogen number density @xmath14 , the distance from the burst @xmath20 and the incident angle @xmath11 fixed .
1 shows the results for illumination by a @xmath10=2 spectrum for both of the scenarios described above .
the illuminating and reflected spectra are displayed as @xmath26 , where @xmath27 is the spectral energy flux and @xmath28 is the photon energy .
the models with the highest ionization parameter ( @xmath29 ) are excellent reflectors , and show almost no fe spectral features .
this is because the surface layer is almost fully ionized , and fe xxvi does not become dominant until @xmath30 .
the temperature at the surface of the slab is @xmath5 1.9 @xmath31 k , the compton temperature for the incident spectrum
. the spectrum drops at @xmath5 50 kev because we included a sharp energy cut - off in the incident spectrum at 100 kev considered here , extending the illuminating spectrum to higher energies ( @xmath32 kev ) would have little effect on the iron ionization and spectral features ( e.g. , ross et al .
1999 ) . ] .
compton reflection produces a slight steeping in the reflected spectrum , which mimics a power law with @xmath33 in the 3 - 30 kev band . for example
, we find @xmath10=2.3 in the @xmath34 sn case and @xmath10=2.14 in the @xmath34 se case . when the illuminating flux is reduced so that @xmath35 , fe k@xmath1 emission and fe k - shell absorption features begin to appear .
most of the k@xmath1 photons originate at @xmath36 and emerge as broad comptonized lines with weak cores .
the h - like k@xmath1 photons at 6.97 kev are generated close to the surface , but are resonantly trapped and removed from the narrow line core by compton scattering .
the he - like inter - combination line at 6.7 kev is not subject to resonant trapping , but is generated at such high @xmath37 ( see panel ( b ) in fig .
2 ) that it is multiply compton scattered on leaving the slab . the broad comptonized emission features blended into the compton smeared k - shell absorption edge , are an important signature of ionized reflection .
the effect of increasing the iron abundance at a fixed ionization parameter is illustrated by the dotted line in fig .
1 . the comptonized line and the absorption feature are increasingly significant for an fe - rich medium . with @xmath38 ,
fe xxv becomes dominant at @xmath39 ( see panel ( a ) in fig .
narrow emission line resulting from fe xxv can now be seen superimposed on the compton - broadened emission bump .
the tiny emission feature just above 8.8 kev results from radiative recombination directly to the ground level of fe xxv ( see the @xmath40 se case in fig 2 ) . in the model with @xmath41 the fe emission
is suppressed because fe xvii - xxii dominates close to the surface , but their k@xmath1 photons are destroyed by the auger effect during resonance trapping ( ross , fabian & brandt 1996 ) .
finally , for @xmath42 the reflection spectrum is similar to that of a cold , neutral slab , and so the narrow emission line at 6.4 kev is dominant . the effect of increasing the incident angle @xmath11 , which is a key ingredient in the pre - ejection models because it is responsible for the observed time - delay , is also illustrated in fig . 1 by the dot - dashed line .
radiation that illuminates the atmosphere more directly ionizes more deeply into the slab than radiation at grazing incidence . at high flux levels the emergent line features
are little changed , however , the k - shell absorption features become more prominent for radiation that impacts closer to the normal of the slab . the ionization structure in the outer layers of the illuminated slab
also depends on the incident radiation spectrum .
3 shows the fe k@xmath1 equivalent width ( ew ) as a function of incident luminosity for a series of reflection spectra with @xmath431.6 , 2.0 & 2.4 , assuming @xmath44 .
the ews were calculated with respect to the reflection spectrum , and the integration was carried out between 5.7 and 7.1 kev . as seen in fig .
1 , the ew of the fe k@xmath1 line decreases with @xmath17 in both grb scenarios .
this behavior continues until @xmath17 is sufficiently low for the narrow fe k@xmath1 emission line to be suppressed by the auger effect .
the ews are generally larger when the illuminating spectrum is steeper ( i.e. , @xmath10 is greater ) .
the weaker ionizing power of steep spectra allow line emission to persist at the highest luminosities , although the ews still end up quite low when @xmath45 . in a grb
, softer spectra may be important if both the line emission arises from the impact of the continuous energy output with the compact remnant and much of the observed continuum emission ( with a flatter @xmath10 ) comes from the afterglow emission directly .
the emission feature observed @xmath5 1.5 days after the grb 991216 burst ( piro et al . 2000 ) had a line luminosity of @xmath46 erg s@xmath18 and an equivalent width ( ew ) of @xmath5 0.5 kev .
the continuum flux from grb 991216 , measured to have @xmath47 , in the 110 kev band was 50100 times stronger than the flux in the line .
as is clear from the above discussion , the emission feature can be explained by reflection if identified with the recombination k@xmath1 line from he - like iron at 6.7 kev .
taking @xmath48 in the se scenario , a continuous ionizing luminosity of @xmath49 erg s@xmath18 ( or a continuous wind luminosity of @xmath50 erg s@xmath18 with @xmath51 ; see rees & mszros 2000 ) would be sufficient to produce the observed line inferred by piro et al .
( 2000 ) . ] .
however , it is possible that the reflected and incident spectra are observed together . with the above parameters , the line luminosity and ew calculated for the total spectrum ( reflected + incident ) are @xmath52 erg s@xmath18 and ew @xmath5 0.250.7 kev . if the illuminating spectrum is a steeper power - law ( @xmath53 2 ) , then a smaller fraction of illuminating photons lie in the 920 kev range which dominates the ionization of fe , and thus an increase in either @xmath17 , @xmath54 or the fe abundance is required in order to reproduce the same line strength ( fig .
3 ) . alternatively ,
an x - ray afterglow with @xmath48 illuminating the walls of a supernova remnant with a luminosity of @xmath55 would produce a line signal with @xmath56 erg s@xmath18 and ew @xmath5 0.20.8 kev .
however , an essential assumption contained in this model is that the material responsible for emission - line features is illuminated by the early afterglow radiation ( or the grb itself ) with @xmath57 .
such high luminosities would cause the spectral features resulting from iron to disappear or , at best , to be extremely weak ( even for larger values of @xmath10 , see fig .
moreover , this early incident radiation can be harder than a @xmath58 spectrum with a significant fraction of the energy above the @xmath59 formation energy threshold , and a high compactness parameter .
this will cause new pairs to be formed in the originally optically thick scattering medium , an effect which amplifies the density of scattering charges and increases the temperature of the illuminating material .
when pairs are produced in sufficient numbers , the iron k@xmath1 emission is suppressed due to the decrease in the number of recombinations .
these effects will be investigated elsewhere ( lazzati , ramirez - ruiz & rees 2001 , in preparation ) .
these problems may be overcome in particular source geometries for which lower luminosities and softer spectra are expected at the edges of the relativistic outflow .
the results presented in fig . 1 show that if the observed fe k@xmath1 line from grb 991216 is identified solely with the h - like line at 6.97 kev , as suggested by piro et al .
( 2000 ) , then this is inconsistent with emission from a photoionized optically thick slab .
their detailed analysis of the _ chandra _ acis - s spectrum shows marginal evidence ( at the @xmath60 level ) of an emission feature at 4.4 @xmath2 0.5 kev .
identifying this feature with the fe recombination edge with rest energy of 9.28 kev gives @xmath4= 1.1 @xmath2 0.1 , consistent with the redshift of a h - like fe k@xmath1 line .
nonetheless , it is important to emphasize that the observations presented by piro et al .
( 2000 ) do not rule out the presence of a he - like fe feature ( or a blend between h and he - like features as shown in fig .
some ambiguity also remains in the rest energy of the emission since doppler blue - shifts of order @xmath61 are expected in both reflection scenarios ( vietri et al .
2000 ; rees & mszros 200 ; mszros & rees 2001 ) and may confuse precise measurements of the line energy . indeed , for an expansion velocity of @xmath61 , the emission feature at 4.4 @xmath2 0.5 kev could be attributed to the fe xxv radiative recombination emission just above 8.8 kev rather than to the fe recombination edge at 9.28 kev .
finally , co produces both a he and h - like emission line : at 7.242 kev and just above 7.526 kev , respectively .
the exact strength of these features , which we do not include in our calculations , could be important in both of the scenarios discussed here and may confuse the identification of low signal - to - noise spectral features .
we have presented calculations of x - ray reflection from thomson - thick slabs for conditions which may arise in the immediate environment of a grb .
comparisons between putative fe k@xmath1 lines detected in x - ray afterglows and those predicted by the computations will be useful in distinguishing between the various line emission mechanisms . we expect that more sensitive data on x - ray afterglow spectral features will impose strong constraints on the nature of grb progenitors and their environments .
we thank m. j. rees , p. mszros , d. lazzati , a. blain and the referee for useful comments and suggestions .
we are particularly grateful to a. c. fabian and r. r. ross for very helpful insights regarding the calculations .
err acknowledges support from conacyt , sep and the ors foundation .
drb thanks the commonwealth scholarship and fellowship plan and the natural sciences and engineering research council of canada for support .
amati , l. , et al . , 2000 , science , 290 , 953 antonelli , a. , et al . , 2000 , apj , 545 , l39 bttcher , m. , 2000 , apj , 539 , 102 bttcher , m. & fryer c. l. , 2001 , apj , 547 , 338 lazzati , d. , campana , s. & ghisellini , g. , 1999 , mnras , 304 , l31 lazzati , d. , ghisellini , g. , amati , l. , frontera , f. , vietri , m. & stella , l. , 2001 , apj , 556 , 471 macfadyen , a.i .
& woosley , s.e . , 1999 ,
apj , 524 , 262 mszros , p. & rees , m. j. , 1998 , mnras , 299 , l10 mszros , p. & rees , m. j. , 2001 , apj , 556 , l37 piro , l. , et al . , 1999 ,
apj , 514 , l73 piro , l. , et al . , 2000 , science , 290 , 955 rees , m. j. & meszaros , p. , 2000 ,
apj , 545 , l73 ross , r. r. , 1979 , apj , 233 , 334 ross , r. r. & weaver , r. , mccray , r. , 1978 , apj , 219 , 292 ross , r. r. & fabian , a.c . , 1993 , mnras , 261 , 74 ross , r. r. , fabian , a.c . & brandt , w.n . , 1996 , mnras , 278 , 1082 ross , r. r. , fabian , a.c . &
young , a.j .
, 1999 , mnras , 306 , 461 vietri , m. , ghisellini , g. , lazzati , d. , fiore , f. & stella , l. , 2000 , apj , 550 , l43 .
vietri , m. & stella , l. , 1998 , apj , 507 , l45 .
weth , c. , meszaros , p. , kallman , t. & rees , m. j. , 2000 , apj , 534 , 581 yoshida , a. , et al . , 1999 , a&as , 138 , 433 . | recent observations of several @xmath0-ray burst ( grb ) afterglows have shown evidence for a large amount of x - ray line emitting material , possibly arising from ionized iron .
a significant detection of an x - ray spectral feature , such as that found in the _ chandra _ observation of grb 991216 , may provide important constraints on the immediate environment of the burst and hence on progenitor models . the large fe k@xmath1 equivalent widths inferred from the x - ray observations favor models in which the line is produced when the primary x - ray emission from the source strikes thomson - thick material and compton scatters into our line of sight .
we present such reflection spectra here , computed in a fully self - consistent manner , and discuss the range of ionization parameters that may be relevant to different models of grbs .
we argue that the presence of a strong hydrogen - like k@xmath1 line is unlikely , because fe xxvi photons would be trapped resonantly and removed from the line core by compton scattering .
in contrast , a strong narrow emission line from he - like fe xxv is prominent in the model spectra . we briefly discuss how these constraints may affect the line energy determination in grb 991216 . |
understanding brain function requires knowing connections between neurons . however , experimental studies of inter - neuronal connectivity are difficult and the connectivity data is scarce . at the same time
neuroanatomists possess much data on cellular morphology and have powerful techniques to image neuronal shapes .
this suggests using morphological data to infer inter - neuronal connections .
such inference must rely on rules which relate shapes of neurons to their connectivity .
the purpose of this paper is to derive such a rule for a frequently encountered feature in the brain organization : a topographic projection .
two layers of neurons are said to form a topographic projection if adjacent neurons of the input layer connect to adjacent neurons of the output layer , figure [ fig : cd ] . as a result ,
output neurons form an orderly map of the input layer .
i characterize inter - neuronal connectivity for a topographic projection by divergence and convergence factors defined as follows , fig .
[ fig : cd ] . _ divergence _
, @xmath0 , of the projection is the number of output neurons which receive connections from an input neuron . _ convergence _ , @xmath1 , of the projection is the number of input neurons which connect with an output neuron .
i assume that these numbers are the same for each neuron in a given layer .
furthermore , each neuron makes the required connections with the nearest neurons of the other layer . in most cases ,
this completely specifies the wiring diagram . a typical topographic wiring diagram shown in fig .
[ fig : cd ] misses an important biological detail . in real brains ,
connections between cell bodies are implemented by neuronal processes : axons carrying nerve pulses away from the cell bodies and dendrites carrying signals towards cell bodies.@xcite therefore each connection is interrupted by a synapse which separates an axon of one neuron from a dendrite of another .
both axons and dendrites branch away from cell bodies forming arbors . in general , a topographic projection with given divergence and convergence
may be implemented by axonal and dendritic arbors of different sizes , which depend on the locations of synapses .
for example , consider a wiring diagram with @xmath2 and @xmath3 , figure [ fig : menorah]a .
narrow axonal arbors may synapse onto wide dendritic arbors , figure [ fig : menorah]b or wide axonal arbors may synapse onto narrow dendritic arbors , figure [ fig : menorah]c .
i call these arrangements type i and type ii , correspondingly
. the question is : which arbor sizes are preferred ?
i propose a rule which specifies the sizes of axonal arbors of input neurons and dendritic arbors of output neurons in a topographic projection : _ high divergence / convergence ratio favors wide axonal and narrow dendritic arbors while low divergence / convergence ratio favors narrow axonal arbors and wide dendritic arbors .
_ alternatively , this rule may be formulated in terms of neuronal densities in the two layers : _ sparser layer has wider arbors .
_ in the above example , divergence / convergence ( and neuronal density ) ratio is 1/6 and , according to the rule , type i arrangement , figure [ fig : menorah]b , is preferred . in this paper
i derive a quantitative version of this rule from the principle of wiring economy which can be summarized as follows .
@xcite,@xcite,@xcite,@xcite,@xcite space constraints require keeping the brain volume to a minimum .
because wiring ( axons and dendrites ) takes up a significant fraction of the volume , evolution has probably designed axonal and dendritic arbors in a way that minimizes their total volume .
therefore we may understand the existing arbor sizes as a result of wiring optimization . to obtain the rule
i formulate and solve a wiring optimization problem .
the goal is to find the sizes of axons and dendrites which minimize the total volume of wiring in a topographic wiring diagram for fixed locations of neurons .
i specify the wiring diagram with divergence and convergence factors . throughout most of the paper
i assume that the cross - sectional area of dendrites and axons are constant and equal .
therefore , the problem reduces to the wiring _ length _ minimization .
my results are trivially extended to the case of unequal fiber diameters as shown below .
purves and coworkers@xcite,@xcite,@xcite have previously reported empirical observations which may be relevant to the present theory .
they found a correlation between convergence and complexity of dendritic arbors in sympathetic ganglia .
conclusive comparison of this data with the theory requires establishing topographic ( or some other ) wiring diagram and measuring axonal arbor sizes in this system . in the next section
i consider a one - dimensional version of the problem . in this case , wirelength is minimized by wide dendritic and no axonal arbors ( type i ) for divergence less than convergence and by no dendritic and wide axonal arbors ( type ii ) in the opposite case . in section [ sec:2d
] , i consider a two - dimensional version of the problem .
if both convergence and divergence are much greater than one , the optimal ratio of dendritic and axonal arbors equals to the square root of convergence / divergence ratio .
i test the rule on the available anatomical data in section [ sec : data ] . for several projections between retinal and cerebellar neurons ,
arbor sizes agree with the rule . in section [ sec : other ] i consider other factors which may affect arbor sizes .
consider two parallel rows of evenly - spaced neurons , figure [ fig : cd ] , with a topographic wiring diagram characterized by divergence , @xmath0 , and convergence , @xmath1 .
the goal is to find axonal and denritic arbor sizes which minimize the combined length of axons and dendrites .
i compare different arbor arrangements by calculating wirelength per unit length of the rows , @xmath4 .
i assume that input / output rows are close to each other and include in the calculation only those parts of the wiring which are parallel to the neuronal rows .
i start by considering a special case where each input neuron connects with only one output neuron ( @xmath2 ) , figure [ fig : menorah]a . there are two limiting arrangements satisfying the wiring diagram : type i has wide dendritic arbors and no axonal arbors , figure [ fig : menorah]b ; type ii has wide axonal arbors and no dendritic arbors , figure [ fig : menorah]c . intuitively , the former arrangement has smaller wirelength : short axons synapsing onto a common bus - like dendrite is better than long axons from each input neuron synapsing onto a short dendrite . to confirm this
i calculate wirelength in the two extreme arrangements for @xmath2 , see methods .
@xmath5 @xmath6 these results show that for @xmath2 and @xmath7 the two arrangements have the same wirelength . for @xmath2 and @xmath8 the arrangement with wide dendritic arbors and no axonal arbors ( type i ) has smaller wirelength than the arrangement with wide axonal arbors and no dendritic arbors ( type ii ) .
i can readily apply this result to another special case , @xmath9 , by invoking the symmetry of the problem in respect to the direction of the signal propagation .
i can interchange words `` axons '' and `` dendrites '' and variables @xmath0 and @xmath1 in the derivation and use the above argument . for @xmath9 and @xmath10 the two extreme arrangements have the same wirelength , while for @xmath11 the arrangement with wide axonal arbors ( type ii ) has shorter wiring than the arrangement with wide dendritic arbors ( type i ) .
next , i consider the case when both convergence and divergence are greater than one ( @xmath12 ) .
for the two extreme arrangements i get ( see methods ) : @xmath13 @xmath14 comparison of the two expressions reveals the following .
if divergence is less than convergence then the optimal arrangement has wide dendritic and no axonal arbors ( type i ) .
if divergence is greater than convergence then the optimal arrangement has wide axonal and no dendritic arbors ( type ii ) .
if convergence and divergence are equal both arrangements have the same wirelength .
i can restate this result by using the identity between the divergence / convergence ratio and the neuronal density ratio ( see methods ) : in the optimal arrangement the sparser layer has wide arbors , while the denser layer has none .
so far i compared extreme arrangements with wide arbors in one row and none in the other .
what about intermediate arrangements , with both axonal and dendritic arbors of non - zero width ? to address this question
i consider the limit of large divergence and convergence factors ( @xmath15 .
i find wirelength as a function of the axonal arbor size @xmath16 ( see methods ) : @xmath17 because @xmath18 i find the following .
if @xmath19 then the minimal wirelength is achieved when @xmath20 , arrangement with wide dendritic and no axonal arbors ( type i ) .
if @xmath21 then the minimal wirelength is achieved when @xmath22 , arrangement with wide axonal and no dendritic arbors ( type ii ) .
if @xmath23 then all possible axonal arbor widths give the same wirelength .
this proves that for @xmath24 extreme arrangements minimize wirelength . in cases of small @xmath1 and
@xmath0 i checked intermediate solutions one by one . in many cases intermediate arrangements
have the same wirelength as the extreme solution .
however , only for a few `` degenerate '' @xmath25 pairs there are equally good intermediate arrangements with the reverse ratio of average axonal and dendritic arbor sizes relative to the extreme solution .
my results are conveniently summarized on the phase diagram , figure [ fig : phase ] , which shows optimal arrangements for various pairs of divergence and convergence factors .
i mark the `` degenerate '' @xmath25 pairs by diamonds on the phase diagram , figure [ fig : phase ] .
what if axons and dendrites have different crossectional areas ?
the principle of wiring economy requires that wire volume rather than wire length should be minimized .
i can modify the arguments of this section by including the cross - sectional areas of the processes .
i find for @xmath26 that if divergence / convergence ratio is less than the ratio of axonal and dendritic cross - sections then the optimal arrangement has wide dendritic and no axonal arbors ( type i ) . in the opposite case
i find wide axonal and no dendritic arbors ( type ii ) .
consider two parallel layers of neurons with densities @xmath27 and @xmath28 .
the topographic wiring diagram has divergence and convergence factors , @xmath0 and @xmath1 , requiring each input neuron to connect with @xmath0 nearest output neurons and each output neuron with @xmath1 nearest input neurons .
again , the problem is to find the arrangement of arbors which minimizes the total length of axons and dendrites . for different arrangements i compare the wirelength per unit area , @xmath4 .
i assume that the two layers are close to each other and include only those parts of the wiring which are parallel to the layers .
i start with a special case where each input neuron connects with only one output neuron ( @xmath2 ) .
consider an example with @xmath29 and neurons arranged on a square grid in each layer , figure [ fig : por]a .
two extreme arrangements satisfy the wiring diagram : type i has wide dendritic arbors and no axonal arbors , figure [ fig : por]b ; type ii has wide axonal arbors and no dendritic arbors , figure [ fig : por]c .
i take the branching angles equal to @xmath30 , an optimal value for constant crossectional area.@xcite assuming `` point '' neurons the ratio of wirelength for type i and type ii arrangements : @xmath31 thus , the type i arrangement with wide dendritic arbors has shorter wire length .
this conclusion holds for other convergence values much greater than one , provided @xmath2 . however , there are other arrangements with non - zero axonal arbors that give the same wire length
. one of them is shown in figure [ fig : por]d .
degenerate arrangements have axonal arbor width @xmath32 where the upper bound is given by the approximate inter - neuronal distance .
this means that the optimal arbor size ratio for @xmath2 @xmath33 by using the symmetry in respect to the direction of signal propagation i adapt this result for the @xmath9 case . for @xmath34 ,
arrangements with wide axonal arbors and narrow dendritic arbors ( @xmath35 ) have minimal wirelength .
these arrangements have arbor size ratio @xmath36 next , i consider the case when both divergence and convergence are greater than one . due to complexity of the problem
i study the limit of large divergence and convergence ( @xmath26 ) .
i find analytically the optimal layout which minimizes the total length of axons and dendrites .
unlike the one - dimensional projection , optimal sizes of both axons and dendrites turn out to be non - zero .
notice that two neurons may form a synapse only if the axonal arbor of the input neuron overlaps with the dendritic arbor of the output neuron in a two - dimensional projection , figure [ fig : ar4 ] .
thus the goal is to design optimal dendritic and axonal arbors so that each dendritic arbor intersects @xmath1 axonal arbors and each axonal arbor intersects @xmath0 dendritic arbors . to be specific , i consider a wiring diagram with convergence exceeding divergence , @xmath37 ( the argument can be readily adapted for the opposite case ) .
i make an assumption , to be verified later , that dendritic arbor diameter @xmath38 is greater than axonal one , @xmath16 . in this regime
each output neuron s dendritic arbor forms a sparse mesh covering the area from which signals are collected , figure [ fig : ar4 ] .
each axonal arbor in that area must intersect the dendritic arbor mesh to satisfy the wiring diagram .
this requires setting mesh size equal to the axonal arbor diameter . by using this requirement
i express the total length of axonal and dendritic arbors as a function of only the axonal arbor size , @xmath16 .
then i find the axonal arbor size which minimizes the total wirelength .
details of the calculation are given in methods .
here , i give an intuitive argument for why in the optimal layout both axonal and dendritic size are non - zero .
consider two extreme layouts . in the first one , dendritic arbors have zero width , type ii . in this arrangement axons
have to reach out to every output neuron . for large convergence , @xmath39 , this is a redundant arrangement because of the many parallel axonal wires whose signals are eventually merged . in the second layout , axonal arbors are absent and dendrites have to reach out to every input neuron . again , because each input neuron connects to many output neurons ( large divergence , @xmath40 ) many dendrites run in parallel inefficiently carrying the same signal .
a non - zero axonal arbor rectifies this inefficiency by carrying signals to several dendrites along one wire .
i find that the optimal ratio of dendritic and axonal arbor diameters equals to the square root of the convergence / divergence ratio , or , alternatively , to the square root of the neuronal density ratio : @xmath41 since i considered the case with @xmath37 this result also justifies the assumption about axonal arbors being smaller than dendritic ones . for arbitrary axonal and dendritic cross - sectional areas , @xmath42 and @xmath43
, expressions of this section are modified .
the wiring economy principle requires minimizing the total volume occupied by axons and dendrites resulting in the following relation for the optimal arrangement : @xmath44 notice that in the optimal arrangement the total axonal volume of input neurons is equal to the total dendritic volume of the output neurons .
this theory predicts a relationship between the con-/divergence ratio and the sizes of axonal and dendritic arbors .
i test these predictions on several cases of topographic projection in two dimensions .
the predictions depend on whether divergence and convergence are both greater than one or not .
therefore , i consider the two regimes separately .
first , i focus on topographic projections of retinal neurons whose divergence factor is equal or close to one . because retinal neurons use mostly graded potentials the difference between axons and dendrites is small and i assume that their cross - sectional areas are equal .
the theory predicts that the ratio of dendritic and axonal arbor sizes must be greater than the square root of the input / output neuronal density ratio , @xmath45 ( eq.[sql ] ) .
i represent the data on the plot of the relative arbor diameter , @xmath46 , vs. the square root of the relative densities , @xmath47 , ( figure [ fig : data ] ) . because neurons located in the same layer may belong to different classes , each having different arbor size and connectivity , i plot data from different classes separately .
all the data points lie above the @xmath48 line in agreement with the prediction .
second , i apply the theory to cerebellar neurons whose divergence and convergence are both greater than one .
i consider a projection from granule cell axons ( parallel fibers ) onto purkinje cells .
ratio of granule cells to purkinje cells is 3300,@xcite , indicating a high convergence / divergence ratio .
this predicts a ratio of dendritic and axonal arbor sizes of 58 .
this is qualitatively in agreement with wide dendritic arbors of purkinje cells and no axonal arbors on parallel fibers .
quantitative comparison is complicated because the projection is not strictly two - dimensional : purkinje dendrites stacked next to each other add up to a significant third dimension .
naively , given that the dendritic arbor size is about 400@xmath49 m eq.[res ] predicts axonal arbor of about 7@xmath49 m .
this is close to the distance between two adjacent purkinje cell arbors of about 9@xmath49 m . because the length of parallel fibers is greater than 7@xmath49 m absence of axonal arbors comes as no surprise . in general , application of the rule
requires some care because it was derived for a simplified model .
i considered a topographic projection only between a single pair of layers .
however , neurons often make connections to different layers .
in particular , dendritic arbors of the output layer may be determined by connections other than to the input layer .
for example , consider the topographic projection from thalamus to the primary visual cortex .
one may think that because the density of magnocellular thalamic afferents is smaller than neurons in layer 4c@xmath50 ( @xmath51 compared with @xmath52)@xcite then the axonal arbors should be wider than the dendritic ones .
although this is true ( @xmath53@xcite compared with @xmath54@xcite ) the majority of inputs to layer 4c@xmath50 are intra - cortical.@xcite therefore , the dendritic arbor size may be determined by these other projections .
one may argue that dendrites and axons have functions other than linking cell bodies to synapses and , therefore , the size of the arbors may be dictated by other considerations .
although i can not rule out this possibility , the _ primary _ function of axons and dendrites is to connect cell bodies to synapses in order to conduct nerve pulses between them . indeed ,
if neurons were not connected more sophisticated effects such as non - linear interactions between different dendritic inputs could not take place .
hence the most basic parameters of axonal and dendritic arbors such as their size should follow from considerations of connectivity .
another possibility is that the size of dendritic arbors is dictated by the surface area needed to arrange all the synapses .
this argument does not specify the arbor size , however : a compact dendrite of elaborate shape can have the same surface area as a wide dendritic arbor .
finally , agreement of the predictions with the existing anatomical data suggests that the rule is based on correct principles .
further extensive testing of the rule is desirable .
violation of the rule in some system would suggest the presence of other overriding considerations in the design of that system , which is also interesting . in conclusion ,
i propose a rule relating connectivity of neurons to their morphology based on the wiring economy principle .
this rule may be used to infer connections between neurons from the sizes of their axonal and dendritic arbors .
i frequently use the following identity@xcite relating convergence / divergence ratio and neuronal densities ratio : @xmath55 to prove it , i calculate the number of connections ( or synapses , if connections are monosynaptic ) per unit length in two ways .
the number of connections ( or synapses ) is the number of input neurons , @xmath27 , times divergence , @xmath0 . at the same time , the number of connections ( or synapses ) is the number of output neurons , @xmath28 , times convergence , @xmath1 .
since the answer should not depend on the argument , @xmath56 and eq.[cnd ] follows trivially .
first , consider the case of @xmath2 .
in type i arrangement ( figure [ fig : menorah]b ) , the size of a dendritic arbor , @xmath38 , is the inter - neuronal spacing @xmath57 times the number of inter - neuronal intervals covered by the arbor , @xmath58 : @xmath59 the number of dendritic arbors per unit length is equal to the density of output neurons @xmath28 . the combined dendritic arbor length per unit length is @xmath60 . since the axonal arbors do not contribute , the total wire length per unit length : @xmath61 by using eq.[cnd ] and recalling that @xmath2 i get eq.[li ] . in type ii arrangement ( figure [ fig : menorah]c ) ,
the wire length is equal to the sum of the lengths of axons converging on each output neuron multiplied by the neuronal density in the output layer @xmath28 : @xmath62 using eq.[cnd ] i express the result in terms of convergence alone @xmath63 and get eq.[lii ] . now consider the case of @xmath12 . by using eq.[cnd ]
i find from eq.[lnn ] that @xmath64 this is eq.[dd ] of the main text . by using the symmetry in respect to the direction of signal propagation
i find eq.[cc ] of the main text .
next , i consider an arrangement with arbitrary sizes of axonal , @xmath16 , and dendritic , @xmath38 , arbors , figure [ fig : ar1 ] , in the limit of @xmath26 . to satisfy the wiring diagram
each input neuron must connect with @xmath0 output neurons and each output neuron must connect with @xmath1 input neurons .
this places a constraint on the sum of axonal and dendritic arbor widths : @xmath65 therefore , axonal arbor width can take values @xmath18 .
the total wirelength per unit length is : @xmath66 using eq.[cnd],[ss ] i get eq.[cll ] of the main text .
i consider the case of @xmath67 , fig.[fig : ar4 ] .
the following calculation is valid to the leading order in @xmath0 and @xmath1 .
i omit numerical factors of order one which depend on the precise geometry of axonal and dendritic arbors .
the total length of a dendritic arbor , @xmath68 , is equal to the number of periods in the mesh @xmath69 times the mesh size , @xmath16 : @xmath70 the size of the dendritic arbor , @xmath38 follows from expressing convergence as the product of the area covered by the dendritic arbor times the density of input neurons @xmath71 : @xmath72 substituting this into eq.[sa ] i find : @xmath73 the length of an axonal arbor is approximately given by its size : @xmath74 then the total wirelength per unit area is : @xmath75 in order to find the optimal axonal arbor size @xmath16 , i differentiate wirelength in respect to @xmath16 and set the derivative to zero .
@xmath76 solution of this equation gives the optimal size of an axonal arbor , @xmath16 : @xmath77 by using eq.[la ] i get the size of the dendritic arbor , @xmath78 the last two equations combined give eq.[res ] of the main text .
i have benefited from helpful discussions with e.m .
callaway , e.j .
chichilnisky , h.j .
karten , c.f .
stevens and t.j .
i am grateful to a.a .
koulakov for making several valuable suggestions .
i thank g.d .
brown for suggesting that the size of axonal and dendritic arbors may be related to con-/divergence .
this research was supported by the sloan foundation . | i consider a topographic projection between two neuronal layers with different densities of neurons . given the number of output neurons connected to each input neuron ( divergence ) and the number of input neurons synapsing on each output neuron ( convergence ) i determine the widths of axonal and dendritic arbors which minimize the total volume of axons and dendrites .
analytical results for one - dimensional and two - dimensional projections can be summarized qualitatively in the following rule : neurons of the sparser layer should have arbors wider than those of the denser layer .
this agrees with the anatomical data from retinal and cerebellar neurons whose morphology and connectivity are known .
the rule may be used to infer connectivity of neurons from their morphology . 2 |
we thank m. gyulassy for helpful discussions . this work was supported in part by the nsfc under the project no . 10825523 and 11221504 , by self - determined research funds of ccnu from the colleges basic research and operation of moe , helmholtz international center for fair within the framework of the loewe program launched by the state of hesse , and us depart of energy under contract no
. de - ac02 - 05ch11231 and within the framework of the jet collaboration .
r. b. neufeld , i. vitev , b. -w .
zhang , phys .
lett . * b704 * , 590 - 595 ( 2011 ) .
h. xing , y. guo , e. wang and x. -n .
wang , nucl .
a * 879 * , 77 ( 2012 ) .
l. d. mclerran and r. venugopalan , phys .
d * 49 * , 2233 ( 1994 ) .
d. kharzeev , y. v. kovchegov and k. tuchin , phys .
d * 68 * , 094013 ( 2003 ) .
j. jalilian - marian and a. h. rezaeian , phys .
d * 85 * , 014017 ( 2012 ) .
x. n. wang and m. gyulassy , phys .
d * 44 * , 3501 ( 1991 ) ; m. gyulassy and x. n. wang , comput .
commun . * 83 * , 307 ( 1994 ) .
w. -t .
deng , x. -n .
wang , r. xu , phys .
rev . * c83 * , 014915 ( 2011 ) ; phys .
* b701 * , 133 - 136 ( 2011 ) .
j. adams _ et al . _ [ star collaboration ] , phys . lett .
b * 637 * , 161 ( 2006 ) [ nucl - ex/0601033 ] .
r. c. hwa and c. b. yang , phys .
lett . * 93 * , 082302 ( 2004 ) .
b. z. kopeliovich , phys .
b243 * , 141 ( 1990 ) .
s. chatrchyan _ et al . _
[ cms collaboration ] , eur .
j. c * 72 * , 1945 ( 2012 ) .
a. leonidov and v. nechitailo , eur .
j. c * 71 * , 1537 ( 2011 ) [ arxiv:1006.0366 [ nucl - th ] ] .
j. casalderrey - solana and e. iancu , jhep * 1108 * , 015 ( 2011 ) [ arxiv:1105.1760 [ hep - ph ] ] . y. mehtar - tani , c. a. salgado and k. tywoniuk , phys . rev .
lett . * 106 * , 122002 ( 2011 ) [ arxiv:1009.2965 [ hep - ph ] ] . | multiple parton scatterings in high - energy @xmath0 collisions involve multi - parton correlation inside the projectile and color coherence of multiple jets which will lead to nuclear modification of final hadron spectra relative to that in @xmath1 collisions .
such modifications of final hadron spectra in @xmath0 collisions are studied within the hijing 2.1 model . besides parton shadowing and transverse momentum broadening , which influence final hadron spectra at intermediate @xmath2 , modification of parton flavor composition due to multiple scattering and hadronization of many - parton jets system
are shown to lead to suppression of final hadrons at large @xmath2 in @xmath0 collisions at the large hadron collider . in search for the formation of quark - gluon plasma ( qgp ) in high - energy heavy - ion collisions , jet quenching or suppression of high-@xmath2 hadrons @xcite can be used for tomographic study of medium properties @xcite . however , initial multiple parton scatterings in cold nuclei before the formation of qgp can also lead to nuclear modification of final high-@xmath2 hadron spectra that needs to be understood for a quantitative study of qgp properties through jet quenching .
parton shadowing or nuclear modification of parton distributions inside large nuclei is one of the most studied cold nuclear effects .
the qcd evolution , however , dramatically reduces the effect on final high-@xmath2 hadron spectra in @xmath0 collisions @xcite .
modification of parton distributions in the projectile nucleon through parton energy loss @xcite is estimated to have a small effect on hadron spectra in the central rapidity region of high - energy @xmath0 collisions .
these cold nuclear effects have also been studied within the color glass condensate ( cgc ) model @xcite .
however , multiple parton correlations inside the projectile nucleon should also be considered .
energy - momentum and valence quark number conservation alone could modify the momentum and flavor dependence of final - state parton and hadron spectra at large @xmath2 .
hadronization of multiple - jet systems from multiple scattering in @xmath0 collisions could also affect final hadron spectra at intermediate and large @xmath2 . in this paper
, we study the nuclear modification of parton and hadron spectra due to parton shadowing , initial and final - state transverse momentum broadening , final - state parton flavor composition and hadronization of multiple jets within the hijing @xcite monte carlo model .
it is based on a two - component model for hadron production in high - energy nucleon and nuclear collisions .
the soft and hard components are separated by a cut - off @xmath3 in the transverse momentum exchange .
hard parton scatterings with @xmath4 are assumed to be described by the perturbative qcd ( pqcd ) and soft interactions are approximated by string excitations with an effective cross section @xmath5 . in hijing 2.0 @xcite , parton distribution functions ( pdf s )
are given by the gluck - reya - vogt(grv ) parameterization @xcite .
an energy - dependent cut - off @xmath6 and soft cross section @xmath7 are assumed in order to fit experimental values of the total and inelastic cross sections of @xmath8 collisions and the energy - dependence of the central rapidity density of charged hadron multiplicities .
hijing 2.0 can describe most features of hadron production in @xmath1 collisions at colliding energies up to 7 tev at the large hadron collider ( lhc ) @xcite . in @xmath0 and @xmath9 collisions
, one has to consider nuclear modification of initial parton distributions due to multiple scatterings .
hijing 2.0 employes a factorized form of parton distributions in nuclei , @xmath10 where @xmath11 is the momentum scale , @xmath12 the impact - parameter , @xmath13 the momentum fraction carried by the parton @xmath14 and @xmath15 is the impact - parameter - dependent nuclear modification factor as given by the new hijing parameterization @xcite .
it was fit to the centrality dependence of the pseudo - rapidity multiplicity density per participant pair @xmath16 at the relativistic heavy - ion collider ( rhic ) , and also describes the data at the lhc energies as well @xcite .
in addition to modification of initial parton distributions , multiple scatterings inside a nucleus should also lead to transverse momentum ( @xmath17 ) broadening of both initial and final - state partons the @xmath17-broadening is responsible for the cronin effect or enhancement of intermediate-@xmath2 hadron spectra in @xmath0 collisions @xcite .
it is implemented in hijing 2.1 as a @xmath17-kick by each binary nucleon - nucleon scattering to the initial partons of a hard scattering as well as final - state partons produced prior to the current scattering .
the @xmath17-kick for each scattering follows a gaussian distribution and an energy dependence of the width , @xmath18 is assumed to describe the measured cronin effect in @xmath0 collisions .
after the @xmath17-kick , longitudinal momenta of partons are reshuffled pair - wise between projectile and target partons in order to ensure four - momentum conservation .
consequently , introduction of the @xmath17-kick will also influence hadron rapidity distributions .
we have to fine - tune the gluon shadowing parameter @xmath19 @xcite again to describe experimental data on rapidity distributions . shown in fig .
[ fig : dndy ] are hijing 2.1 results on rapidity distributions of charged hadrons in minimum - biased @xmath20 ( upper panel ) and @xmath21 ( lower panel ) collisions ( solid ) at @xmath22 gev and 5.0 tev , respectively , as compared to @xmath20 data from the star experiment @xcite . here
the gluon shadowing parameter is set to @xmath23 .
also shown are hijing2.1 results for @xmath1 collisions ( dashed ) as compared to the phobos data @xcite and @xmath24 collisions without parton shadowing ( dot - dashed ) .
gluon shadowing in hijing 2.1 is seen to suppress the hadron yield in mid and forward rapidity region , especially at the lhc energy .
measurement of the hadron rapidity distribution at lhc therefore would provide further constraints on the gluon shadowing .
and @xmath1 ( dashed ) collisions at @xmath22 gev ( upper ) and 5.0 tev ( lower ) as compared to data from the star @xcite and phobos @xcite experiment.,scaledwidth=40.0% ] the hard processes in hijing are based on lowest order pqcd in which the single jet inclusive cross section , @xmath25 is directly proportional to nuclear parton distributions @xmath26 . here @xmath27 are fractional momenta of the initial partons , @xmath28 are rapidities of the final - state parton jets and @xmath29 accounts for higher - order corrections .
the nuclear thickness function is normalized to @xmath30 . in the central rapidity region ( @xmath31 ) , transverse momenta of final - state jets determine the momentum fractions @xmath32 of the initial colliding partons .
therefore , @xmath2 spectra of produced partons via hard scatterings in @xmath0 collisions should depend on the nuclear modification of parton distributions in nuclei . to study parton shadowing and other cold nuclear effects
, we calculate the nuclear modification factor for final - state parton and hadron @xmath2 spectra , @xmath33 where @xmath34 is the average number of binary nucleon - nucleon interactions in @xmath0 collisions .
spectra in @xmath24 collisions at @xmath22 gev ( upper ) and 5.0 tev ( lower ) from hijing 2.1 with ( solid ) , without ( dashed ) parton shadowing and with de - coherent hard scatterings ( dhc ) ( dot - dashed and doted).,scaledwidth=40.0% ] spectra in @xmath24 collisions at @xmath22 gev ( upper ) and 5.0 tev ( lower ) from hijing 2.1 with ( solid ) , without ( dashed ) parton shadowing and with de - coherent hard scatterings ( dhc ) ( dot - dashed and doted).,scaledwidth=40.0% ] shown in fig .
[ rparton ] are parton modification factors from the hijing 2.1 for minimum - biased @xmath20 and @xmath21 collisions at @xmath22 gev and 5.0 tev , respectively .
the enhancement of parton spectra at intermediate @xmath2 is the signature effect of @xmath17-broadening , which should disappear at large @xmath2 . in the hijing model , hard scatterings are simulated each time the projectile parton interact with a target parton alongside soft interactions with the remnants of colliding nucleons .
hijing ensures energy - momentum conservation by subtracting energy - momentum transfers in the previous hard and soft interactions from the projectile and thus restricting energy available for subsequent hard collisions .
since the time scale of hard scatterings is much shorter than soft interactions , such a coupling between hard and soft interactions might not be physical .
as an option in our study , we turn off such a coupling and do not restrict energy available for hard scatterings .
we denote the results as dhc ( de - coherent hard scattering ) .
one can see from fig .
[ rparton ] , both parton shadowing and soft - hard coupling suppresses high-@xmath2 parton production .
these features of parton spectra will be reflected in the final hadron spectra after hadronization . in @xmath1 (
dotted ) @xmath24 collisions at the rhic ( upper ) and lhc ( lower ) energies from hijing 2.1 without ( solid ) and with ( dashed ) de - coherent hard scatterings ( dhc).,scaledwidth=40.0% ] in @xmath1 ( dotted ) @xmath24 collisions at the rhic ( upper ) and lhc ( lower ) energies from hijing 2.1 without ( solid ) and with ( dashed ) de - coherent hard scatterings ( dhc).,scaledwidth=40.0% ] collisions at rhic and @xmath21 at lhc .
the ratio in @xmath1 collisions is also shown ( long dashed ) .
the rhic experimental data are from star @xcite .
, scaledwidth=40.0% ] collisions at rhic and @xmath21 at lhc .
the ratio in @xmath1 collisions is also shown ( long dashed ) .
the rhic experimental data are from star @xcite .
, scaledwidth=40.0% ] in @xmath0 collisions , the projectile proton scatters with multiple nucleons inside the target , each with finite probability of hard parton scattering .
the leading contribution comes from independent hard scatterings , each involving independent initial partons from projectile and target nucleons . in principle
, a fast projectile parton can also suffer multiple hard scatterings which will affect the final - state parton spectra but do not increase the projectile s contribution to the final - state parton multiplicity .
momentum correlation of multiple partons inside the projectile is neglected beyond the conservation of the total momentum .
flavor conservation , however , will limit the availability of valence quarks from the projectile for each of these multiple independent hard scatterings as the projectile plunges through the target nucleus .
such limitation on the number of valence quarks available for each independent hard scattering will lead to a change in the flavor composition of produced partons per average binary nucleon - nucleon collision .
shown in fig .
[ qgratio ] are the ratios of produced quark and gluon @xmath2 spectra in the central rapidity region of @xmath20 and @xmath21 collisions at @xmath22 gev and 5.0 tev , respectively , as compared to that in @xmath1 collisions .
the quark number conservation in multiple scatterings is seen to lead to smaller quark / gluon ratio in @xmath0 relative to @xmath1 collisions .
the suppressed quark yield and high quark / gluon ratio , especially at rhic , will lead to the modification of final - state parton spectra at large @xmath2 as shown in fig .
[ rparton ] . because gluon fragmentation functions are softer than quarks , the increased fraction of gluon jets in @xmath0 collisions will lead to apparent suppression of high @xmath2 hadron spectra relative to @xmath1 collisions .
the modified flavor composition of the final parton due to multiple scattering in pa collisions could also change the flavor composition of final hadrons due to flavor - dependence of the fragmentation functions .
gluon jets in general have higher proton - to - pion ratio than in quark jets .
hijing has a strong gluon shadowing which suppresses larger @xmath2 gluon jets .
this should reduce the proton - to - pion ratio in hijing 2.1 results as compared to the results without gluon shadowing as shown in fig .
[ p - pi - ratio ] .
this reduction of proton - to - pion ratio by gluon shadowing is much stronger at the lhc energy .
nevertheless , the proton - to - pion ratio in @xmath20 collisions at rhic and @xmath21 collisions at lhc will still be larger than that in @xmath1 collisions because of the enhanced gluon - to - quark ratio in @xmath0 collisions relative to that in @xmath1 as shown in fig .
[ qgratio ] .
such a trend is already seen in the star data @xcite at rhic .
the flavor composition of final hadrons in @xmath0 collisions could also be influenced by other nonperturbative mechanisms such as parton recombination during hadronization @xcite and hadronic final - state rescattering @xcite .
another process that can also modify final hadron spectra in @xmath0 collisions is jet hadronization or fragmentation .
hijing follows the same routine as in pythia that groups jet partons together with project and target remnants to form color - singlet string systems . in hijing ,
jet parton showers from all hard scatterings , including initial and final - state radiations , are ordered in rapidity and gluons are then connected to the valence quark and diquark of the associated projectile or target nucleons as kinks to form string systems .
these string systems are hadronized according to the lund string model @xcite .
the projectile can undergo multiple scatterings in @xmath0 collisions .
its string system therefore has more gluons attached as compared to that in @xmath1 collisions .
hadrons from such a string system are expected to be softer than independent fragmentation of individual gluons . to test the sensitivity of nuclear modification of final hadron spectra to jet hadronization
, we also consider the independent fragmentation mechanism in hijing for both @xmath1 and @xmath0 collisions .
collisions from hijing 2.1 with different options : default ( solid ) , dhc ( dashed ) , dhc without parton shadowing ( dot - dashed ) and dhc without parton shadowing and independent fragmentation ( dotted ) , as compared to data from the phenix and star experiment @xcite .
the arrowed lines indicate the most possible behavior of the nuclear modification factors at lhc between low and high @xmath2 regions.,title="fig:",scaledwidth=40.0% ] collisions from hijing 2.1 with different options : default ( solid ) , dhc ( dashed ) , dhc without parton shadowing ( dot - dashed ) and dhc without parton shadowing and independent fragmentation ( dotted ) , as compared to data from the phenix and star experiment @xcite .
the arrowed lines indicate the most possible behavior of the nuclear modification factors at lhc between low and high @xmath2 regions.,title="fig:",scaledwidth=40.0% ] shown in fig .
[ rhadron ] are nuclear modification factors for charged hadrons in @xmath20 and @xmath21 collisions at @xmath22 gev and 5.0 tev , respectively , as compared to existing data from both the phenix and star expok , eriment at rhic @xcite . at low @xmath2 , hadron production is dominated by hadronization of soft strings which is proportional to the number of participant nucleons @xmath35 .
the nuclear modification factor should approach @xmath36 in the limit @xmath37 . at large @xmath2 ,
incoherent hard parton scatterings become dominate and @xmath38 without any nuclear modification .
the peak in the nuclear modification factor is generally known as the cronin enhancement of hadron spectra at intermediate @xmath2 due to @xmath17-broadening through multiple scatterings .
such a peak has been observed in pa collisions at the fermilab energies ( @xmath39 gev ) @xcite and the position of the peak at around 4 gev/@xmath40 has no apparent energy dependence and remains the same in hijing results at both rhic and lhc energies .
nuclear shadowing of parton distributions at lhc will only affect the overall value of @xmath41 at intermediate @xmath2 , but the peak - feature and its position remain the same .
this is a unique prediction of the @xmath17-broadening due to multiple scattering as implemented in hijing2.1 .
it is in sharp contrast to the prediction of disappearance of the peak in gluon saturation model or relocation to much higher @xmath2 due to strong anti - shadowing in the collinear pqcd parton model
. experimental test of these features will shed light on the parton dynamics in high - energy nuclei . at very large @xmath2 ,
parton shadowing in nuclei , soft - hard coupling and enhanced gluonic jets due to valence quark conservation all lead to suppression of charged hadron spectra .
fragmentation of the projectile string system with multiple gluon jets in @xmath0 collisions seems to further suppress high-@xmath2 hadron spectra as compared to the independent fragmentation ( dotted ) .
these nuclear effects are consistent with existing @xmath20 data at rhic
. however , @xmath21 collisions at lhc will be able to test these scenarios within the available @xmath2 range .
since parton shadowing will disappear at large @xmath2 @xcite due to qcd evolution and hard scatterings are de - coherent from soft interactions , the nuclear modification factor at lhc will likely follow the default and dhc results at low @xmath2 and approach dhc without shadowing at large @xmath2 with possible further modifications due to hadronization of multiple jets . in fig .
[ rhadron ] , we indicate this trend by joining these two set of results with arrowed lines at intermediate @xmath2 tev from hijing 2.1 with different options : default ( solid ) , dhc ( dashed ) , dhc without parton shadowing ( dot - dashed ) and dhc without parton shadowing and independent fragmentation ( dotted ) , as compared to data from the alice and cms experiment @xcite .
jet quenching due to final - state interaction is not considered in the hijing calculations.,scaledwidth=40.0% ] tev from hijing 2.1 with different options : default ( solid ) , dhc ( dashed ) , dhc without parton shadowing ( dot - dashed ) and dhc without parton shadowing and independent fragmentation ( dotted ) , as compared to data from the alice and cms experiment @xcite .
jet quenching due to final - state interaction is not considered in the hijing calculations.,scaledwidth=40.0% ] the initial - state interaction and fragmentation of multiple jet systems should also affect the final hadron spectra in high - energy heavy - ion collisions in addition to jet quenching due to final - state interaction . the existing experimental data on hadron spectra modification should provide some constraints on the initial - state parton interactions and the final hadronization mechanism . shown in fig .
[ raa ] are nuclear modification factors for final parton ( upper panel ) and hadron spectra in central @xmath42 collisions at @xmath43 tev .
note that there is no jet quenching effect in the calculation shown . on the parton level
, the nuclear modification at large @xmath2 mainly comes from parton shadowing which should be small if the scale dependence of shadowing is taken into account .
the final hadron spectra at large @xmath2 are however significantly suppressed due to the lund string fragmentation of multiple parton systems ( solid and dashed curves in the lower panel ) leaving no room for additional suppression by jet quenching as compared to experimental data from alice @xcite and cms @xcite .
this is a strong indication that the lund string fragmentation for hadronization is no longer a valid mode for multiple gluonic jet systems in heavy - ion collisions where a deconfined quark - gluon plasma is formed .
recent studies show that multiple parton scatterings inside a quark - gluon plasma can in fact change the the color flow of the jet parton shower from that in vacuum @xcite .
one therefore should expect the fragmentation processes in heavy - ion collisions to be significantly modified from that in vacuum .
an independent jet fragmentation mechanism such as parton recombination model might be a better alternative where local parton phase distributions determine the final hadron spectra and global color flow of the multiple jet system is no longer important .
since multiple gluonic jet systems in pa collisions can be considered as precursors of the quark - gluon plasma , investigation of the final hadron spectra at large @xmath2 in @xmath0 collisions can shed light on the hadronization mechanism of energetic jets after they escape jet quenching and emerge from the quark - gluon plasma in high - energy heavy - ion collisions . in summary
, we have studied the nuclear modification of hadron spectra in @xmath0 collisions at both the rhic and lhc energies within hijing2.1 monte carlo model and their sensitivities to cold nuclear effects .
multiple parton correlations in the projectile due to valence quark number conservation , string fragmentation of multiple gluon jets as well as parton shadowing can all lead to suppression of high-@xmath2 hadron spectra in @xmath0 relative to @xmath1 collisions .
comparisons between hijing2.1 results without jet quenching and existing data on large @xmath2 hadron suppression at lhc indicate an independence fragmentation mechanism for multiple jet systems such as in pa and aa collisions .
experimental test of these effects in @xmath21 collisions at lhc and its implications for hard scatterings in @xmath9 collisions is crucial for the study of jet quenching and extraction of jet transport properties in the hot quark - gluon plasma . |
it is noteworthy that except for one or two possible exceptions in the meson sector , the hundreds of known hadrons can be described as bound states of three quarks ( @xmath0 ) , in the case of baryons , or a quark and anti - quark ( @xmath6 ) , in the case of mesons .
there have been reports of possible evidence of mesons with exotic @xmath7 quantum numbers , not possible for @xmath6 , but until 2003 , no reports of baryons with quantum numbers inconsistent with @xmath0 .
there were searches in the 1960 s and 1970 s for what was then referred to as the @xmath8 a baryon with positive strangeness .
the review by hey and kelly @xcite discusses these early searches , mostly in bubble chamber experiments , and a more recent review by trilling of the current exotic baryon sightings @xcite lists publications reporting on bubble chamber experiments with incident @xmath9 and @xmath10 beams searching for the @xmath8 .
those experiments found no enhancements in @xmath11 baryon channels .
indeed the failure to find such flavor exotic baryons in the 1960 s and 1970 s lent credence to the then nascent quark model . in 1997 , d. diakonov , v. petrov and m. polyakov @xcite , using a chiral soliton model , predicted a @xmath12 exotic anti - decuplet of pentaquark baryons ( see figure [ pentagram ] ) including a @xmath13 baryon with positive strangeness , isospin zero and a mass of 1530 mev/@xmath5 with a width of 15 mev/@xmath5 .
evidence for that state , the @xmath14 , has been reported by several experiments . that anti - decuplet is also predicted to include a @xmath15 and a @xmath16 for
which evidence has also been reported .
yet another experiment claims an anti - charm pentaquark baryon , @xmath17 .
these findings have generated much excitement in the nuclear and particle physics theory community nearly 300 papers have been written on the interpretation of these states and their properties . but as will be discussed below , these purported discoveries have also inspired recent searches that fail to find evidence for these states .
the predicted anti - decuplet @xcite of pentaquark baryons indicating , in particular , the states @xmath18 , @xmath15 and @xmath16 .
evidence for these states have been presented as well as for a @xmath17 .
other searches for these states have yielded null results . ]
table [ posex ] lists the experiments claiming evidence for pentaquark states .
there are eleven experiments claiming a @xmath14 , a state with @xmath11 , a mass of about 1.54 gev/@xmath5 and a width consistent with being less than the mass resolutions of the the experiments .
the state is observed through its decay into either @xmath19 or @xmath20 .
the first five reactions listed in table [ posex ] used a photon probe of relatively low energy , a few gev , and the purported pentaquark is observed in the @xmath19 mode .
the first sighting of the @xmath14 was from the leps experiment at spring8 @xcite followed shortly thereafter by the clas experiment at jefferson lab @xcite .
the saphir experiment is also a low - energy photon experiment @xcite .
llllcl experiment&reaction&state&mode&reference + leps(1)&@xmath21&@xmath14&@xmath19&@xcite + leps(2)&@xmath22&@xmath14&@xmath19&@xcite + clas(d)&@xmath23&@xmath14&@xmath19&@xcite + clas(p)&@xmath24&@xmath14&@xmath19 & @xcite + saphir&@xmath25&@xmath14&@xmath19 & @xcite + cosy&@xmath26&@xmath14&@xmath20 & @xcite + jinr&@xmath27&@xmath14&@xmath20&@xcite + svd&@xmath28&@xmath14&@xmath20 & @xcite + diana&@xmath29&@xmath14&@xmath20 & @xcite + @xmath30bc & @xmath31&@xmath14&@xmath20 & @xcite + nomad & @xmath31&@xmath14&@xmath20 & @xcite + hermes&quasi - real photoproduction&@xmath14 & @xmath20&@xcite + zeus&@xmath32&@xmath14&@xmath20&@xcite + na49&@xmath33&@xmath34&@xmath35&@xcite + h1&@xmath36&@xmath37&@xmath38&@xcite + the final state neutron in the above five experiments is not detected . in the leps experiment the assumed @xmath19
effective mass is actually the missing mass recoiling against the @xmath39 . in the clas(d )
experiment the assumption is that the final state proton , which is detected , is the spectator nucleon and the @xmath19 effective mass is the mass recoiling against the @xmath40 system . in the clas(p )
experiment the neutron is inferred from missing mass and in the saphir experiment the neutron is inferred from kinematic fitting .
the cosy experiment used a low - momentum proton beam @xcite spanning the momentum range from 2.85 to 3.3 gev/@xmath41 .
the jinr result comes from an analysis of collisions in a propane bubble chamber exposed to a 10 gev/@xmath41 proton beam .
the svd experiment @xcite at ihep studied @xmath42 collisions at 70 gev/@xmath41 .
both the diana and @xmath30bc groups re - analyzed old data from a liquid xenon bubble chamber ( in the case of diana ) @xcite and the cern bebc and the fnal 15-foot chamber ( in the case of @xmath30bc ) @xcite the former using a low - energy @xmath43 beam , the the latter neutrinos .
the nomad experiment at cern was built to search for neutrino oscillations @xcite .
the hermes experiment at desy found evidence for the @xmath14 with quasi - real photons @xcite and zeus claims evidence @xcite for the @xmath14 in @xmath44 collisions .
the evidence for the the @xmath14 presented in references @xcite is in the @xmath19 mode , which is manifestly flavor - exotic while the other experiments report the @xmath14 in the @xmath20 mode , which is a linear combination of @xmath11 and @xmath45 .
several of these experiments studied the @xmath46 spectrum and found no evidence for a @xmath47 , thus concluding that @xmath48 for the @xmath14 . at first glance
, the evidence for the @xmath14 seems convincing .
the experiments claim signals with a statistical significance ranging from 4 to 7 standard deviations over background ( we will return to this issue later ) .
the evidence comes from low - energy and high energy experiments and experiments with a variety of beams : photons ( real and quasi - real ) , electrons , protons , neutrinos and positively charged kaons .
evidence for the pentaquark cascade states , @xmath49 and @xmath50 , comes from a single experiment , na49 at cern @xcite , in the @xmath51 and @xmath52 modes in proton - proton collisions at @xmath53 gev .
the reported mass is 1.862 gev/@xmath5 and the width is consistent with being below detector resolution . and
evidence for the anti - charmed pentaquark , @xmath37 , also comes from a single experiment , h1 at hera @xcite , in the @xmath54 and @xmath55 modes at a mass of 3.1 gev/@xmath5 and a width consistent with detector resolution .
the data were collected in @xmath44 collisions at @xmath56 of 300 and 320 gev .
table [ negex ] lists recent experiments that have searched for and failed to find evidence for pentaquark signals . as discussed earlier @xcite
, searches for @xmath11 baryons in the 1960 s and 1970 s also failed to find evidence for their existence .
the experiments in table [ negex ] are listed in alphabetical order by name . for each of the three pentaquark states ( @xmath14 , @xmath34 and @xmath37 ) , we indicate with a @xmath57 that the state was observed or with a @xmath58 that the state was searched for and not observed .
the entry ( @xmath59 ) indicates that the state was not searched for . of the 18 entries in table [ negex ] , 16 searches for the @xmath14 yielded null results .
one experiment , wa89 @xcite , that failed to find the @xmath34 , did not search for the @xmath14 and another experiment , zeus , that searched for and failed to find either the @xmath34 or @xmath37 did claim positive evidence for the @xmath14 @xcite .
all of the experiments ( 9 of them ) searching for the @xmath34 failed to find the state and all of the experiments ( 6 of them ) searching for the @xmath37 failed to find that state . as will be discussed below , the experiments listed in table [ negex ] are characterized by high statistics and excellent mass resolution , of order 1 to 2 mev/@xmath5 in most cases .
these experiments also see the relevant _ benchmark _ states , such as the @xmath60 , @xmath61 , @xmath62 , @xmath63 , @xmath64 or @xmath65 , with statistics that overwhelm experiments that report positive pentaquark sightings . lllllcl experiment&search reaction&@xmath14&@xmath34&@xmath37&reference + aleph&hadronic z decays&@xmath66&@xmath66&@xmath66&@xcite + babar&@xmath67&@xmath66&@xmath66&@xmath68&@xcite + belle&@xmath69&@xmath66&@xmath59&@xmath66 & @xcite + bes&@xmath70&@xmath66&@xmath59&@xmath66 & @xcite + cdf&@xmath71&@xmath66&@xmath66&@xmath66 & @xcite + compass&@xmath72&@xmath66&@xmath66&@xmath73 & @xcite + delphi&hadronic z decays&@xmath66&@xmath73&@xmath73 & @xcite + e690&@xmath74&@xmath66&@xmath66&@xmath73&@xcite + focus&@xmath75&@xmath66&@xmath66&@xmath66 & @xcite + hera - b&@xmath76&@xmath66&@xmath66&@xmath73&@xcite + hypercp&@xmath77&@xmath66&@xmath73&@xmath73 & @xcite + lass&@xmath78&@xmath66&@xmath73&@xmath73 & @xcite + l3&@xmath79&@xmath66&@xmath73&@xmath73 & @xcite + phenix&@xmath80&@xmath66&@xmath59&@xmath73&@xcite + selex&@xmath81&@xmath66&@xmath73&@xmath73 & @xcite + sphinx&@xmath82&@xmath66&@xmath73&@xmath73 & @xcite + wa89&@xmath83&@xmath73&@xmath66&@xmath73 & @xcite + zeus&@xmath84&@xmath85&@xmath66&@xmath66 & @xcite + the aleph , delphi and l3 experiments ran at lep . the aleph @xcite and delphi @xcite searches studied hadronic @xmath86 decays while l3 @xcite searched for the @xmath14 in photon - photon collisions .
the babar , belle and bes experiments are also all carried out at @xmath87 colliders .
babar searched @xcite in final states from @xmath67 as well as 40 mev below resonance while bes @xcite searched in @xmath88 .
and belle searched for @xmath14 in kaon interactions in the detector material @xcite a search that yielded 16000 @xmath89 decays but no @xmath14 signal .
the cdf experiment at fnal searched for all three pentaquark states @xcite in @xmath90 collisions .
the e690 @xcite , focus @xcite , hypercp @xcite and selex @xcite searches were all carried out in fixed target experiments at fnal .
focus used a photon beam produced by bremsstrahlung of 300 gev electrons and positrons , e690 a 800 gev/@xmath41 proton beam , hypercp a mixed beam of @xmath91 , @xmath43 and protons ranging in momentum from 50 to 250 gev/@xmath41 and selex a 600 gev/@xmath41 mixed beam of @xmath9 , @xmath10 and @xmath92 .
hera - b @xcite and sphinx @xcite at hera and ihep ( protvino , russia ) also were fixed target experiments , both using proton beams 900 gev/@xmath41 at hera and 70 gev/@xmath41 at ihep .
compass @xcite is a fixed target experiment at cern using a 160 gev/@xmath41 @xmath93 beam .
the lass spectrometer at slac collected data using a 11 gev/@xmath41 @xmath43 beam and these data were recently re - analyzed @xcite .
wa89 is a fixed target at experiment that used a 340 gev/@xmath41 @xmath92 beam at cern @xcite .
the phenix detector searched for the @xmath14 in @xmath94 collisions @xcite at rhic .
since the @xmath34 and @xmath37 have only been observed by one experiment each , we discuss these states first and then discuss the @xmath14 .
the @xmath34 has only been observed by na49 @xcite and searched for , but not observed , by 9 experiments @xcite as shown in table [ negex ] . in the @xmath51 mode
na49 reports 38 signal ( @xmath95 ) events over a background ( @xmath96 ) of 43 events and thus claim a signal significance ( @xmath97 ) of 4.2 @xmath98 assuming @xmath99 . when they combine events from the @xmath52 mode as well then @xmath100 , @xmath101 and @xmath102 .
na49 observes a total of 1640 @xmath103 in their sample .
by contrast aleph , hera - b , cdf , babar and wa89 have samples that exceed the na49 sample of @xmath64 by factors of 2 , 10 , 22 , 157 and 412 respectively and all but na49 fail to observe the @xmath34 .
perhaps even more relevant are the numbers of @xmath65 states observed .
na49 finds 150 @xmath65 s @xcite while aleph and zeus also see a comparable number but cdf sees 6 times more and babar and e690 see 100 times more .
not only are the statistics of the experiments that do not observe the @xmath34 significantly higher , but the typical mass resolution of 1 to 2 mev/@xmath5 is much better than the resolution of na49 .
it is hard to reconcile the existence of the @xmath34 with this overwhelming negative evidence .
the @xmath37 has only been observed by h1 @xcite and searched for , but not observed , by 6 experiments @xcite as shown in table [ negex ] , h1 sees @xmath104 , @xmath105 and claim @xmath106 .
the @xmath37 is observed in its decay to @xmath54 and @xmath55 .
h1 has @xmath107 @xmath62 s in their sample and by contrast the number of @xmath62 s observed by focus and cdf exceed the h1 sample by factors of 12 and 178 respectively .
the failure of zeus to find the @xmath37 is particularly relevant because h1 and zeus are similar experiments operating at the same accelerator facility .
it is hard to reconcile the existence of the @xmath37 with this overwhelming negative evidence .
as shown in table [ posex ] , the @xmath14 has been observed in 13 reactions where two of the reports are from the leps group @xcite and another two from the clas group @xcite . as shown in table [ negex ] , the @xmath14 has been searched for by 16 separate experiments and all those came up with null results .
an interesting benchmark to compare the positive sightings with the null results is the number of observed @xmath108 decays .
the number of claimed signal events @xmath109 and observed @xmath63 decays @xmath110 for leps(1 ) is @xmath111 19(25 ) . for the other positive sightings , @xmath112 for leps(2 ) , clas(d ) , saphir , hermes and zeus are 56(162 ) , 43(212 ) , 55(530 ) , 51(850 ) and 230(193 ) respectively .
by contrast , the number of @xmath63 decays observed by aleph , babar , belle , cdf , e690 , hera - b and sphinx ( in thousands of events ) are 2.8k , 100k , 15.5k , 3.3k , 5k and 23k .
so , for example , the babar yield of @xmath63 decays exceeds that of leps(1 ) by a factor of 4000 .
it seems difficult to reconcile the positive sightings with the null results since each of the two set of experiments spans a wide variety of production mechanisms . and in general the experiments yielding null results have higher statistics and superior resolution .
since the existence of a flavor - exotic baryon state would have such an important impact on our understanding of hadronic physics it is essential that the positive evidence be examined critically . in particular
, we will discuss * the statistical significance of the purported signals including a discussion of the background estimation ; * the effect of kinematical reflections giving rise to enhancements that could fluctuate to sharp peaks ; * other possibilities of generating sharp peaks including _ ghost tracks _ , kinematic cuts and other mechanisms ; * effects that could impair the isolation of exclusive reactions ; and * problems of width estimates and consistency of masses among experiments .
lcccccc experiment & signal & background & + & @xmath95 & @xmath96 & published & @xmath113 & @xmath114 & @xmath115 + leps(1 ) @xcite & @xmath116 & @xmath117 & @xmath118 & @xmath118 & @xmath119 & @xmath120 + leps(2 ) @xcite & @xmath121 & @xmath122 & & @xmath123 & @xmath124 & @xmath125 + clas(d ) @xcite & @xmath126 & @xmath127 & @xmath128 & @xmath129 & @xmath123 & @xmath130 + clas(p ) @xcite & @xmath131 & @xmath132 & @xmath133 & @xmath134 & @xmath135 & @xmath136 + saphir @xcite & @xmath137 & @xmath121 & @xmath138 & @xmath139 & @xmath128 & @xmath140 + cosy @xcite & @xmath141 & @xmath142 & @xmath143 & @xmath129 & @xmath135 & @xmath144 + jinr @xcite & @xmath145 & @xmath146 & @xmath147 & @xmath148 & @xmath149 & @xmath150 + svd @xcite & @xmath132 & @xmath151 & @xmath152 & @xmath153 & @xmath154 & @xmath155 + diana @xcite & @xmath156 & @xmath157 & @xmath123 & @xmath123 & @xmath158 & @xmath159 + @xmath30bc @xcite&@xmath160&@xmath161 & @xmath162 & @xmath163 & @xmath130 & @xmath164 + nomad @xcite&@xmath165 & @xmath166 & @xmath140 & @xmath140 & @xmath158 & @xmath159 + hermes @xcite & @xmath167 & @xmath168 & @xmath169 & @xmath170 & @xmath153 & @xmath159 + zeus @xcite & @xmath171&@xmath172 & @xmath118 & @xmath173 & @xmath148 & @xmath135 + table [ statistics ] lists the number of signal events and background events for each of the positive sightings along with the published statistical significance of the @xmath14 .
the simple estimators rely on two numbers .
the first is the number of background counts under the peak , @xmath96 , while the second is the number of signal counts in the peak , @xmath95 above background .
from these two numbers , there are three commonly used significance estimators .
1 . @xmath174 2 .
@xmath175 3 .
@xmath176 the first of these neglects the statistical uncertainty of the background and is normally an over - estimate of the true significance .
this is the most commonly quoted estimator for observed pentaquark signals .
the second estimator assumes a smooth background with a well defined shape .
finally , the third method takes into account the uncertainty in a statistically independent background under the signal .
this latter method should normally underestimate the true significance of an observed signal .
all three estimators are computed for each sighting in table [ statistics ] .
since all of these experiments report on a signal with an unknown production mechanism over a background that is not understood , the best statistical measure is likely to be case ( iii ) , @xmath177 . in figure [ allplots ]
we show the @xmath19 or @xmath178 mass distributions for the experiments listed in table [ posex ] absent any background curves . only the mass region from 1.4 to 1.7
gev/@xmath5 is shown .
error bars are based simply on statistical errors . a comparison of the mass distributions for the @xmath19 or @xmath178 mass distributions for the experiments listed in table [ posex ] .
only the mass region from 1.4 to 1.7 gev/@xmath5 is shown .
error bars are based simply on statistical errors . ] for the reaction @xmath179 the configuration of momentum vectors is shown in figure ( a ) in the @xmath180 rest frame where the boost direction to this frame defines the axis . as discussed in @xcite the decays of mesons , such as the @xmath181 , @xmath182 , @xmath183 and @xmath184 , with their characteristic decay structure and interference with each other , will influence the shape of the @xmath19 ( and @xmath19 ) mass distribution .
the curve shown in figure ( b ) is a possible description of the background for the observed @xmath19 spectrum reported in @xcite and results from a simultaneous fit of the @xmath19 , @xmath185 and @xmath180 effective mass distributions with an assumption that the @xmath181 , @xmath182 , @xmath183 and @xmath184 are produced and then decay into @xmath180 . ]
it was pointed out in reference @xcite that for lower energy photon experiments @xcite the production of mesons , such as the @xmath181 , @xmath182 , @xmath183 and @xmath184 , with subsequent decays into @xmath180 could lead to structure in the the @xmath19 mass distribution .
this comes about because of the decay pattern of the decaying mesons given by @xmath186 and is illustrated in figure [ parent ] . in part ( a ) of figure [ parent ] the momentum vectors in the rest frame of the decaying mesons is shown .
the boost direction is along the line - of - flight of the meson and in this _ helicity frame _ that direction defines the @xmath187axis . when the polar angle @xmath188 is small , the @xmath43 and @xmath189 momenta are nearly collinear and the @xmath19 effective mass is small . for @xmath190
the effective mass is large .
there is strong forward - backward peaking because of the @xmath186 and this reflects in the @xmath19 effective mass spectrum .
but because of the constrained kinematics , owing to the low beam momenta , only the lower peak survives .
the curve shown in figure [ parent](b ) is a possible description of the background for the observed @xmath19 spectrum reported in @xcite and results from a simultaneous fit of the @xmath19 , @xmath185 and @xmath180 effective mass distributions with an assumption that the @xmath181 , @xmath182 , @xmath183 and @xmath184 are produced and then decay into @xmath180 . in 1969
@xcite performed a search for the @xmath8 in @xmath191 at 6 and 8 gev/@xmath41 by looking at the missing mass recoiling against the @xmath39 ( as is the case as well for the leps(1 ) , leps(2 ) and clas(d ) searches ) .
they found a peak at around 1.6 gev/@xmath5 that they eventually ascribed to a kinematic reflection due to production of mesons decaying into @xmath192 .
this was supported by monte carlo studies and the observations that the peak position changed as the beam energy changed . using as a parent distribution , the distribution shown in figure [ parent ] ,
20 histograms were generated with statistics equal to the @xmath19 distribution presented in @xcite also shown in figure ( b ) .
of the 20 histograms , three were selected that had a sharp peak near the mass of @xmath14 .
these are shown in figures ( a ) , ( c ) and ( d ) . ]
although such a mechanism can not produce a sharp peak in the @xmath19 mass distribution it will reduce the statistical significance of a purported signal in the low - mass region where this background peaks . and it could well lead to fluctuations into a sharp peak with low statistics as is illustrated in figure [ fakes ] . using as a parent distribution the distribution shown in figure [ parent ] ,
twenty histograms were randomly generated with statistics equal to the @xmath19 distribution presented in @xcite also shown in part ( b ) of figure [ parent ] . of the 20 histograms generated , three were selected that had a sharp peak near the mass of @xmath14 .
these are shown in parts ( a ) , ( c ) and ( d ) of figure [ fakes ] . the use of this mechanism to estimate the background for the clas(d ) @xcite data would result in a 20% probability of fluctuating into a sharp peak at the @xmath14 mass . of the 13 @xmath14 sightings list in table [ posex ] ,
8 are of the decay @xmath193 and of those all but two , cosy and diana , are inclusive measurements .
it was pointed out by zavertyaev @xcite and longo @xcite that ghost tracks associated with the decay @xmath194 could lead to a spurious sharp spike at precisely 1.54 gev/@xmath5 if @xmath195 2 gev/@xmath41 .
the inclusive experiments should produce @xmath196 s copiously .
we illustrate this mechanism in figure [ fakes ] . in part
( a ) of that figure we show a schematic of a @xmath194 decay where the track reconstruction duplicated the proton track yielding an extra spurious positively charged track that has been assigned the mass of a pion . when combined with the @xmath197 from the @xmath198 the effective mass clusters about 0.5 gev/@xmath5 as in part ( b ) and when the ghost track is combined with the @xmath198 decay products the effective mass clusters around 1.5 gev/@xmath5 as seen in part ( c ) . in the shaded distributions
the @xmath199 mass is required to be near the @xmath200 .
the mean of the shaded portion of the distribution in part ( c ) is 1.54 gev/@xmath5 , the mass of the @xmath14 . in this
study the @xmath198 momentum in the lab frame was uniform from 2 to 100 gev/@xmath41 .
the diana experiment shows a peak at the @xmath14 mass of 1.54 gev/@xmath5 ( see figure [ allplots ] ) .
but as pointed out by zavertyaev @xcite , among the observed reactions should be charge exchange off a neutron , @xmath201 .
this two - body reaction should yield a fixed @xmath20 effective mass if the @xmath43 beam momentum is fixed .
indeed the beam momentum spectrum is consistent with the position of the observed peak in the diana spectrum .
figure ( a ) is a schematic of the decay @xmath202 .
the effect of spurious _ ghost _ tracks from the reconstruction software is considered . in this case
a @xmath91 track is generated . when combined with the @xmath197 from the @xmath198 the effective mass clusters about 0.5 gev/@xmath5 as in figure ( b ) and when
the ghost track is combined with the @xmath198 decay products the effective mass clusters around 1.5 gev/@xmath5 as seen in figure ( c ) . in the shaded distributions
the @xmath199 mass is required to be near the @xmath200 .
the mean of the shaded portion of the distribution in figure ( c ) is 1.54 gev/@xmath5 , the mass of the @xmath14 . in this
study the @xmath198 momentum in the lab frame was uniform from 2 to 100 gev/@xmath41 . ]
the clas(p ) reaction @xmath203 @xcite is an exclusive reaction uncomplicated by nuclear target effects ( see the next subsection ) .
after cuts are made to isolate this reaction the published @xmath19 mass spectrum shows no structure .
but then kinematic cuts are made to remove processes that are potential backgrounds to @xmath14 production such as : 1 .
@xmath204 followed by @xmath205 2 .
@xmath206 followed by @xmath207 and @xmath208 3 .
@xmath209 followed by @xmath207 to remove these `` backgrounds '' the following cuts are made : @xmath210 ( gev/@xmath41)@xmath211 and @xmath212 where @xmath213 is the angle of the @xmath43 in the overall center - of - mass . the result is a @xmath19 mass distribution sharply reduced in statistics but with a peak at the @xmath14 .
furthermore it is claimed that the @xmath214 effective mass shows a peak at 2.4 gev/@xmath5 leading to the conclusion that an important mode for @xmath14 production is @xmath215 followed by @xmath216 .
the authors do not discuss the stability of observed peaks with respect to the kinematic cuts .
we did monte carlo studies of reactions such as @xmath217 and found that the kinematic cuts can produce peaks observed in the clas(p ) data depending on details of the spin - state of the @xmath182 .
stoler also pointed out @xcite that the these data do show an anti - correlation of @xmath14 signal with the @xmath218 signal which is surprising since one might expect associated production of the @xmath14 , _
i.e. _ @xmath219 to be favored .
however , this anti - coorelation would be expected from a kinematical reflection .
the leps(1 ) and leps(2 ) reactions take place off a carbon nucleus leading to complications due to fermi motion of the target nucleon . because only the @xmath43 and @xmath39 are measured
, they then must assume that the unobserved neutron carries off the remaining missing momentum .
in the case of clas(d ) , the @xmath14 is observed in the effective mass of @xmath19 where the @xmath189 is missing .
however , to be able to reconstruct the reaction , the proton is required to have sufficient energy to escape the target .
the proton is required to somehow be involved in the reaction , which leads to complications in understanding both the reaction itself and the backgrounds to the reaction .
the cosy experiment @xcite depends on identification of the reaction @xmath220 but this experiment is highly unconstrained .
this only information about this experiment is a collection of space - points from three pairs of hodoscope planes downstream of a `` point '' target .
there is no momentum measurement , time measurement nor particle identification .
the @xmath200 is identified through its decay into @xmath221 between the first and second hodoscope planes and the @xmath222 is identified as a kink. this experiment is seriously under - constrained and subject to backgrounds . reported masses , with error bars , of the @xmath14 . ]
figure [ masses2 ] shows a plot of the masses reported for the @xmath14 .
the quoted masses are inconsistent with each other although the reported widths are less than experimental resolution and these experiments accurately reproduce masses of established resonances .
cahn and trilling @xcite analyzed the data from the diana experiment and conclude that the width of the @xmath14 has a width of less than 1 mev/@xmath5 .
et al _ @xcite reanalyzed @xmath223 and conclude that the addition of the @xmath14 with a width of about 5 mev/@xmath5 is improbable but the data are consistent with a width of the @xmath14 less than 1 mev/@xmath5 .
the experimental evidence for the pentaquark states @xmath14 , @xmath34 and @xmath37 has been reviewed .
the @xmath14 has been observed in 13 reactions where two of the reports are from the leps group @xcite and another two from the clas group @xcite .
the @xmath14 has been searched for by 16 separate experiments and all those came up with null results .
the positive sightings are complicated by nuclear target effects , mechanisms that can generate spurious peaks and problems with kinematic identification of final states .
furthermore , stringent cuts , the stability of which are not well understood , are required to isolate the peaks . on the other hand , experiments that have searched for the @xmath14 with null results have high statistics and superior resolution .
the reported masses for the @xmath14 are inconsistent with each other and the width is unusually small for a typical hadronic state .
there have been attempts , _
e.g. _ karliner and lipkin @xcite , to understand all these results in terms of models that would suppress pentaquark production in some processes and not others .
if it is real , the @xmath14 is exotic not only in its flavor quantum numbers but also in its production and its decay .
the @xmath34 has only been observed by na49 @xcite and searched for , but not observed , by 9 experiments @xcite .
it is hard to reconcile the existence of the @xmath34 with this overwhelming negative evidence .
the @xmath37 has only been observed by h1 @xcite and searched for , but not observed , by 6 experiments @xcite .
it is hard to reconcile the existence of the @xmath37 with this overwhelming negative evidence .
based on all this , the conclusion is that the experimental evidence for pentaquarks is very weak .
the authors wish to acknowledge helpful discussions with d. christian , m. longo , r. mitchell , m. reyes and s. teige .
this work was supported in part by the department of energy .
99 hey a and kelly r , 1983 _ phys . rep . _
* 96 * 71 trilling g , 2004 _ a possible exotic baryon resonance _ , a review from the review of particle physics pdg.lbl.gov diakonov d , petrov v and polyakov m , 1997 _ z. phys .
_ * 359 , * 305 nakano t _
et al _ , 2003 _ phys .
lett . _ * 91 * ,
012002 nakano t , 2004 presented at qnp2004 www.qnp2004.org stepanyan s _
et al _ ( clas collaboration ) , 2003 _ phys .
lett . _ * 91 * , 252001 - 1 kubarovsky v _
et al _ ( clas collaboration ) , 2004 _ phys .
lett . _ * 92 * , 032001 - 1 barth j _
et al _ ( saphir collaboration ) , 2003 _ phys .
lett . _ * b572 * , 127 abdel - bary m _
et al _ ( cosy collaboration ) , 2004 _ phys .
_ * b595 * , 127 aslanyan p , emelyanenko v and rikhkvitzkaya , 2004 hep - ex/0403044 aleev a _ et al . _ ( svd collaboration ) , 2004 hep - ex/0401024 barmin v _
et al _ , 2003 _ phys .
nucl . _ * 66 * , 1715 asratyan a , dolgolenko a and kubantsev a , 2004 _ phys .
nucl . _ * 67 * , 682 camilleri l , 2004 presented at the _ neutrinos 2004 paris _ , ` neutrino2004.in2p3.fr ` airapetian a _ et al _ ( hermes collaboration ) , 2004 _ phys . lett . _ * b585 * , 213 chekanov s _
et al _ ( zeus collaboration ) , 2004 _ phys .
* b591 * , 7 alt c _ et al _
( na49 collaboration ) 2004 _ phys .
lett . _ * 92 * , 042003 - 1 aktas , a _ et al _
( h1 collaboration ) , 2004 _ phys .
_ * b588 * , 17 schael s _ et al _ ( aleph collaboration ) , 2004 _ phys .
_ * b599 * 1 and cern - ph - ep-2004 - 038 aubert b _ et al _ ( babar collaboration ) , 2004 presented at ichep2004 ichep04.ihep.ac.cn and hep - ex/0408064 mizuk r ( belle collaboration ) , 2004 presented at ichep2004 ichep04.ihep.ac.cn and hep - ex/0411005 bai j _ et al _
( bes collaboration ) , 2004 _ phys .
_ * d70 * 012004 and hep - ex/0402012 wang m - j _ et al . _
( cdf collaboration ) , 2004 presented at qnp2004 www.qnp2004.org ; gorelov i _ et al _ ( the cdf collaboration ) , 2004 presented at dis 2004 www.saske.sk/dis04 and hep - ex/0408025 ; litvintsev d _
et al _ ( the cdf collaboration ) , 2004 presented at beach2004 capp.iit.edu/beach04 and hep - ex/0410024 brona g and badelek b ( compass collaboration ) , 2004 wwwcompass.cern.ch/compass/notes/2004-5 lin c , 2004 2004 presented at ichep2004 ichep04.ihep.ac.cn christian d _ et al _
( e690 collaboration ) , 2004 presented at qnp2004 www.qnp2004.org stenson , k ( focus collaboration ) , 2004 presented at the 2004 dpf meeting hep - ex/0412021 abt i _
et al _ ( hera - b collaboration ) , 2004 _ phys .
_ * 93 * 212003 and hep - ex/0408048 ; knoepfle , k _ et al _
( hera - b collaboration ) , 2004 j. phys .
* g30 * s1363 and hep - ex/0403020 longo m _ et al _ ( hypercp collaboration ) , 2004 presented at qnp2004 www.qnp2004.org and hep - ex/0410027 napolitano j , cummings j and witkowski m , 2004 hep - ex/0412031 armstrong s , 2004 presented at beach2004 capp.iit.edu/beach04 and hep - ex/0410080 pinkenburg c ( phenix collaboration ) , 2004 _ j.phys .
_ * g30 * s1201 and nucl - ex/0404001 engelfried j ( selex collaboration ) , 2004 presented at quark confinement 2004 www.eurocongress.it/quark antipov yu _ et al . _
( sphinx collaboration ) , 2004 _ eur .
j. _ * a21 * 455 and hep - ex/0407026 salur s ( star collaboration ) , 2004 hep - ex/0403009 and nucl - ex/0410039 adamovich m _
( wa89 colaboration ) , 2004 hep - ex/0405042 ( zeus collaboration ) , 2004 presented at ichep2004 ichep04.ihep.ac.cn chekanov s _
et al _ ( zeus collaboration ) , 2004 _ eur.phys.j .
_ * c38 * 29 and hep - ex/0409033 fischer h and wenig s , 2004 hep - ex/0401014 dzierba a , krop d , swat m , teige s and szczepaniak a , 2004 _ phys .
* d69 * 051901 - 1 anderson e _ et al _ , 1969 _ phys .
lett . _ * b29 * , 136 zavertyaev m , 2003 hep - ph/0311250 longo m , 2004 presentation of hypercp at qnp2004 www.qnp2004.org stoler p , 2004 presentation of clas results at qnp2004 www.qnp2004.org cahn r and trilling g , 2004 _ phys . rev . _ * d69 * , 011501 - 1 arndt r , strakovsky i and workman r , 2003 _ phys . rev . _ * c68 * 042201 haidenbauer j and krein g , 2003 _ phys . rev . _ * d68 * 052201 karliner m and lipkin h , 2003 _ phys . lett . _ * b575 * , 249 | several experimental groups have reported evidence for baryons with flavor exotic quantum numbers that can not be explained as @xmath0 bound states but require a minimum of five quarks
@xmath1 .
these pentaquark states include the @xmath2 , the @xmath3 and the @xmath4 .
the reported widths of these new states are consistent with experimental resolution and may be as narrow as a few mev/@xmath5 or less .
prior to 2003 , experimental searches for flavor exotic baryons spanning several decades yielded negative results .
there have also been a number of searches carried out since the reports of these new pentaquark states that do not confirm their existence .
this review of both the positive and negative reports seeks to understand the current situation regarding the experimental evidence for pentaquarks . . |
quantum monte carlo ( qmc ) techniques are among the most powerful and versatile methods for the computer simulation of a large variety of interesting quantum systems encountered in physics , chemistry or biology @xcite . in particular , qmc is capable of delivering numerically exact results . despite the great potential of this method , there are several restrictions and handicaps inherent to all qmc techniques , the perhaps most pressing one being due to the _ sign problem _ @xcite
. there are various sign problems , namely the _ fermionic sign problem _ encountered in equilibrium ( imaginary - time ) simulations of strongly correlated many - fermion systems , and the _ dynamical sign problem _ in real - time ( dynamical ) simulations , which already shows up for a single particle .
unfortunately , apart from variational or approximate treatments ( such as the fixed - node approximation ) , a completely general and totally satisfactory solution to the sign problem in qmc simulations is still lacking .
nevertheless , over the past few years considerable and substantial progress has been achieved along several different lines without introducing approximations into the qmc scheme , see , for instance , refs .
@xcite . in these notes focus is put on one specific class of qmc methods called _ path - integral monte carlo ( pimc)_. pimc is based on a discretized path - integral representation of the quantities of interest .
the sign problem then arises when different paths that contribute to averages carry different sign ( or complex - valued phase ) .
for instance , for the fermionic sign problem , as a consequence of exchange , one typically has to deal with determinants , so that non - positive - definite fermionic density matrix elements arise .
the sign cancellations arising from sampling fermion paths then manifest themselves as a signal - to - noise ratio , @xmath0 , that vanishes exponentially with both particle number @xmath1 and inverse temperature @xmath2 ; here @xmath3 is a system - dependent energy scale .
the small signal surviving the interference of many fermion paths is then inevitably lost in the large background noise of the stochastic simulation .
similarly , when studying the real - time dynamics of even a single particle , the quantum - mechanical time evolution operator @xmath4 attaches a complex - valued phase to each quantum path , which in turn gives rise to the dynamical sign problem .
again the signal - to - noise ratio will vanish exponentially , @xmath5 , where @xmath6 is the maximum real time under study and @xmath7 a system- and implementation - specific characteristic time scale .
the exponential scaling is typical for the `` naive '' approach , where one simple uses the absolute value of the complex - valued weight function for the mc sampling , and includes the phase information in the accumulation . in ref .
@xcite a general scheme for tackling the sign problem in pimc simulations was proposed .
the method has been applied to interacting electrons in a quantum dot @xcite and to the real - time dynamics of simple few - degrees - of - freedom systems @xcite .
the generalization of the algorithm to the case of effective actions which are long - ranged along the trotter or real - time direction along with the application to the dissipative two - state ( spin - boson ) dynamics has been given in refs .
this _ multilevel blocking _ ( mlb ) approach represents the systematic implementation of a simple _ blocking strategy_. the blocking strategy states that by sampling groups of paths ( `` blocks '' ) at the same time , the sign problem can always be reduced compared to sampling single paths as would be done normally ; for a proof , see sec . [ block ] below .
by suitably bunching paths together into sufficiently small blocks , the sign cancellations among paths within the same block can be accounted for without the sign problem , simply because there is no sign problem for a sufficiently small system .
the mlb approach is based on a recursive implementation of this idea , i.e. after building `` elementary '' blocks , new blocks are formed out of these , and the process is then iterated .
this leads to a powerful implementation of the blocking strategy .
this algorithm is able to turn the exponential severity of the sign problem into an `` only '' algebraic one .
this is still difficult enough but in practice implies that significantly larger systems ( lower temperature , longer real time ) can be studied by mlb - pimc than under the naive pimc approach .
nevertheless , it should be stressed that mlb is definitely _ not _ a black - box scheme .
there are several potential pitfalls related to incorrect or inadequate choices of certain mlb parameters , and one needs to be quite careful in applying this technique @xcite . given some experience
, however , it represents a powerful handle to relieve the sign problem , with potential for additional improvement .
the plan for the remainder of these notes is as follows . in sec .
[ mlbsec ] the mlb approach is discussed in detail , first in an intuitive way in sec .
[ block ] , and subsequently on a more formal or technical level in sec .
[ mlbform ] for fermions .
the modifications for real - time simulations are summarized in sec .
[ realtime ] .
this is then followed by a discussion of the performance and the accuracy of this algorithm in sec . [ performance ] .
in sec . [ effact ] , the generalization to effective - action problems is outlined .
to demonstrate the power of this approach , mlb results are presented in sec .
[ appl ] . as an example for the fermionic sign problem ,
low - temperature simulation results for strongly correlated electrons in a 2d quantum dot are shown in sec .
[ appl : dot ] .
the remainder of that section is then concerned with the dynamical sign problem . after presenting results for several simple model systems in sec .
[ appl : simple ] , the dynamics of the dissipative two - state system is discussed in sec .
[ appl : sb ] .
finally , some conclusions can be found in sec .
before diving into the details of the mlb approach , the underlying idea ( `` blocking strategy '' ) will be explained , focusing for simplicity on fermionic imaginary - time simulations . for those interested in working with this method ,
technical details and practical guidelines are provided in secs .
[ mlbform ] to [ performance ] . in the last part
the generalization of mlb to pimc simulations of the effective - action type is described .
let us consider a many - fermion system whose state is described by a set of quantum numbers @xmath8 denoting , e.g. the positions and spins of _ all _ particles .
these quantum numbers can correspond to electrons living on a lattice or in continuous space . for notational simplicity , we focus on calculating the equilibrium expectation value of a diagonal operator or correlation function ( this can be easily generalized ) , @xmath9 where @xmath10 represents either a summation for the case of a discrete system or an integration for a continuous system , and @xmath11 denotes the ( reduced ) density matrix of the system . in pimc applications ,
imaginary time is discretized into @xmath12 slices of length @xmath13 .
inserting complete sets at each slice @xmath14 , and denoting the corresponding configuration on slice @xmath15 by @xmath16 , the diagonal elements of the density matrix at @xmath17 entering eq .
( [ fir ] ) are : @xmath18 . to proceed
, one then has to construct accurate analytical approximations for the short - time propagator .
this formulation of the problem excludes effective actions such as those arising from an integration over the fermions using the hubbard - stratonovich transformation @xcite , since that generally leads to long - ranged imaginary - time interactions .
the mlb approach suitable for such a situation @xcite is described below in sec .
[ effact ] .
since we are dealing with a many - fermion system , the short - time propagators need to be antisymmetrized , leading to the appearance of determinants causing the sign problem . strictly speaking
, the antisymmetrization has to be done only on one time slice , but the intrinsic sign problem is much better behaved if one antisymmetrizes on all time slices .
choosing the absolute value of the product of the short - time propagators in eq .
( [ rho ] ) as the positive definite mc weight function @xmath19 $ ] , one has to keep the sign @xmath20 $ ] associated with a particular discretized path @xmath21 during the accumulation procedure , @xmath22 \phi[x ] a[x ] } { \sum_{x } p[x ] \phi[x ] } \;.\ ] ] assuming that there are no further exclusivity problems in the numerator so that @xmath23 $ ] is well - behaved , one can then analyze the sign problem in terms of the variance of the denominator , @xmath24 where @xmath25 is the number of mc samples taken and stochastic averages are calculated with @xmath12 as the weight function . for the fermion sign problem , where @xmath26 and hence @xmath27 , the variance of the signal is controlled by the size of @xmath28 .
remarkably , one can achieve considerable progress by _ blocking paths together_. instead of sampling single paths along the mc trajectory , one can consider sampling sets of paths called blocks .
under such a blocking operation , the stochastic estimate for @xmath29 takes the form @xmath30 \phi[x ] a[x ] \right ) } { \sum_{b } \left(\sum_{x\in b } p[x ] \phi[x ] \right ) } \;,\ ] ] where one first sums over the configurations belonging to a block @xmath31 in a way that is not affected by the sign problem , and then stochastically sums over the blocks . the summation within a block
must therefore be done non - stochastically , or alternatively the block size must be chosen sufficiently small .
of course , there is considerable freedom in how to choose this blocking .
let us analyze the variance @xmath32 of the denominator of eq .
( [ start2 ] ) .
first define new sampling functions in terms of the blocks which are then sampled stochastically , @xmath33 = \left| \sum_{x \in b } p[x ] \phi[x ] \right| \;,\quad \phi'[b ] = { \rm sgn } \left(\sum_{x \in b } p[x ] \phi[x]\right ) \;.\ ] ] rewriting the average sign in the new representation , i.e. using @xmath34 $ ] as the weight , then inserting the definition of @xmath35 and @xmath36 in the numerator , @xmath37 \rangle = \frac{\sum_{b } p'[b ] \phi'[b ] } { \sum_b p'[b ] } = \frac{\sum_{x } p[x ] \phi [ x ] } { \sum_b p'[b ] } \;,\ ] ] and comparing to the average sign in the standard representation using @xmath19 $ ] as the weight , one finds @xmath38 } { \sum_{b } p'[b ] } \;.\ ] ] by virtue of the schwarz inequality , @xmath39 = \sum_{b } \left| \sum_{x\in b } p[x ] \phi[x ] \right|\leq \sum_{b } \left| \sum_{x\in b } p[x ] \right| = \sum_{x } p[x]\;,\ ] ] it follows that _ for any kind of blocking _ , the average sign improves , @xmath40 .
furthermore , since @xmath41 , eq . ( [ var ] ) implies that @xmath42 and hence _ the signal - to - noise ratio is always improved upon blocking configurations together_. clearly , the worst blocking one could possibly choose would be to group the configurations into two separate blocks , one collecting all paths with positive sign and the other with negative sign . in this case , blocking yields no improvement whatsoever , and the `` @xmath43 '' becomes `` @xmath44 '' in eq .
( [ eqi ] ) .
it is apparent from eq .
( [ eqi ] ) that the blocking strategy provides a systematic handle to reduce the sign problem . in our realization of the blocking strategy ,
a block is defined by all paths that differ only at one time slice , i.e. only @xmath16 is updated with all other @xmath45 being frozen .
a direct implementation of the blocking strategy does indeed improve the sign problem but will not remove its exponential severity .
the reason is simply that for a sufficiently large system , there will be too many blocks , and once the signals coming from these blocks are allowed to interfere , one again runs into the sign problem , albeit with a smaller scale @xmath3 entering the signal - to - noise ratio .
the resolution to this problem comes from the multilevel blocking ( mlb ) approach @xcite where one _ applies the blocking strategy in a recursive manner to the blocks _ again . in a sense ,
new blocks containing a sufficiently small number of elementary ones are formed , and this process is repeated until only one block is left .
each step of this hierarchy is called _ level _ in what follows .
blocks are then defined living on these _ levels _ , and after taking care of the sign cancellations within all blocks on a given ( fine ) level , the resulting sign information is transferred to the next ( coarser ) level . on each step , the blocking strategy ensures that no sign problem occurs _
provided one has chosen sufficiently small block sizes_. by doing this recursively , the sign problem on all the coarser levels can be handled in the same manner .
it is then possible to proceed without numerical instabilities from the bottom up to the top level .
the cancellations arising at the top level will create a sign problem again , which is however strongly reduced .
as is argued below , the resulting sign problem is characterized by an only algebraic severity . in many ways ,
the mlb idea is related to the renormalization group approach .
but instead of integrating out information on fine levels , sign cancellations are `` synthesized '' within a given level and subsequently their effects are transferred to coarser levels . while the renormalization group filters out information judged `` relevant '' and then ignores the `` irrelevant '' part , no such approximation is introduced here .
therefore our approach is actually closer in spirit to the multi - grid algorithm @xcite .
the technical implementation of mlb is discussed next following ref .
@xcite . to keep notation simple
, the slice index @xmath15 is used as a shorthand notation for the quantum numbers @xmath46 . from eq .
( [ rho ] ) the _
level-0 bonds _ , which are simply the short - time propagators , then follow in the form : @xmath47 , where @xmath48 defines the trotter number @xmath49 , have to be specified .
each slice @xmath15 belongs to a unique level @xmath50 , such that @xmath51 and @xmath52 is a nonnegative integer .
for instance , the slices @xmath53 belong to @xmath54 , @xmath55 belong to @xmath56 , etc .
, such that there are @xmath57 ( but @xmath58 ) different slices on level @xmath50 , see fig . [ fig0 ] .
an elementary blocking is achieved by grouping together configurations that differ only at slice @xmath15 , so only @xmath46 varies in that block while all @xmath59 remain fixed .
sampling on level @xmath50 therefore extends over configurations @xmath60 living on the @xmath61 different slices .
in the mlb scheme , one moves recursively from the finest ( @xmath54 ) up to the coarsest level ( @xmath62 ) , and the measurement of the diagonal operator is done only at the top level using the configuration @xmath63 .
a mc sweep starts by changing only configurations associated with the slices on level @xmath64 according to the standard weight @xmath65 generating a mc trajectory containing @xmath66 samples for each slice on level @xmath54 .
these @xmath67 samples are stored and they are used to generate additional coarser interactions among the higher - level slices , @xmath68 \rangle_{{\cal p}_0[m+1 ] } = k^{-1 } \sum_{m+1 } { \rm sgn } [ ( m , m+1)_0 ( m+1,m+2)_0 ] \;,\ ] ] where the summation @xmath69 extends over the @xmath66 samples available for slice @xmath15 . as will be discussed later on , the important mlb parameter @xmath66 subsequently called _ sample number _
should be chosen as large as possible to ensure that the second equality in eq .
( [ l1b ] ) is justified .
the _ level-1 bonds _ ( [ l1b ] ) contain precious and crucial information about the sign cancellations that occured on the previous level @xmath54 . using these bonds , the density matrix ( [ rho ] )
is rewritten as @xmath70 comparing this to eq .
( [ rho ] ) , the entire sign problem has been transferred to the next coarser level by using the level-1 bonds . in the next step ,
the sampling is carried out on level @xmath56 in order to generate the next - level bonds , i.e. only slices @xmath71 are updated , using the weight @xmath72 with @xmath73 moving the level-1 configurations modifies the level-0 bonds , which in turn requires that the level-1 bonds be updated . a direct re - calculation of these bonds according to eq .
( [ l1b ] ) would be too costly . instead
, the stored configurations on level @xmath54 are used to perform an importance sampling of the new level-1 bonds . under the test move @xmath74 ,
i.e. @xmath75 , on level @xmath56 , the bond ( [ l1b ] ) can be obtained from @xmath76 where @xmath69 runs over the previously stored mc configurations @xmath77 .
note that for small values of @xmath66 , eq .
( [ eq8 ] ) is only approximative , and thus a sufficiently large value of @xmath66 should be chosen . with the aid of eq .
( [ eq8 ] ) , the updated level-1 bonds follow with only moderate computational effort . generating a sequence of @xmath66 samples for each slice on level @xmath56 and storing them , _
level-2 bonds _ can be calculated in analogy to eq .
( [ l1b ] ) , @xmath78 \rangle_{{\cal p}_1 { \cal p}_0 } \;.\ ] ] finally , the process is iterated up to the top level @xmath62 using the obvious recursive generalization of eqs .
( [ l1b ] ) and ( [ l2b ] ) to define _
level-@xmath50 bonds_. thereby the diagonal elements of the density matrix are obtained as @xmath79 by virtue of this algorithm , the sign problem is transferred step by step up to the coarsest level .
the expectation value ( [ fir ] ) can thus be computed from @xmath80 the manifestly positive definite mc weight @xmath81 used for the averaging in eq .
( [ expec ] ) can be read off from eq .
( [ rfinal ] ) , @xmath82 the denominator in eq .
( [ expec ] ) gives the average sign and indicates to what extent the sign problem has been solved . for proper choice of mlb parameters , in particular the sample number @xmath66 , this method can solve the sign problem .
the price to pay for the stability of the algorithm is the increased memory requirement @xmath83 associated with having to store the sampled configurations .
the same method , described so far for fermions , can be applied with minor modifications to a computation of real - time correlation functions or occupation probabilities . for concreteness ,
let us focus on an equilibrium time - correlation function , @xmath84 similarly other dynamical properties like the thermally symmetrized correlation function @xcite , @xmath85 with @xmath86 being the partition function , can be computed . in terms of path integrals ,
the traces in ( [ corr ] ) involve two quantum paths , one propagated backward in time for the duration @xmath87 and the other propagated in complex time for the duration @xmath88 .
discretizing each of the two paths into @xmath12 slices , the entire cyclic path has a total of @xmath89 slices .
a slice on the first half of them has length @xmath90 and on the second half @xmath91 . denoting the quantum numbers ( e.g. spin or position variables ) at slice @xmath52 by @xmath92 ,
the correlation function ( [ corr ] ) reads @xmath93 where the level-0 bond @xmath94 is again the short - time propagator between slices @xmath52 and @xmath95 , and @xmath96 .
first assign all slices along the discretized path to different levels @xmath97 , where @xmath49 , in close analogy to the treatment for fermions , see fig .
each slice @xmath98 belongs to a unique level @xmath50 , such that @xmath99 and @xmath100 is a nonnegative integer .
for instance , slices @xmath101 belong to level @xmath102 , slices @xmath103 to @xmath104 , etc .
the mlb algorithm starts by sampling only configurations which are allowed to vary on slices associated with the finest level @xmath102 , using the weight @xmath105 .
the short - time level-@xmath106 bonds are then employed to synthesize longer - time level-@xmath107 bonds that connect the even-@xmath52 slices .
subsequently the level-1 bonds are used to synthesize level-@xmath108 bonds , and so on . in this way
the mlb algorithm moves recursively from the finest level ( @xmath102 ) up to increasingly coarser levels until @xmath109 , where the measurement is done using @xmath110 and @xmath111 . generating a mc trajectory containing @xmath66 samples for each slice on level @xmath54 and storing these samples , the level-1 bonds ( [ l1b ] ) can be computed , where the `` sgn '' has to be replaced by the complex - valued phase @xmath112 = e^{i{\rm arg}(z)}$ ] .
their benefit becomes clear when rewriting the integrand of the denominator in ( [ naive ] ) as @xmath113 comparing this to ( [ naive ] ) , the entire sign problem has been transferred to the next coarser level . in the next step ,
the sampling is carried out on level @xmath56 in order to compute the next - level bonds , using the weight @xmath72 with @xmath114 . generating a sequence of @xmath66 samples for each slice on level @xmath56 , and storing these samples ,
level-2 bonds can be calculated , @xmath115 \rangle_{{\cal p}_0 { \cal p}_1 } \;.\ ] ] this process is then iterated up to the top level .
finally , the correlation function ( [ corr ] ) can be computed from @xmath116 \rangle_{\cal p } } { \langle{\phi}[(\bbox{r}_p,\bbox{r}_{2p})_l ( \bbox{r}_{2p},\bbox{r}_p)_l ] \rangle_{\cal p } } \;,\ ] ] with the positive definite mc weight @xmath117 .
the denominator in eq .
( [ expecta ] ) gives the _ average phase _ and indicates to what extent the sign problem has been solved . with a suitable choice of mlb parameters ,
the average phase remains close to unity even for long times .
next questions concerning the _ exactness _ and _ performance _ of the mlb approach are addressed , in particular the dependence on the sample number @xmath66 .
clearly , @xmath66 needs to be sufficiently large to produce a reliable estimate for the level-@xmath50 bonds . if these bonds could be calculated exactly
corresponding to the limit @xmath118 , the manipulations leading to eq .
( [ rfinal ] ) yield the exact result .
hence for large enough @xmath66 , the mlb algorithm must become exact and completely solve the sign problem . since the level-@xmath50 bonds can however only be computed for finite @xmath66
, the weight function @xmath81 amounts to using a noisy estimator , which in turn can introduce _ bias _ into the algorithm @xcite . in principle , this problem could be avoided by using a linear acceptance criterion instead of the algorithmically simpler metropolis choice @xcite which was employed in the applications reported here . but
even with the metropolis choice , the bias can be made arbitrarily small by increasing @xmath66 .
therefore , with sufficient computer memory , the mlb approach can be made to give numerically exact results .
one might then worry about the actual value of @xmath119 required to obtain stable and exact results .
if the critical value @xmath120 were to scale exponentially with @xmath121 and/or system size , the sign problem would be present in disguise again .
although a rigorous non - exponential bound on @xmath120 has not yet been established , our experience with the mlb algorithm indicates that this scaling is at worst algebraic .
this is corroborated by a recent careful study of this issue @xcite .
therefore the _ exponential severity of the sign problem is replaced by an algebraic one under mlb_. a heuristic argument supporting this statement goes as follows .
if one needs @xmath66 samples for each slice on a given level in order to have satisfactory statistics despite of the sign problem , the total number of paths needed in the naive approach depends exponentially on @xmath12 , namely @xmath122 .
this is precisely the well - known exponential severity of the sign problem under the naive approach .
however , with mlb the work on the last level , which is the only one affected by a sign problem provided @xmath66 was chosen sufficiently large , scales only @xmath123 .
note that it does not scale @xmath124 because one must update the level-@xmath50 bonds on all @xmath48 finer levels as well .
so in mlb , the work needed to sample the @xmath125 paths with satisfactory statistical accuracy grows @xmath126 , i.e. algebraically with @xmath12 .
an important point to mention at this point concerns the high - temperature ( or short - time ) limit , where the direct pimc simulation does not face a significant sign problem . in this case , however , the above - mentioned bias problematics of the mlb - pimc is quite serious and can give erroneous results .
fortunately , since that regime is of little interest to mlb , this is not a serious restriction
. a more detailed discussion of this point can be found in ref .
@xcite . to elucidate how the mlb algorithm works in practice , in table [ tabledot ] simulation results for @xmath127 electrons in a quantum dot at various values of @xmath66 are listed . for details , see sec .
[ appl : dot ] .
compared to the naive approach where @xmath128 , using a moderate @xmath129 already increases the average sign from @xmath130 to @xmath131 , making it possible to obtain more accurate results from much fewer samples .
the data in table [ tabledot ] also confirms that the bias can be systematically eliminated by increasing @xmath66 , so that the energy found at @xmath132 essentially represents the exact result ( within error bars ) .
the value @xmath133 is quite typical for many applications . for a simple model system ,
a value @xmath134 was found in ref .
table [ tabledot ] also shows that the cpu time per sample scales linearly with @xmath66 , whereas memory requirements grow @xmath83 .
.[tabledot ] mlb results for @xmath127 and @xmath135 ( see sec .
[ appl : dot ] ) .
@xmath25 is the number of samples ( in @xmath136 ) , @xmath137 the total cpu time ( in hours ) , mb the required memory ( in mega - bytes ) , and @xmath138 the average sign .
bracketed numbers are error estimates . [ cols="<,<,<,<,<,<",options="header " , ] below only results for the _ occupation probability _
@xmath139 under the nonequilibrium initial preparation @xmath140 are presented .
@xmath141 gives the time - dependent difference of the quantum - mechanical occupation probabilities of the left and right states , with the particle initially confined to the left state . to obtain @xmath141 numerically , in a discretized path - integral representation one traces out the bath to get a long - ranged effective action , the _ influence functional _ @xcite . in discretized form the tls path
is represented by spins @xmath142 on the forward- and backward - paths , respectively .
the total action @xmath143 consists of three terms .
first , there is the `` free '' action @xmath144 determined by the bare tls propagator @xmath145 , @xmath146 where @xmath6 is the maximum time and @xmath12 the trotter number .
next there is the influence functional , @xmath147 , which contains the long - ranged interaction among the spins , @xmath148 where @xmath149 is given by @xmath150/4\ ] ] for @xmath151 , and @xmath152 . in the scaling regime at @xmath153 ,
this effective action produces interactions @xmath154 between the spins ( `` inverse - square ising model '' ) .
the contribution @xmath155 describes the interactions with the imaginary - time branch where @xmath156 , with the damping kernel @xmath157 the most difficult case for pimc corresponds to an unbiased two - state system at zero temperature , @xmath158 . to check the code ,
the case @xmath159 was studied in some detail , where the exact solution @xcite is very simple , @xmath160 .
this exact solution only holds in the scaling limit , which is already reached for @xmath161 where the mlb - pimc simulations yield precisely this result .
typical mlb parameters and the respective average sign are listed in table [ tablesb ] .
the first line in table [ tablesb ] corresponds to the naive approach .
it is then clear that the average sign and hence the signal - to - noise ratio can be dramatically improved thus allowing for a study of long timescales @xmath6 . for a fixed number of levels @xmath48
, the average sign grows by increasing the parameter @xmath66 .
alternatively , for fixed @xmath66 , the average sign increases with @xmath48 .
evidently , the latter procedure is more efficient in curing the sign problem , but at the same time computationally expensive . in practice
, it is then necessary to find a suitable compromise .
figure [ fsb ] shows scaling curves for @xmath141 at @xmath162 for @xmath161 and @xmath163 .
the first value for @xmath164 is within the scaling regime .
this is confirmed by a comparison to the noninteracting blip approximation ( niba ) @xcite , which is known to be very accurate for @xmath165 in the scaling limit .
however , for @xmath163 , scaling concepts and also niba are expected to fail dramatically .
this is seen in the simulations .
mlb results show that away from the scaling limit , quantum coherence is able to persist for much longer , and both frequency and decay rate of the oscillations differ significantly from the predictions of niba . in electron transfer reactions in the adiabatic - to - nonadiabatic crossover regime ,
such coherence effects can then strongly influence the low - temperature dynamics .
one obvious and important consequence of these coherence effects is the breakdown of a rate description , implying that theories based on an imaginary - time formalism might not be appropriate in this regime .
a detailed analysis of this crossover regime using mlb is currently in progress @xcite .
these notes summarize our previous activities using the multilevel blocking approach to the sign problem in path - integral monte carlo simulations .
the approach holds substantial promise towards relieving ( and eventually overcoming ) the sign problem , but clearly there is still much room for improvement .
the applications presented here demonstrate unambiguously that there are general and powerful handles to relieve the sign problem , even though a problem characterized by an intrinsic sign problem is still much harder than one without .
we hope that especially young researchers will be attracted to work on this subject themselves .
see , e.g. , _ quantum monte carlo methods in condensed matter physics _
, ed . by m. suzuki ( world scientific , singapore , 1993 ) , and references therein .
see , for instance , e.y .
, j. gubernatis , r.t .
scalettar , s.r .
white , d.j .
scalapino , and r.l .
sugar , phys .
b * 41 * , 9301 ( 1990 ) . c.h .
mak , r. egger , and h. weber - gottschick , phys .
81 * , 4533 ( 1998 ) .
s. chandrasekharan and u .- j .
wiese , phys .
lett . * 83 * , 3116 ( 1999 ) .
kalos and f. pederiva , phys .
lett . * 85 * , 3547 ( 2000 ) .
r. egger , w. husler , c.h .
mak , and h. grabert , phys .
* 82 * , 3320 ( 1999 ) ; * 83 * , 462(e ) ( 1999 ) .
mak and r. egger , j. chem .
phys . * 110 * , 12 ( 1999 ) .
r. egger , l. mhlbacher , and c.h .
mak , phys .
e * 61 * , 5961 ( 2000 ) .
l. mhlbacher and r. egger , in preparation .
dikovsky and c.h .
mak , phys .
b * 63 * , 235105 ( 2001 ) .
j. goodman and a.d .
sokal , phys .
d * 40 * , 2035 ( 1989 ) .
d. thirumalai and b.j .
berne , j. chem .
* 79 * , 5029 ( 1984 ) ; _ ibid . _ * 81 * , 2512 ( 1984 ) .
a.d . kennedy and j. kuti , phys .
* 54 * , 2473 ( 1985 ) .
u. weiss , _ quantum dissipative systems _
, vol.2 ( world scientific , singapore , 1999 ) . r.c .
ashoori , nature * 379 * , 413 ( 1996 ) .
n.b . zhitenev _
_ , science * 285 * , 715 ( 1999 ) .
s. tarucha _
et al . _ ,
lett . * 77 * , 3613 ( 1996 ) .
f. bolton , phys .
lett . * 73 * , 158 ( 1994 ) .
f. pederiva , c.j .
umrigar , and e. lipparini , phys .
b * 62 * , 8120 ( 2000 ) .
j. harting , o. mlken , and p. borrmann , phys .
b * 62 * , 10207 ( 2000 ) .
filinov , m. bonitz , and yu.e .
lozovik , phys .
lett . * 86 * , 3851 ( 2001 ) .
b. reusch , private communication .
b. tanatar and d.m .
ceperley , phys .
b * 39 * , 5005 ( 1989 ) . | this article gives an introduction to the multilevel blocking ( mlb ) approach to both the fermion and the dynamical sign problem in path - integral monte carlo simulations .
mlb is able to substantially relieve the sign problem in many situations . besides an exposition of the method ,
its accuracy and several potential pitfalls are discussed , providing guidelines for the proper choice of certain mlb parameters .
simulation results are shown for strongly interacting electrons in a 2d parabolic quantum dot , the real - time dynamics of several simple model systems , and the dissipative two - state dynamics ( spin - boson problem ) . |
the intermetallic compound zr@xmath0al has been of special interest as a nuclear structural material although a drawback of the material is embrittlement under certain irradiation conditions @xcite .
recently , it was shown that nanostructuring of materials can enhance the irradiation resistance , e.g. for the intermetallic compound niti @xcite .
one successful method to achieve nanostructuring in bulk materials is severe plastic deformation ( spd ) .
the formation of the nanocrystalline structure can occur directly by grain refinement of the coarse grained material or by crystallization of spd induced amorphous material .
for example , crystallization of the intermetallic alloy niti amorphized by severe plastic deformation can lead to nanocrystalline structures and by modification of the deformation path and the heat treatment , properties of the alloy can be tailored @xcite .
bulk intermetallic alloys can be deformed severely using high pressure torsion ( hpt ) . for zr@xmath0al , it was shown that hpt at room temperature leads to a final grain size of approximately 20 nm , but amorphization of a significant volume fraction of the sample has not been encountered @xcite .
furthermore , it was observed that upon hpt deformation , zr@xmath0al exhibits inhomogeneous microstructures , as were also observed in the case of l1@xmath1 structured ni@xmath0al @xcite .
deformation by cold rolling with intermediate foldings ( repeated cold rolling - rcr ) is an alternative promising deformation route to produce bulk nanocrystalline materials @xcite .
the minimum final grain sizes that can be achieved for a material are often smaller than the corresponding ones after spd under high pressure ( e.g. hpt , equal - channel angular pressing ) @xcite .
in addition , rcr can lead to amorphization .
it is the aim of this work to study the effect of rcr on the grain refinement of zr@xmath0al and the behaviour of the refined material upon heating to different temperatures .
crystallization of the amorphous phase is also of interest since the zr - al system exhibits several intermetallic phases ( cf .
[ zral ] ) and the driving force to form the l1@xmath1 structure was reported to be rather weak compared to the one of neighbouring phases @xcite .
-phase at 25at% al is stable up to 1261 k , at higher temperatures it decomposes to bcc @xmath2-zr and zr@xmath1al , which is of the b8@xmath1-structure @xcite.,scaledwidth=45.0% ]
zr@xmath0al was alloyed with an initial composition of zr with 27 at.% al .
the alloy was homogenized at 1160 k for 24 h leading to the ordered l1@xmath1 structure with 10% residual zr@xmath1al and @xmath3-zr .
the material was cut to sheets 10 mm x 10 mm x 0.8 mm in size .
the thickness of the sheets was reduced to 0.25 mm by repeated cold rolling .
then they were folded and again cold rolled until a thickness of 0.25 mm was achieved .
this process was repeated up to 80 times . during rolling the alloy
was placed between two spring steel plates .
the deformed material with the highest strain ( 80 foldings , @xmath4 ) was heated to different temperatures at different heating rates in a differential scanning calorimeter ( dsc ) .
for the baseline subtraction , the material was kept at the maximum temperature after heating until the exothermic signal was negligible to avoid exothermic processes in the subsequent run that provided the baseline .
undeformed samples , samples deformed for 80 foldings and samples heated to 773 k and 973 k at 20 k min@xmath5 after deformation for 80 foldings were investigated by x - ray diffraction ( xrd ) and transmission electron microscopy ( tem ) .
the xrd data were analysed to get information about the coherently scattering domain ( csd ) size and the crystal structure of the samples .
the tem preparation was done by cutting disks suitable for tem preparation from samples of different states , subsequent grinding and dimpling and finally electropolishing using the same parameters as described in @xcite .
acceleration voltages of 200 kv and 300 kv were used for conventional tem and high - resolution tem , respectively .
tem analysis of samples deformed for 5 , 10 and 20 foldings reveals that already at these relatively low strains ( @xmath6 , @xmath7 , @xmath8 ) , grain refinement is clearly visible . at 80 foldings ( @xmath4 )
, xrd shows two broad peaks ( cf . fig .
[ ascr]a ) indicating an amorphous sample .
complementary tem images taken of the same material yield additional important information ( cf .
[ ascr]b ) : a mostly homogeneous intensity distribution with a few remaining crystalline regions that are 100 to 200 nm in diameter ( fig . [ ascr]c ) . from dark - field images
, it can be seen that the whole crystalline region is oriented in a similar way , so it is rather one crystallite than an agglomeration of many small crystallites .
the complex contrast is an indication for a very high defect density in the crystallite .
in addition , small crystallites were identified using high - resolution tem in the amorphous matrix ( fig .
[ ascr]d ) ; they show up as dark dots in fig .
[ ascr]b ( the large ( @xmath9 500 nm ) roundish bright areas originate from thickness variations in the tem foil developed as an artifact during electropolishing . ) . combining the xrd and tem results , a small crystalline volume fraction of a few percent in a mostly amorphous sample
was identified as the sample structure after 80 foldings .
10 nm ) can be seen in the amorphous matrix .
( c ) tem dark - field image of one larger crystallite .
the contrast shows that the crystallite contains a lot of structural defects .
( d ) high - resolution tem image of a small crystallite embedded in an amorphous matrix . ] 10 nm ) can be seen in the amorphous matrix .
( c ) tem dark - field image of one larger crystallite .
the contrast shows that the crystallite contains a lot of structural defects .
( d ) high - resolution tem image of a small crystallite embedded in an amorphous matrix . ]
10 nm ) can be seen in the amorphous matrix .
( c ) tem dark - field image of one larger crystallite .
the contrast shows that the crystallite contains a lot of structural defects .
( d ) high - resolution tem image of a small crystallite embedded in an amorphous matrix . ]
dsc experiments were conducted to clarify the thermal stability of the as - deformed material .
[ dsc ] shows a baseline - subtracted dsc curve for a heating rate of 5 k min@xmath5 .
the baseline was obtained by a second run after annealing at 973 k for 30 min .
two exothermic peaks are revealed .
the first peak ( peak a ) that starts at 650 k and has its maximum at 720 k is caused by crystallization of the amorphous phase . to determine the enthalpy of peak a , the following method ( cf .
inset in fig .
[ dsc ] ) was chosen under the assumption that the peak is symmetric and that peak b has a negligible influence on the low - temperature half of peak a : from the left integration limit @xmath10 ( @xmath11 was determined using the tangent method ) , the enthalpy between a horizontal baseline and the dsc curve was integrated until @xmath12 ( grey area in inset in fig .
[ dsc ] ) .
this value was taken as the enthalpy of the low - temperature half of the enthalpy of the peak . due to the above - mentioned assumptions , doubling this value yields the total enthalpy of the peak .
the determination of the enthalpies for four different heating rates yielded a crystallization enthalpy of @xmath133 j / g .
this statistical uncertainty , however , seems to underestimate the real one , that might be as high as 20% , due to systematic instrumental errors and the assumption of peak b not having an influence on the low - temperature half of peak a. a kissinger plot using the four different heating rates ranging from 5 k min@xmath5 to 50 k min@xmath5 leads to an activation enthalpy of @xmath14 ev .
the second exothermic peak with a plateau - like behaviour at high temperatures is attributed to a superposition of grain growth and phase transformation processes . ) of zr@xmath0al deformed by 80 foldings . a crystallization peak
( marked a ) can be seen at 720 k. a large peak ( marked b ) occurs at higher temperature . the temperatures to which the tem samples were heated are indicated .
a sketch of the integration method is shown in the inset.,scaledwidth=50.0% ] using a heating rate of 20 k min@xmath5 , samples were heated to 773 k and to 973k , i.e. after the crystallization peak and to the maximum temperature of the dsc curve , respectively .
[ dsc4]a shows a tem bright - field image of the sample heated to 773 k revealing nanocrystalline material with a grain size of approximately 10 - 20 nm .
this grain size estimation is in reasonable agreement with a williamson - hall analysis of the csd size as deduced from the corresponding xrd spectrum ( cf .
[ dsc4]b ) yielding a volume - weighted mean crystalline diameter of 40 nm .
apart from the csd size , the xrd results show that the b8@xmath1-structure is clearly predominant in the sample heated to 773 k. all the major peaks can be related to peaks of this structure , which is the equilibrium phase for zr@xmath1al @xcite .
this result was also confirmed by analyzing the corresponding electron diffraction ring patterns using the pasad software @xcite .
after heating the sample to 973 k , defect - free grains grow to a size of @xmath15 100 nm as observed in the tem bright - field image ( cf .
[ dsc4]c ) . the corresponding xrd curve ( cf .
[ dsc4]d ) shows that the l1@xmath1 structure ( i.e. the equilibrium structure for zr@xmath0al ) starts to form .
it should be mentioned that after heating to 973 k at 20 k min@xmath5 and immediate subsequent cooling at the same rate , the b8@xmath1 structure is still predominant .
al deformed by 80 foldings and subsequent heat treatment . for the xrd peaks ,
the underlying structures and equilibrium compounds are indicated by symbols ; the pronounced background at low angles comes from the glass sample holder .
( a ) tem image after heating to 773 k. the grain size is 10 - 20 nm .
( b ) xrd after heating to 773 k. all the peaks correspond to the b8@xmath1 structure .
( c ) tem image after heating to 973 k ; defect - free grains with a grain size of @xmath15 100 nm .
( d ) xrd after heating to 973 k. the major peaks mainly correspond to the b8@xmath1 structure ; minor peaks to the @xmath3-zr , l1@xmath1 and the equilibrium zr@xmath0al@xmath1 structure . ]
al deformed by 80 foldings and subsequent heat treatment . for the xrd peaks ,
the underlying structures and equilibrium compounds are indicated by symbols ; the pronounced background at low angles comes from the glass sample holder .
( a ) tem image after heating to 773 k. the grain size is 10 - 20 nm .
( b ) xrd after heating to 773 k. all the peaks correspond to the b8@xmath1 structure .
( c ) tem image after heating to 973 k ; defect - free grains with a grain size of @xmath15 100 nm .
( d ) xrd after heating to 973 k. the major peaks mainly correspond to the b8@xmath1 structure ; minor peaks to the @xmath3-zr , l1@xmath1 and the equilibrium zr@xmath0al@xmath1 structure . ]
al deformed by 80 foldings and subsequent heat treatment . for the xrd peaks ,
the underlying structures and equilibrium compounds are indicated by symbols ; the pronounced background at low angles comes from the glass sample holder .
( a ) tem image after heating to 773 k. the grain size is 10 - 20 nm .
( b ) xrd after heating to 773 k. all the peaks correspond to the b8@xmath1 structure .
( c ) tem image after heating to 973 k ; defect - free grains with a grain size of @xmath15 100 nm .
( d ) xrd after heating to 973 k. the major peaks mainly correspond to the b8@xmath1 structure ; minor peaks to the @xmath3-zr , l1@xmath1 and the equilibrium zr@xmath0al@xmath1 structure . ]
tem analyses of samples deformed by rcr show that pronounced grain refinement takes place at much lower strain levels than it was reported for zr@xmath0al deformed by high pressure torsion @xcite . after 80 foldings ( @xmath16 ) , at least 90% of the volume is amorphous and the grain structure is finer than that in the hpt experiment .
this is in agreement with the results showing that the grain sizes after rcr can be smaller than those achieved by other spd techniques @xcite . as depicted in fig .
[ ascr]b , a homogeneously distributed crystalline debris is retained in the amorphous material . upon heating to 773 k , the crystallites that are only a few nm large can serve as pre - existing nuclei and lead to the formation of a fine nanocrystalline structure ( 10 - 20 nm , cf . fig . [
dsc4]a ) because of their dense distribution .
a similar behaviour was observed for niti , that was rendered amorphous by severe plastic deformation and afterwards heated in in - situ tem experiments @xcite .
peak a of the dsc curve ( cf .
[ dsc ] ) is therefore attributed to crystallization and grain growth until impingement .
the peak temperature is about 70 k lower than that reported in @xcite for ball - milled amorphous zr@xmath0al .
this difference can be explained by the presence of the crystalline debris in our samples . a similar behaviour was observed in niti with a peak temperature reduction of 100 k @xcite .
therefore , it is concluded that the higher peak temperature reported for zr@xmath0al in @xcite indicates that the material did not contain a crystalline debris .
peak b of the dsc curve ( cf .
[ dsc ] ) is interpreted to be caused by grain growth and phase transformations between different intermetallic phases ( cf .
[ dsc4]c and d ) . by experimental and theoretical evaluation , ma et al @xcite deduced an enthalpy release caused by devitrification of zr@xmath0al of @xmath15 85 j / g , which is about twice the amount that is determined here .
possible reasons for the higher crystallization enthalpy are an additional contribution of grain growth , which is part of dsc peak b ( cf . fig .
[ dsc ] ) in this work or a principal difference in the structure of the amorphous samples caused by the different processing conditions ( strains , strain rates , deformation temperatures , ... ) x - ray results show that the equilibrium l1@xmath1 structure does not form in the initial crystallization process since only b8@xmath1 structured zr@xmath1al was detected after heating to 773 k ( cf . fig .
[ dsc4]b ) .
it is well established that nanocrystalline material can have a different crystal structure than its coarse crystalline counterpart at the same composition and temperature @xcite . in the present case , antisites or vacancies on al sites of the ordered b8@xmath1 structure could in addition
compensate for the difference between the nominal as - cast composition and that of stoichiometric zr@xmath1al . from the dsc curve
, the formation enthalpy of the present b8@xmath1 structure ( containing a high number of point defects causing excess enthalpy ) is determined to lie between the one of the amorphous phase and the one of the l1@xmath1 structure as given in @xcite .
this makes the transformation path @xmath17 thermodynamically reasonable .
a comparison with data of zr@xmath18al@xmath0 and zr@xmath18al@xmath19 favors the substitution with vacancies @xcite . assuming a vacancy density that compensates for the aluminium deficiency in the formation of the b8@xmath1 structure and ignoring grain size effects causing excess enthalpy gives a reasonable upper limit of 2.7 ev for the formation enthalpy of a vacancy . after heating to 973 k ,
the zr@xmath1al phase is still predominant , but some @xmath3-zr , l1@xmath1 structured zr@xmath0al and a little zr@xmath0al@xmath1 of the space group p4@xmath1/mnm starts to form ( cf .
[ dsc4]d ) .
the heat treatments presented in this work resulted in a very fine nanocrystalline structure dominated by the b8@xmath1 phase ( 10 - 20 nm ) .
it is suggested that heat treatment of amorphous samples at low temperatures ( @xmath15 770 k ) for several hours or even days might give rise to the formation of the l1@xmath1 structure without causing excessive grain growth .
it is concluded that the crystalline debris resulting from the amorphization by rcr is important for reducing the grain size of the crystallized material by providing a dense network of homogeneously distributed pre - existing nuclei .
* zr@xmath0al can be rendered amorphous by repeated cold rolling as concluded from the x - ray results .
the tem studies show that the deformed material contains a crystalline debris that is not resolved by x - ray diffraction . *
heating to 773 k leads to the crystallization of small ( 10 - 20 nm in diameter ) b8@xmath1-structured nanocrystals . in this nanocrystalline structure ,
the non - equilibrium zr@xmath1al phase is clearly predominant . *
it is concluded that during heating , the crystalline debris ( of the zr@xmath0al phase retained in the amorphous phase ) acts as nuclei for the crystallization process of the non - equilibrium phase zr@xmath1al .
* when heating to 973 k , the equilibrium l1@xmath1 phase starts to form and the crystals grow to a size of about 100 nm .
the authors thank prof . erland m. schulson ( dartmouth college , new hampshire , usa ) for the kind provision of zr@xmath20al . d.g .
, h.p.k . and c.r .
acknowledge support by the research project ` bulk nanostructured materials ' within the research focus ` materials science ' of the university of vienna and by the austrian science fund ( fwf):[p22440 ] . d.g . acknowledges the support by the ` nims internship program ' of the national institute for materials science , tsukuba , japan and by the ig ` experimental materials science - nanostructured materials ' , a college for phd students at the university of vienna . | the intermetallic compound zr@xmath0al is severely deformed by the method of repeated cold rolling . by x - ray diffraction it is shown that this leads to amorphization .
tem investigations reveal that a homogeneously distributed debris of very small nanocrystals is present in the amorphous matrix that is not resolved by x - ray diffraction . after heating to 773 k
, the crystallization of the amorphous structure leads to a fully nanocrystalline structure of small grains ( 10 - 20 nm in diameter ) of the non - equilibrium zr@xmath1al phase .
it is concluded that the debris retained in the amorphous phase acts as nuclei . after heating to 973
k the grains grow to about 100 nm in diameter and the compound zr@xmath0al starts to form , that is corresponding to the alloy composition . |
evolutionary dynamics arise from the interplay of mutations acting on genotypes and natural selection acting on phenotypes . understanding the structure of the genotype - phenotype ( gp ) map is therefore critical for understanding evolutionary processes .
we address a simple question about structure : are the genotypes positively correlated ? that is , will the mutational neighbours of a genotype be more likely to map to similar phenotypes than expected from random chance ?
john maynard smith and others have argued that the intuitive answer is yes .
here we quantify these intuitions by comparing model gp maps for rna secondary structure , protein tertiary structure , and protein quaternary structure to a random gp map .
we find strong neutral correlations : point mutations are orders of magnitude more likely than expected by random chance to link genotypes that map to the same phenotype , which vitally increases the potential for evolutionary innovation by generating neutral networks .
if gp maps were uncorrelated like the random map , evolution may not even be possible .
we also find correlations for non - neutral mutations : mutational neighbourhoods are less diverse than expected by random chance .
such local heterogeneity slows down the rate at which new phenotypic variation can be found . but non - neutral correlations
also enhance evolvability by lowering the probability of mutating to a deleterious non - folding or non - assembling phenotype .
in a classic paper @xcite , published in 1970 , john maynard smith introduced several key ideas for describing the structure of genotype - phenotype ( gp ) maps .
he first outlined the concept of a protein space , the set of all possible sets of amino acid chains , and suggested that for evolution to smoothly proceed , these should be connected as networks of functional protein phenotypes that can be interconverted by ( point ) mutations .
he then argued that one criterion for such networks to exist is for a protein @xmath0 to have at least one mutationally accessible neighbour which is `` meaningful , in the sense of being as good or better than @xmath0 in some environment '' .
in other words , if @xmath0 has @xmath1 mutational neighbours , then the frequency @xmath2 of `` meaningful '' proteins in its mutational neighborhood should satisfy @xmath3 .
he pointed out that this was likely to be true in part due to the ubiquity of neutral mutations , which had been famously proposed by kimura @xcite and king and jukes @xcite just a few years prior to his paper .
but he also gave a second reason for expecting connected networks , namely that , `` there is almost certainly a higher probability that a sequence will be meaningful if it is a neighbour of an existing functional protein than if it is selected at random . ''
this concept that mutational neighbours differ from the random expectation is what we will call _ genetic correlations_. following maynard smith , many authors have explored the role of networks of genotypes connected by single point mutations .
lipman and wilbur @xcite first showed that large networks of mutationally connected genotypes mapping to the same phenotype are found in the hydrophobic - polar ( hp ) model for protein folding , introduced by dill @xcite .
they also pointed out that neutral mutations allow a population to traverse these networks , facilitating access to a larger variety of alternate phenotypes .
schuster and colleagues @xcite developed these themes further using detailed models for the secondary structure of rna @xcite .
they coined the term `` neutral network '' to describe sets of mutationally connected genotypes that map to the same secondary structure phenotype .
as rna secondary structure is fairly easy to calculate and thermodynamics based models such as the vienna package are thought to provide an accurate prediction of real rna secondary structure @xcite , the nature of neutral networks in these models has been extensively studied @xcite . since these pioneering works ,
neutral networks have been considered in gp maps of other biological processes , including models for gene networks @xcite metabolic networks @xcite and the polyomino model for self - assembling protein quaternary structure @xcite . from these studies of model systems
a number of basic principles have emerged , much of which has been reviewed in important books by wagner @xcite .
firstly , for neutral networks to exist , the gp map should exhibit _ redundancy _ , where multiple genotypes map onto the same phenotype .
this many - to - one nature of the mappings is illustrated in fig .
redundancy is of course closely linked to the existence of neutral mutations @xcite , although the relationship between these concepts is not entirely unambiguous . in the theory of neutral evolution , a mutation may lead to a slightly different phenotype , but as long as the change in fitness is small enough not to be visible to selection , it is considered to be effectively neutral @xcite . whether selection can act depends on the degree of phenotypic change , the environment , and other factors such as the population size and mutation rate .
therefore , identifying whether or not a mutation is neutral can be complex , and the answer may vary as parameters external to the gp map change with time .
so while redundancy only couples identical phenotypes , and so is a more restrictive concept than neutral mutations , it has the advantage of sidestepping the subtle issues listed above and is therefore more easily applicable to the study of a static gp map .
the second basic principle to emerge is that the number of genotypes per phenotype ( the redundancy ) can vary , leading to _ phenotype bias _ , as depicted in fig .
[ fig1 ] b. thirdly , it is generally the case that the larger the redundancy , the greater the mean mutational robustness of genotypes mapping to that phenotype .
fourthly , the larger the neutral network , the greater the variety of alternative phenotypes within one ( non - neutral ) point mutation of the whole neutral network , leading to a positive correlation between robustness and measures of evolvability that count the number of different phenotypes that are potentially accessible @xcite .
finally , a key principle emphasised by maynard smith @xcite , but which has earlier roots in concepts such as the shifting balance theory of sewall wright @xcite , is that neutral mutations allow a population to access , over time , a wider variety of potential alternative phenotypes than would be available around a single genotype @xcite .
evidence for the key role of these networks in promoting evolutionary innovation has been found , for example , in experiments on rna structures @xcite and transcription factors @xcite .
the main focus of this paper is genetic correlations . to explore and quantify how they affect concepts such as neutral networks , robustness and evolvability
, we study genetic correlations in three of the gp maps mentioned above .
these are the sequence to rna secondary structure map and hp model for protein folding ( tertiary structure ) , which have been extensively studied , as well as the more recently introduced polyomino model for self - assembling protein quaternary structure .
several properties of these three gp maps have recently been compared @xcite and we summarise some of the similarities and differences between them in the methods section . however , a detailed investigation of their genetic correlations has not yet been considered .
a key question we consider is how to define genetic correlations in a quantitative way .
for that one needs some kind of uncorrelated null model for how genotypes are distributed over phenotypes with which to compare the full biophysical systems .
we employ a model we call the _ random gp map_. it has the same number of genotypes mapping to each phenotype as the biophysical gp map to which it is being compared , as well as the same basic type of genotype space ( alphabet size and genome length ) , and nodes ( genomes ) that are linked by single mutations if they differ by one locus .
the difference is that the genotypes are randomly distributed across the genotype space .
of course one does not expect biological systems to have such a random distribution , but it is not at all straightforward to think of a better null expectation .
the great advantage of having such a null model , even if one knows that it is has limitations , is that it allows us to quantitatively contrast how the biophysical genotypes are organised across the genetic space , which should shed light on nature of the correlations that maynard smith introduced .
the paper is organised as follows .
we first define our models in the methods section .
next we examine _ neutral correlations _
, schematically illustrated in fig .
[ fig1 ] c , through considering various measures of robustness that quantify the relative likelihood that mutationally neighbouring genotypes possess the same phenotype .
we then perform a similar analysis comparing the biological gp maps to the random map for _ non - neutral correlations_. since these different kinds of correlations all modulate the way that novel variation arises through random mutations , we finish by commenting on how correlations affect subtle interplay of robustness and evolvability @xcite , and also briefly suggest a few other forms of correlation that could be studied in gp maps .
we consider three separate gp maps for low - level self - assembling biological systems .
firstly , a model for rna secondary structure @xcite that determines which bases in an rna sequence form bonded pairs . secondly , the hp lattice model for protein tertiary structure @xcite , that determines the three - dimensional shape of a folded amino acid chain . and thirdly , the polyomino model for protein quaternary structures@xcite , where quaternary structure of proteins is the topological arrangement of separate folded amino acid chains .
all three of these models have been previously compared in ref .
@xcite , and below we briefly outline the three different systems . in the widely studied rna secondary structure gp map @xcite ,
the genotypes are sequences made of an alphabet of four different nucleotides , and phenotypes are the secondary structures , which describe the bonding pattern in the folded structure with the lowest free energy for the given sequence . here
, we use the popular vienna package @xcite , which uses an empirical free - energy model and dynamic programming techniques to efficiently find the lowest free energy structures .
we use version 1.8.5 with all parameters set to their default values . the rna gp map for length @xmath4
is referred to as rna@xmath4 .
we present results from rna12 , with @xmath5 genotypes mapping to @xmath6 folded pheontypes , rna15 with @xmath7 genotypes mapping to @xmath8 folded phenotypes and and rna20 , with @xmath9 genotypes mapping to @xmath10 folded secondary structure phenotypes .
the hp lattice model represents proteins as a linear chain of either hydrophobic ( h ) or polar ( p ) amino acids on a lattice @xcite .
a simple interaction energy function is used between non - adjacent molecules in the chain .
hydrophobic - hydrophobic interactions are energetically favourable .
we use an implementation with the interaction energy @xmath11 between the different potential pairs being classified as @xmath12 , with @xmath13 , as widely chosen by other authors , see e.g. @xcite .
a phenotype is defined for each shape ( fold ) on the lattice that is the unique free energy minimum of at least one sequence .
if a sequence has more than one structure as its minimum , then it is considered not to fold properly , and so is categorised as belonging to the ( potentially deleterious ) general non - folding phenotype .
the compact hp model restricts the possible folds to those which are maximally compact , in an attempt to capture the globular nature of _ in vivo _
proteins @xcite .
we make use of both compact and non - compact hp gp maps by considering both the gp map for all folds of length , @xmath14 ( denoted hp24 ) and all compact folds on the @xmath15 square grid ( of length @xmath16 , denoted hp5x5 ) . for the non - compact hp24 gp map
there are @xmath17 genotypes and @xmath18 folded phenotypes , while for the compact hp5x5 model the @xmath19 genotypes map to a much smaller set of @xmath20 unique folded phenotypes .
protein quaternary structure describes the topological arrangement of separate folded proteins that self - assemble into a well - defined cluster .
the polyomino gp map is a recently introduced lattice model which , in the spirit of the hp lattice model for tertiary structure , provides a simplified but tractable gp map for protein quaternary structure , as described in more detail in @xcite . very briefly , it employs a set of @xmath21 square tiles with @xmath22 interface types , together with a set of rules that denote how the interfaces bind .
these sets are specified by genotypes in the form of linear strings . in this work ,
@xmath23 and we specify that the interface types interact in ordered odd - even pairs , such that the following interface types interact ( @xmath24 , @xmath25 , @xmath26 ) .
the conversion of the tile set ( genotype ) into a bounded shape ( phenotype ) is achieved through simulating a lattice based self - assembly process .
this begins with seeding where the first tile encoded in the genotype is placed on the lattice .
thereafter other tiles are randomly placed .
if they form a bond they are kept , otherwise they are discarded .
this assembly process is repeated until either no more bonds can be formed or else the number of tiles grows without limit . for each set of tiles , assembly
is attempted several times .
if a given set of tiles self - assembles into a unique bounded shape , the genotype is considered to map to that shape ( phenotype ) .
if on the other hand , the tile set does not always assemble to the same shape , or if it assembles to an unbounded shape ( as occurs in sickle cell anaemia for example ) , then the genotype maps to the _ undefined phenotype _ ( und ) .
the gp map resulting from @xmath21 kit tiles and @xmath22 interface types is denoted as @xmath27 . in this work , we consider @xmath28 which has @xmath29 genotypes mapping to @xmath30 different self - assembling phenotypes and the larger space @xmath31 which has @xmath32 genotypes mapping to @xmath33 phenotypes .
all parameters used for simulations were the same as in ref .
@xcite .
we also distinguish a _ deleterious phenotype _ ( del ) in all three gp maps . for the rna gp map ,
this occurs when the unbonded strand is the free - energy minimum , so that the strand does not fold . for the hp model
this occurs when a sequence does not have a unique ground state structure , which is interpreted as the protein not folding . for the polyomino gp map
this occurs when the set of tiles produces an und phenotype .
note that , depending on the environment , many other phenotypes may also be deleterious , but the del phenotype will always be an evolutionary dead end . for rna
this del phenotype makes up @xmath34 of rna12 , @xmath35 of rna15 and @xmath36 of rna20 . in the hp model
the fraction is typically larger , consisting of @xmath37 of the hp24 map and @xmath38 for the hp5x5 mapping , while for the polyomino gp map we find @xmath39 of @xmath28 and @xmath40 of @xmath31 . for the rna
this fraction decreases with increasing genotype length @xmath4 ( asymptotically the fraction @xmath41 , of genotypes mapping to the deleterious phenotype scales as @xmath42 @xcite ) while for the polyomino gp map @xmath41 increases for larger system size , and in the hp model the trend remains ambiguous @xcite .
recently , direct comparisons between the properties of these gp maps have begun to be considered @xcite .
a common feature of all three gp maps is that they correspond to self - assembly processes of molecular structures in biological systems .
thermodynamics drives the assembly process in each case .
they differ in that the rna and hp model have linked units that are fixed in size , while the polyomino model has disconnected units that can assemble into multiple sizes .
the gp maps share properties such as redundancy , bias and phenotypic robustness @xcite ( see fig . [ fig1 ] ) , which are seen in a much wider range of gp maps @xcite .
however , the nature of these properties varies between the gp maps .
for example , the polyomino and rna gp maps typically have fewer phenotypes and larger neutral networks than phenotypes in the hp model .
another key variable that differs between the gp maps is the base @xmath43 . for rna @xmath44 , hp @xmath45 and for polyominoes @xmath46 .
these different genotypic topologies as well as the way in basic units interact to generate phenotypes can affect the formation of neutral networks .
for example , in rna , a cg bond in a stem motif can not be turned into a gc bond without breaking the bond @xcite , a form of neutral reciprocal sign epistasis @xcite . for a secondary structure stem motif of @xmath47 bonds ,
this phenomenon breaks the neutral set up into around @xmath48 separate components , often of similar size , that are connected by point mutations internally , but which need at least two mutations to be connected together .
these different components of the full neutral set are separate neutral networks . with @xmath46 and asymmetric bonding of bases ( @xmath24 and @xmath49 being distinct ) , the polyomino gp map may be similarly susceptible to this form of epistasis .
the hp lattice model does not have this feature due to bonds being formed from non - adjacent neighbouring hh interactions .
how these connectivity patterns scale with increasing @xmath4 remains an open question , in part because the spaces grow exponentially with length , and so rapidly become much more difficult to comprehensively analyse . as discussed in the introduction , in order to quantify genetic correlations we must first define an uncorrelated null model to which the biophysical gp maps can be compared . here
we employ a random gp map that was recently explicitly introduced for analysing whole gp map properties in ref .
@xcite , but has also been used implicitly in many earlier works see e.g. @xcite , although we believe this is the first time this random model has been used to _ define _ correlations .
the random gp map shares the following properties with the biological gp map to which it is being compared : the same alphabet size @xmath43 , genome length @xmath4 , number of 1-mutation neighbours @xmath50 , number of genotypes @xmath51 , number of phenotypes @xmath52 , and frequencies @xmath53 , defined as the fraction of all genotypes that possess phenotype @xmath54 .
we summarise the gp map nomenclature used in this paper in table [ table1 ] , which compares what is shared and what is different between the biological and the random gp maps . with these key global gp map properties fixed
, the only difference between a biological gp map and its associated random gp map is that the @xmath55 genotypes for each phenotype @xmath54 are each randomly assigned to the set of @xmath56 possible genotypes .
as phenotypes are randomly assigned , departures in properties between the two versions of the gp map may be considered to be due to _ correlations _ , that is , in contrast to the random gp map , the mutational neighbourhood of a genotype in the biological gp maps is affected , for example , by what phenotype it maps to .
we note that these correlations can be very complex , and depend not only on the identity of the phenotype , but also on higher order features such as the identity of one or two or more phenotypes in the direct neighbourhood ( higher order correlations ) .
it is almost certainly also true that , depending on the phenotype , the correlations may also depend on which of the @xmath4 genomic positions is mutated ( see e.g. maynard smith s word game described in the discussion ) .
however , in this paper we mainly focus on the simplest kinds of correlations ; for example for neutral correlations we mainly look at effects that are captured by the concept of robustness . other null models are conceivable .
for example , in ref .
@xcite the authors used an approach based on network theory @xcite , comparing the topology of neutral components to erds - rnyi networks and scale - free networks . while this type of network is helpful for understanding the topology of neutral networks themselves , a focus of this work here is not just how genotypes of the same phenotype connect to each other , but how phenotypes are arranged in relation to each other . besides their simplicity
, an advantage of using the random gp maps for this purpose is that the overall connectivity of the genotype space is left intact , along with several global properties of the map , allowing the way phenotypes are arranged to be directly considered .
moreover , this random map has been used ( implicitly and explicitly ) throughout the literature , see e.g. @xcite , and so it is of general interest to carefully analyse some of its properties .
x r properties shared by random and biological gp maps & symbol + _ alphabet size _ : & @xmath43 + _ genotype length _ : & @xmath4 + number of 1-mutation neighbours of a genotype : & @xmath50 + number of genotypes : & @xmath51 + number of phenotypes : & @xmath52 + _ redundancy _ : the size of neutral set or the number of genotypes that map to phenotype @xmath54 & @xmath57 + _ phenotype frequency _ : the fraction of genotypes that map to phenotype @xmath54 , & @xmath58 + properties that differ between random and biological gp maps & symbol + : all genotypes that map to phenotype @xmath54 & @xmath59 + _ neutral component _ : a subset of @xmath59 that is fully connected by point mutations . also called a _
neutral network_. & @xmath60 + the number of 1-mutation neighbours of genotype @xmath61 mapping to phenotype @xmath54 & @xmath62 + _ phenotype robustness _ : mean robustness of all genotypes mapping to a phenotype @xmath54 & @xmath63 + _ phenotype mutation probability _ : probability that a point mutation from a genotype mapping to phenotype @xmath54 will generate a genotype mapping to phenotype @xmath64 & @xmath65 +
the concept of robustness to mutations is well established in the literature @xcite .
it is intimately tied to neutral correlations , in fact robustness helps quantify the amount of neutral correlation present . before we study the more novel topic of non - neutral correlations , it is therefore interesting to compare various measures of robustness between the biological gp maps and the random uncorrelated gp map .
the 1-robustness of a single genotype @xmath61 that maps to phenotype @xmath54 is straightforwardly defined as the number of genotypes @xmath62 that map to @xmath54 that are accessible within one point mutation of @xmath61 .
the _ phenotypic robustness _
@xmath63 of a phenotype @xmath54 is defined the average of the 1-robustness over the entire neutral set @xmath59 @xcite .
this can be expressed algebraically as @xmath66 in a random gp map , phenotypes are arranged randomly over genotypes so the probability that a genotype leads to phenotype @xmath54 is given by its frequency @xmath53 , independently of the identity of its neighbours . the phenotypic robustness therefore is simply @xmath67 and the mean number of neutral neighbours is @xmath68 which is the expectation value for a binomial distribution with @xmath50 trials and probability of a given neighbour being @xmath53 .
it is independent of the identity of the genotype @xmath61 .
we define _ neutral correlations _ as the difference in how genotypes mapping to the same phenotype are distributed in a biologically relevant gp map as compared to the associated random gp map null model .
one way of characterising these neutral correlations is by comparing the phenotype robustness @xmath63 to the random expectation @xmath69 .
the violation of this equality is a sufficient ( though not necessary ) condition for the existence of neutral correlations .
moreover , we define a phenotype @xmath54 to have _ positive _ neutral correlations if @xmath70 is satisfied .
this is intuitive when robustness is greater than @xmath53 then phenotypes are closer to each other in the genotype network than would be expected by random chance . using the above definitions around neutral correlations , we explicitly consider the robustness in the various gp maps . in fig .
[ fig2]a , we compare the phenotypic robustness across our three biological gp maps to the robustness of the associated random gp map .
the figure confirms both the analytical result derived above for the random model that @xmath71 ( we only show one schematic random map in the figure , but the others have the same behaviour ) . in sharp contrast , for the biological gp maps we find that , very roughly , @xmath72 , so that the robustness is much larger than would be expected for the null model , in fact by several orders of magnitude for smaller @xmath53 .
since @xmath73 , this indicates the presence of extremely strong neutral correlations in these biological gp maps .
of course the fact that @xmath70 is not a new finding , but it is instructive to show this trend displayed explicitly for entire mappings in the three kinds of biological systems .
we next extend phenotype robustness to @xmath47-mutations .
_ generalised robustness _ or _
@xmath74 , measures phenotypic robustness for a greater number of mutations .
it is defined as the robustness of a genotype with phenotype @xmath54 to @xmath47 independent mutations to its genotype , rather than just the single mutation discussed above .
this can be expressed algebraically as @xmath75 where @xmath76 is the number of @xmath47-mutant neighbours of @xmath61 with phenotype @xmath54 and the normalisation on the right - hand of the sum is the total number of @xmath47-mutants . in the same way as for the phenotype robustness
, the @xmath47-robustness is averaged across the neutral set @xmath59 of all genotypes that map to phenotype @xmath54 .
a further quantity we define is the _
average n - robustness _ @xmath77 which is the average of the @xmath47-robustness over all phenotypes in a given gp map : @xmath78 where @xmath79 is the set of all @xmath52 phenotypes in the gp map .
in contrast to the two previous definitions that measure robustness for a single phenotype , it is a general property of the whole gp map .
one could imagine generalising this further to a subset of the phenotypes , for example those whose frequencies @xmath53 are greater than the average @xmath80 . to establish the @xmath47-robustness and average @xmath47-robustness in the random gp map , the same logic can be applied as in the previous section . since the probability of finding a phenotype is uniformly distributed over the genotype space , the @xmath47-robustness is given by @xmath81 with the @xmath47-robustness the same for all @xmath47 , leading to an average @xmath47-robustness : @xmath82 since the phenotype frequencies in a gp map sum to unity .
the inequality @xmath83 can be used to define whether a biological gp map possesses neutral correlations as a whole .
we consider the average @xmath47-robustness against the radius @xmath47 for the three gp maps @xmath28 , rna12 and hp24 .
a sample of 100 genotypes for each phenotype in the respective systems is taken ( apart from hp24 where a sample of 100 randomly chosen phenotypes is made due to the large number of phenotypes ) and the @xmath47-robustness is measured and averaged over phenotypes . in fig .
[ fig2]b , we plot the average @xmath47-robustness at each radius along with the flat expectation lines from eq .
[ eq : av_nrobustness ] for the null models . in all three cases
we observe a decay from greater than the null values for small radii to slightly less than the null expectation at larger radii .
the reason for this drop below the random expectation can be understood intuitively : given that positive neutral correlations are present , the over - representation for small radii must be balanced at larger radii by under - representation in order for the number of genotypes to balance .
we also define a _ neutral correlation length _
@xmath84 which measures the mutational hamming distance over which neutral correlations extend .
we define @xmath84 for a phenotype to be equal to the smallest value of @xmath47 where @xmath85 and for the gp map when @xmath86 .
we find that @xmath87 for the rna12 model , @xmath88 for the polyomino @xmath28 model and @xmath89 for the hp24 model .
the neutral correlation length is smaller for the hp model than the other two systems .
as discussed in the methods section , and illustrated in fig.[fig2]a , the hp model typically has phenotypes with smaller frequencies / robustness than the other two systems suggesting neutral networks that do not expand to the same diameter which would reduce the expected neutral correlation length .
all three models are of fairly small genome length @xmath4 , so one should be careful of reading too much into the numerical values of these correlation lengths
. however , it may very well be that this ordering of models will persist for larger @xmath4 .
having illustrated the concept of positive neutral correlations
measured by ( generalised ) robustness greater than that of the random null model
we next show how other properties of neutral networks are affected by their presence .
the neutral set @xmath59 is the set of all genotypes mapping to phenotype @xmath54 .
a component is the subset of the neutral set @xmath59 that is connected by single point mutations .
we use this term because it is commonly used in graph theory to denote a set that is connected .
although the literature can be somewhat ambiguous , with the term neutral networks sometimes referring to the neutral set , and sometimes to a neutral component , we take a neutral network to be synonymous to a neutral component in this paper because if we have only point mutations then a population can only explore a neutral component and may not be able access the whole neutral set .
there are several reasons why a neutral set may not be fully connected by neutral point mutations .
if the genotypes are too diffusely spread out over the full genotype space , then they may be disconnected .
but in some cases basic biophysical constraints , such as the neutral reciprocal sign epistasis described in the methods section , also lead to fragmentation .
we begin by comparing the size of neutral components in the random null model to those found in our biological gp maps . in the random model , there are two important threshold values : firstly , the _ giant component onset _
, when a phenotype s components change from being largely isolated to forming larger connected clusters , and secondly , the _ single component onset _ where virtually all genotypes are taken up by a single giant connected component .
as each genotype has many neighbours , a simple mean - field - like approach from _ percolation theory _ for random graphs @xcite should be fairly accurate .
this suggests that the giant component onset begins when the average number of neighbours of a given genotype with the same phenotype is approximately unity , which was also the criterion used by john maynard smith @xcite . for the null model , where phenotypes are assigned to genotypes completely randomly ,
this reduces to an explicit threshold frequency @xmath90 such that we expect the giant components for phenotypes with @xmath91 .
it can be shown analytically in the limit @xmath92 @xcite that there is another transition at @xmath93 where , for @xmath94 , all the components coalesce into one single giant component , so that the neutral set should be ( nearly ) fully connected . while the giant component threshold @xmath95 scales as @xmath96 , so that it decreases for larger maps , the single component threshold @xmath97 from eq .
[ eq : lambda ] is independent of genome length @xmath4 , and only varies with alphabet size .
for example , @xmath98 for @xmath45 and @xmath99 for @xmath44 .
these are large frequencies that are unlikely to be reached for more than a single phenotype in any realistic gp maps . in fig .
[ fig3 ] , we plot how the largest component size ( left ) and number of components ( right ) varies with frequency in both a null model ( @xmath44 , @xmath100 ) and three gp maps @xmath28 , rna12 and hp24 .
we first focus on the simple schematic null model .
data is calculated by averaging over 100 independent realisations of the random mapping of genotypes to phenotypes in a way that preserves the frequencies .
the largest component size , and the number of components formed by the phenotype , are then measured .
these values are shown in fig .
[ fig3 ] for an array of frequencies in the schematic null gp map . below the giant component onset @xmath101 , most genotypes are completely isolated the total number of neutral components scales with @xmath53 . around the giant component threshold @xmath95 ,
this scaling changes markedly , and instead the size of the largest neutral components scales linearly with @xmath53 and takes up the majority of the genotypes in the neutral set .
the number of components continues to decline until @xmath53 exceeds the single component connectivity threshold @xmath99 , at which point there is just one component and the neutral set is completely connected .
we next consider the biological gp maps relative to the behaviour exhibited by the null model .
firstly , all three gp maps have much larger maximum neutral set sizes than the random model .
this is not surprising , as fig .
[ fig2]a shows that , due to positive neutral correlations , @xmath102 for most phenotypes in each system ( @xmath103 for @xmath44 , @xmath100 is shown as a dotted red line in the plot ) .
once the probability of having a neutral neighbour is above the @xmath95 threshold , we expect large networks . for hp24 and rna12 , the largest neutral component size clearly grows linearly with frequency , and so scales linearly with the size of the neutral set . for the polyomino space
@xmath28 this scaling is less evident , but the components are still much larger than their random counterparts would be . secondly , for all three models , the number of components does not vary much with @xmath53 , in contrast to the random model where this number scales , as expected , with the neutral set size if @xmath104 . since these components typically have robustness above @xmath95 or even @xmath97 , the reason there are still multiple components must be due to biophysical constraints which are not present in the random model , such as the neutral non - reciprocal sign epistasis discussed earlier for rna .
these effects are to first order independent of @xmath53 which explains why the number of components does not correlate with @xmath53 . in each of these three models
the largest `` phenotype '' of all is the deleterious non - folding or non - assembling one .
its frequency exceeds the threshold @xmath97 and its neutral set is fully connected .
differences between the three biophysical gp maps observed in fig .
[ fig2 ] can be fairly easily explained by some of the differences highlighted in the methods section .
for example , the number of phenotypes per genotype is largest in the hp model , and smallest in the polyomino model , which explains why they group at different frequencies . while the number of components in fig .
[ fig2 ] b ) is generally much lower in the biological models than it is in the random model , the hp model has significantly fewer components than the rna and polyomino models do .
this may be due to the fact that the hp model does not exhibit effects such as neutral reciprocal sign epistasis , which fragments neutral sets in the other two systems .
we conclude that due to positive neutral correlations and the concomitant higher robustness , the biophysical models considered here have large neutral networks even for frequencies @xmath53 that are several orders of magnitude lower than the random model large component threshold @xmath95 .
the abject failure of the random model to predict the robustness and the neutral network size highlights the importance of neutral correlations in these systems .
we next consider non - neutral mutations .
the first question is : are two different phenotypes , on average , more or less likely to be connected to each other than one would expect by chance ? to address this question , we employ a generalisation of robustness , namely the _ phenotype mutation probability _ @xmath105 of @xmath64 with respect to @xmath54 , defined as the fraction of 1-point mutations of genotypes in the neutral set for phenotype @xmath54 that map to phenotype @xmath64 .
this can be written as : @xmath106 thus @xmath105 averages a local property , @xmath107 the number of genotypes that map to phenotype @xmath64 found the 1-mutation neighbourhood of a genotype that maps to phenotype @xmath54 over the entire neutral set @xmath59 .
note that this phenotype mutation probability is not symmetric ( @xmath108 ) and that , if @xmath109 , it reduces to the phenotype robustness @xmath110 .
it has recently been shown @xcite that @xmath105 is a key quantity for incorporating the structure of a gp map into population genetic calculations . in the null model
we expect @xmath111 to be an excellent approximation @xcite , with the caveat that it must be possible for enough genotypes to be sampled .
what do we mean by enough genotypes ? given a phenotype @xmath54 with redundancy @xmath57 , there are at most @xmath112 unique neighbours available .
this number provides an upper bound in reality , several neighbours of one genotype will also be neighbours of another genotype with the same phenotype , resulting in a reduction in the number of unique neighbours .
nevertheless , this allows us to define a minimum threshold @xmath113 if @xmath114 , then the expected number of genotypes with phenotype @xmath64 found around phenotype @xmath54 is less than one , and the probability that @xmath115 due to statistical fluctuations becomes appreciable .
further detail on how @xmath105 and @xmath57 relate when the threshold is not satisfied , which is mainly relevant for smaller gp maps and for lower @xmath57 , is provided in appendix [ s1_text ] . here
we focus on phenotypes with larger @xmath57 , in the larger gp maps of the previous section . in fig .
[ fig4 ] , we plot the relationship between the phenotype mutation probability @xmath105 and global frequency @xmath116 around the rna20 phenotype with the second largest neutral set , the assembling phenotype for @xmath31 with the largest neutral set , and the hp5x5 folding phenotype with the largest neutral set . for phenotypes in @xmath31 and hp5x5 , with such large numbers of genotypes ,
all phenotypes have @xmath116 values that are significantly above @xmath117 ( vertical dotted lines ) , which is the approximate threshold at which at least one genotype of phenotype @xmath64 would be expected to be found .
a small fraction of phenotypes lie close to the @xmath117 threshold for rna20 , but by far the majority may be expected to be effectively sampled . for rna20 and @xmath31 , we observe a very strong and highly significant positive correlation with the random null model expectation @xmath111 . in hp5x5 , there is also a strong positive correlation , though less strong than in the rna and polyomino cases , with a greater number of phenotypes falling below the one - to - one expectation .
we did not plot the non - compact model hp24 because most of its frequencies are below the threshold @xmath118 ( see supporting information ) . to summarise , in contrast to the robustness @xmath119 where neutral mutations lead to strong deviations from the null model , the non - neutral phenotype mutation probabilities follow the random model expectation that @xmath120 remarkably well .
there are still important deviations , especially for those phenotypes that can not be reached due to biophysical constraints so that @xmath121 @xcite .
moreover , it may be an interesting exercise to look more closely at phenotypes for which @xmath105 is significantly greater or less than @xmath116 as such deviations could signal similarities or differences between phenotypes .
for example , two rna phenotypes with similar hairpin topology , but perhaps a difference of one bond in a stem may have a larger probability of interconverting than topologically more dissimilar rna phenotypes .
the difference between @xmath105 and @xmath116 could then be used to quantify the difference between phenotypes @xmath54 and @xmath64 .
these more subtle types of correlation are beyond the scope of this paper . at any rate ,
compared to the result in the previous sections showing the strength of neutral correlations , the dominant agreement with the random model is apparent .
however , given that @xmath105 is averaged over a neutral set , it may be that there are _ local _ non - neutral correlations that are obscured by the averaging . with this in mind
, we next investigate such local correlations .
we first describe _ non - neutral local over - representation correlations _ which mean that , given phenotype @xmath64 is found in the 1-mutation neighbourhood of a genotype @xmath61 ( which maps to phenotype @xmath122 ) , then phenotype @xmath64 will appear a greater number of times in total than predicted by @xmath116 or @xmath105 in this 1-mutation neighbourhood , as pointed out in ref .
these correlations are illustrated in fig .
[ fig5]a . to measure 1-mutation neighbourhoods
, we sample randomly chosen genotypes @xmath61 from the neutral set @xmath59 , with a genotype of phenotype @xmath64 in its neighbourhood .
we then measure the phenotype of all other neighbours of @xmath61 . from this sample ,
we obtain the probability @xmath123 of @xmath64 occurring @xmath124 times in the 1-mutation neighbourhood of a genotype mapping to phenotype @xmath54 , given that @xmath64 occurs at least once .
two control null expectations may also be derived for @xmath123 . in the random model where phenotypes are randomly assigned , given @xmath64
is in the 1-mutation neighbourhood of a genotype @xmath61 ( at a specific genotype @xmath125 ) , the probability may be calculated as a binomial probability based upon the overall frequency of @xmath64 , leading to @xmath126 a second null expectation calculates the binomial probability by replacing @xmath116 in eq .
[ eq : null1 ] above by replacing the phenotype mutation probability @xmath105 for the gp map instead : @xmath127 in contrast to @xmath128 , this form accounts for any overall phenotypic heterogeneity known to be present in the gp map .
we compare the actual local prevalence against these two null expectations in fig .
for rna20 , @xmath31 and hp5x5 we chose the same three phenotypes for phenotype @xmath64 as we did in the previous section , while for phenotype @xmath54 we choose one - by - one the next @xmath129 largest ( non - deleterious ) phenotypes available in the gp map . by sampling 10,000 neighbourhoods for each of the @xmath129 phenotypes for @xmath54
, we calculate an average for @xmath123 across the phenotypes ( @xmath130 ) and compare this in fig .
[ fig6 ] to the averages for the null expectations @xmath131 and @xmath132 . for each biological gp map , @xmath64 is more likely to be over - represented , that is to appear multiple times if it appears at least once when compared to the null expectations , leading to a skewed distribution compared to the control case .
the most striking result is seen in rna20 , where there is a substantial tail to the distribution .
we use average measures here to provide the general profile , smoothing out particular features that may occur between individual pairs of phenotypes @xmath64 and @xmath54 , but the local over - representation is seen for any of the phenotype pairs considered .
one consequence of these local over - representation correlations is that the probability that a genotype with phenotype @xmath54 has phenotype @xmath64 _ at least once _ in its 1-mutant neighbourhood is less than expected from @xmath105 .
this is because those genotypes that have phenotype @xmath54 in their 1-mutant neighbourhood typically do so a greater number of times than expected .
this must therefore be compensated with fewer genotypes around which phenotype @xmath54 actually appears at all , which we confirm numerically . in the rna20 gp map , with
the most frequent of the set of phenotypes used for @xmath54 and the next most frequent used as @xmath64 , the probability of finding @xmath64 at least once is @xmath133 versus a null expectation of @xmath134 .
thus these correlations lead to heterogeneity in the connections between phenotypes .
how these correlations affect evolutionary dynamics will depend on the regime being explored @xcite .
if the population is neutrally exploring genotypes that map to phenotype @xmath54 , then in the monomorphic regime of evolutionary dynamics , where @xmath135 , this heterogeneity will lead to a significant drop in the rate at which @xmath64 is first discovered by neutral exploration .
in the polymorphic regime where @xmath136 , and different individuals in the population have different genotypes , the rate at which novel variation with phenotype @xmath64 occurs may not be that different from the expectation given by @xmath105 , at least if the population is spread across a large enough number of different genotypes to average over local heterogeneity @xcite .
we next examine _ non - neutral local mutational neighbourhood correlations _ which are illustrated schematically in fig .
they show that the 1-mutation neighbourhoods of two genotypes connected by a neutral point mutation are more likely to have similar phenotypic compositions than would be expected by two randomly chosen neutral non - neighbouring genotypes of the same phenotype .
this type of correlation has already been demonstrated to exist for rna @xcite . to measure the similarity of neighbouring genotypes mutational neighbourhoods
, we consider the local quantity @xmath137 , which becomes @xmath105 when averaged over the whole neutral set @xmath59 .
we compare the @xmath138 for neighbouring genotypes with non - neighbouring genotypes in both the null model and biological gp maps .
the similarity or difference could be measured in several different ways .
the statistical measure we employ here is the bhattacharyya coefficient @xcite , which for two discrete probability distributions @xmath139 and @xmath140 may be expressed as @xmath141 varying between 0 and 1 for maximally dissimilar and identical discrete probability distributions respectively . to quantify whether neutral neighbours @xmath61 and
@xmath142 have more similar phenotype distributions in comparison to non - neighbouring neutral genotype pairs @xmath61 and @xmath143 , we compared the _ similarity ratio _ of the bhattacharyya coefficients , @xmath144 , using the @xmath138 to define the distributions .
a ratio greater than unity indicates that the phenotype distributions around neutral neighbours are more similar than the randomly selected neutral pair , and vice versa .
we remove the @xmath145 mutual neighbours of @xmath61 and @xmath142 from the distributions as these will automatically contribute to similarity between the neighbourhoods in a trivial manner which we wish to exclude . in fig .
[ fig7 ] we plot histograms of the similarity ratio for 10,000 samples of @xmath61 , @xmath142 and @xmath143 in rna20 , @xmath31 and hp5x5 , where the phenotype sampled has the second largest frequency in rna20 , and the largest frequency in @xmath31 and hp5x5 ( excluding the del phenotype ) . for 10,000 samples
the means are @xmath146 for rna20 , @xmath147 for @xmath31 and @xmath148 for hp5x5 , where the error is the standard error on the mean . for rna20 and @xmath31 , a clear skew in the overall distribution may be visually observed , demonstrating that neutral neighbours , on average , have more similar mutational neighbourhoods .
hp5x5 also has the mean of its distribution at a value slightly larger than unity but it is much more marginal in this case , and the skew is harder to detect .
we note that in general , the non - neutral correlations are weakest for the hp5x5 model .
finally , just as is the case for the non - neutral local over - representation correlations of the previous section , these local mutational neighbourhood correlations also reduce the rate at which novel phenotypes would be discovered by neutral exploration since a neutral neighbour is more likely to have some of the same phenotypes in its mutational neighbourhood , and so fewer alternatives .
the final , and perhaps most important , type of non - neutral correlation we consider is the accessibility of the deleterious phenotype from folding or self - assembling phenotypes , which we call _ non - neutral deleterious phenotype correlations_. this type of non - neutral correlation is closest to the type of correlation suggested by maynard smith @xcite . in fig .
[ fig8 ] we plot histograms of the ratio @xmath149 for all phenotypes @xmath54 in @xmath31 and hp5x5 , and the top 20 most frequent ( largest @xmath53 ) in rna20 ( limited due to computational expense of this larger system ) . in all cases , we see that the deleterious phenotype is significantly less frequent around the non - deleterious phenotypes .
this behaviour contrasts to non - deleterious phenotypes , for which @xmath150 . as a corollary of this effect
, we also find @xmath151 equal to @xmath152 , @xmath153 and @xmath154 for rna12 , @xmath31 and hp5x5 respectively , illustrating positive correlations , that is , a corresponding local over - representation of the deleterious phenotype in its own mutational neighbourhoods .
moreover , for @xmath155 we find that @xmath156 suggesting that these positive neutral correlations may become stronger for larger @xmath4 .
we find that the del phenotype forms only a single component in rna12 , @xmath28 , hp24 and hp5x5 .
this result is unsurprising because the large size of the del phenotype in each gp map ( 85% , 54% , 98% and 82% respectively ) means that the frequencies are all well above the single component threshold @xmath97 of eq .
[ eq : lambda ] , which would lead to the expectation of a single component even in the random null model .
in this paper , we have explored the role of genetic correlations , which we defined and quantified as the difference in how genotypes are mutationally connected for biologically relevant gp maps , compared to a random null model with the same global properties ( alphabet size , genome length , and number of genotypes per phenotype ) .
genetic correlations provide a simple conceptual framework within which a number of topological properties of gp maps can naturally be captured .
* neutral correlations : * we first explored the phenotype robustness for all three gp maps , showing that @xmath70 for all phenotypes , a result which is not unexpected in the literature , but to our knowledge has not been compared for a set of whole gp maps before . since @xmath71 for the random model , the extent to which @xmath63 differs from @xmath53 can be viewed as a measure of the extent of the neutral correlations .
we also introduced the concept of @xmath47-robustness , which measures robustness over @xmath47 mutations . from this
we derived another criterion that measures the presence of neutral correlations by averaging this measure over all phenotypes and comparing to the null expectation : if @xmath157 then there are positive neutral correlations .
we find that the enhanced probability of encountering a genotype mapping to the same phenotype can extend to multiple mutations @xmath47 away from genotypes .
the extent of the correlations in sequence space can be quantified by the number of mutations @xmath158 at which the criterion is violated , a measure we call the correlation length of the neutral mutations .
we find that that @xmath158 is largest for the rna12 model , and smallest for the hp24 model .
how @xmath158 or even the relative ordering of the correlation lengths between the different systems will scale with increasing genome length @xmath4 remains an open question .
it should also be emphasised that the full complexity of neutral correlations for a phenotype are only partly be captured with the measures we introduced here , which average over the entire neutral set . as can be seen in fig 3a of ref .
@xcite , a single neutral set can have significant local heterogeneity in its internal connections . since neutral sets are frequently so vast that they can not be fully explored by populations on evolutionary time - scales , such local heterogeneities may also have implications for evolutionary dynamics .
thus local measures of heterogeneity and robustness , which may be influenced , for example , by the identities of multiple neighbours , or the position of a mutation along a genome , may also be important to develop in future work
. we found , for the three biological gp maps that we studied , that the dominant relationship of robustness with frequency is @xmath159 , a scaling that has already been pointed out earlier for rna @xcite . in an interesting paper that applies concepts from network theory @xcite to neutral sets ,
aguirre et .
@xcite rationalise this scaling for rna by separating out the mutational behaviour of bound and unbound bases .
it would be interesting to see if a more general argument could be developed to explain the logarithmic scaling across all the systems we studied .
moreover , these results also pose fascinating questions relating to why or how the constraints in a gp map lead to the kinds of neutral correlations they do .
some clues to the underlying causes of neutral correlations and robustness can be gleaned from maynard smith s original paper , where he illustrated the concept of a neutral network with the parlour game of transitioning between connecting two words in the english language by changing one letter at a time , with each change also generating a valid word .
he used the example of changing `` word '' to `` gene '' in four steps as illustrated in fig .
there are @xmath160 different possible 4-letter words , but , according to the merriam - webster official scrabble players dictionary 2014 , only @xmath161 , or just over @xmath162 , are valid english words .
if we consider the set of valid words to be a phenotype , then it has frequency of only @xmath163 , just under the giant component threshold @xmath164 , and well below the single component threshold @xmath165 . on average the probability that a 4-letter word has a valid neighbour is just below one .
however , if we measure the phenotype robustness , that is the mean probability that a valid 4-letter word has valid words in its @xmath166 neighbours , we find that @xmath167 , or on average each word has 11 neighbours , which makes the game much simpler than if the random expectation held .
the reasons this game exhibits such a large enhancement of the robustness over @xmath53 clearly arise from neutral correlations in english words .
for example , vowels are more likely to appear in specific places in words than would be expected by chance .
the second letter of 4-letter english words has a 74% chance of being a vowel compared to the 5/26 = 19% overall average probability per locus .
so if a word has a vowel placed at the second letter , it is much more likely to have neighbouring words using the same vowel , as can be seen in fig .
this example illustrates how basic properties ( in this case of language properties , in the case of our models , biophysical properties ) can generate correlations , which can then result in high robustness .
a question often debated in the literature is the extent to which mutational robustness is selected for . here we argue that a major enhancement of robustness , often by many orders of magnitude over the random expectation , is not caused by selection , but rather emerges from the internal constraints of a gp map the way that genotypes map to phenotypes which naturally lead to positive neutral correlations
. it may still be the case that more robust genotypes can be selected for within a neutral set , or that these genotypes are favoured in certain dynamic regimes @xcite .
it may also be true that in some cases a particular phenotype is preferred by selection because it is more robust than an alternative one .
but even if this is so , it is important to keep in mind that natural selection is still acting on variation that is already naturally quite robust due to correlations caused by biophysical constraints .
the relationship between robustness and evolvability has been the subject of much discussion in the literature @xcite . here
we show , as already anticipated by maynard smith @xcite , that if the phenotype robustness is roughly larger than @xmath168 , so that the expected number of neutral neighbours is greater than one , then the phenotype will exhibit large neutral networks
. in the random model , large networks will typically be very rare , but neutral correlations mean that robustness above the @xmath95 threshold is common for the biophysical gp maps .
the effect can be very large .
for example , for @xmath169 rna , a recent study @xcite suggests that there are about @xmath170 phenotypes , so that the mean frequency is @xmath171 .
in fact all phenotype frequencies are well below the threshold @xmath172 above which we expect extended neutral networks . on the other hand ,
the mean robustness of all phenotypes was estimated to be @xmath173 .
neutral correlations increase the probability of a nearest neighbour generating the same phenotype by on average about @xmath174 orders of magnitude over the mean expectation of the null model , lifting robustness well above the threshold @xmath95 .
thus the most important way that neutral correlations contribute to evolvability is by naturally creating robustness greater than the threshold needed to generate percolating networks which provide access to phenotypic novelty .
in fact it may very well be that without neutral correlations and its attendant robustness , evolution as we know it would not be possible * non - neutral correlations : * _ non - neutral mutations _
are important for the generation of novel variation .
for all three gp maps , the probability @xmath105 that a phenotype @xmath64 is found by a point mutation from genotypes mapping to phenotype @xmath54 is , to first order , given simply by the global frequency : @xmath175 , which is independent of @xmath54 . since @xmath116 can span many orders of magnitude , the rate at which variation appears ( which scales as @xmath176 if a population is neutrally exploring phenotype @xmath54 @xcite ) can also range over many orders of magnitude in these systems .
these large differences can lead , both in the monomorphic and polymorphic regimes , to effects such as the _ arrival of the frequent _
@xcite , where frequent phenotypes ( with larger @xmath116 ) fix in a population even when alternate phenotypes that are much more fit , but much less frequent , are accessible in principle .
the reason these fitter phenotypes are not fixed is because they are unlikely to be found on evolutionary time - scales .
natural selection can only work on variation that actually arises .
in the alternative case where the system is effectively in steady state , so that a less frequent phenotype has a realistic probability to arise in a population , it can still be the case , especially at larger mutation rates , that a phenotype with lower fitness but larger frequency ( and robustness ) will fix , an effect known as the _ survival of the flattest _
@xcite .
finally , we note that @xmath105 can be viewed as a non - neutral generalisation of the phenotypic robustness , but that @xmath177 scales very differently with @xmath53 than @xmath105 does when @xmath122 . in the latter case local correlations more or less cancel out when averaged over the whole neutral set , so that @xmath175 , while in the former case the local correlations do not cancel out at all because robustness is fundamentally a local quantity .
it is quite striking that in all three models , a very large number of phenotypes are indeed connected to one another .
the hp model merits further discussion in this regard . in a recent review @xcite ,
rna space was compared to `` a bowl of spaghetti '' , because the neutral spaces were connected to most other phenotypes , while proteins were compared to a``plum pudding '' , where the neutral networks were more likely to be isolated from one another .
we indeed find that the neutral networks in the hp24 model are not well connected , but locate the origin of this effect in the large @xmath52/@xmath56 ratio for hp24 , which means that many networks are below the threshold of eq .
( [ eq : sample_threshold ] ) for connections .
by contrast , the compact hp5x5 model with many fewer phenotypes but a similar sized genotype space is well connected , more like `` spaghetti '' than like a `` plum pudding '' . what happens for real proteins , without the simplifying assumptions and small system sizes typically studied in the hp model @xcite , remains an open question .
another type of heterogeneity in the mapping of genotypes to phenotypes can be quantified as _
local _ non - neutral correlations , which occur when the local neighbourhood of genotypes are different from the global expectation given by @xmath105 or @xmath116 .
we investigated two types of correlation ( although one could imagine many more ) :
i ) non - neutral local over - representation correlations which result in phenotypes being more likely to be found multiple times around genotypes , and ii ) non - neutral local mutational neighbourhood correlations , which mean that two genotypes connected by a neutral point mutation have mutational neighbourhoods that are more similar than do two randomly selected genotypes in a neutral set .
these two types of correlation mean that the diversity of phenotypes in the direct neighbourhood of a genotype is lower than expected from the random model or even from the averaged phenotype mutation coefficients @xmath105
. thus the rate at which a neutrally exploring population encounters novel variation will be reduced due to these correlations .
how this effect influences evolvability is complex , because the term is used in many different ways in the literature @xcite .
one type of evolvability simply measures the number of different phenotypes that are connected by single mutations to a neutral set @xcite . while non - neutral correlations may not affect this number very much
, they will affect the rate at which neutral exploration finds these new phenotypes .
this lowering of the rate at which novelty appears may have a larger impact on other measures of evolvability .
each of the three models has a deleterious phenotype which either does not fold ( for rna and the hp protein model ) or does not properly assemble ( in the polyomino model for protein clusters ) .
the third type of non - neutral correlations we considered were iii ) non - neutral deleterious phenotype correlations . for all three gp maps ,
the folding or assembling phenotypes have fewer mutational connections to the deleterious phenotypes than would be expected by the global frequency @xmath178 .
this last result is perhaps the most interesting type of non - neutral correlation .
it was already predicted by john maynard smith in his classic 1970 paper @xcite , where he argued that
meaningful `` proteins were more likely to be neighbours of other ' ' meaningful `` proteins , and by extension , that the probability of finding a deleterious phenotype in the mutational neighbourhood of a ' ' meaningful " protein would be less than by random chance .
such an effect can enhance evolutionary dynamics , because non - deleterious phenotypes are more strongly connected by mutations than expected by random chance , and so the population can more easily access potentially meaningful novel variation .
of course in practice , whether or not even the folding or self - assembling phenotypes are in fact `` meaningful '' will depend on the environment and other factors , but to first order a reduced propensity to mutate to manifestly deleterious phenotypes should be an evolutionary advantage .
* evolvability : * while the effect of neutral correlations on robustness is straightforward , how correlations affect evolvability is more complex , not just because the concept itself is more diffuse , but also because the relationships between correlations and evolvability are more varied .
nevertheless , we can summarise how different correlations affect evolvability as follows : 1 .
positive neutral correlations , measured by the presence of greater phenotypic robustness than would be expected by chance , is critical for the formation of large neutral networks .
these networks are , in turn , a key facilitating factor for the ability of a population to access novel variation by neutral evolution over the network @xcite . without positive neutral correlations , and the associated phenotype robustness
, evolvability would be hugely suppressed .
2 . the non - folding or non - assembling deleterious set of phenotypes
are ( positively ) neutrally correlated and anti - correlated with the set of folding or assembling phenotypes .
this correlation increases the potential phenotypic variation that is accessible by reducing the likelihood of mutations leading to a seriously deleterious phenotype .
local non - neutral correlations generally mean that the amount of novel variation available after mutations is smaller than one might expect from a random model .
these correlations will reduce evolvability .
for example , huynen et al .
@xcite showed that the innovation rate for a random walk on the neutral space of the @xmath179 secondary structure of trna@xmath180 is about @xmath181 , even though each genotype has @xmath182 neighbours .
this significant lowering of the innovation rate is due to local non - neutral correlations , and may be a more general effect .
4 . for all three gp maps we found that @xmath183 , which means that the exponential variation in the frequency of phenotypes is reflected in the probability that a novel phenotype is found by mutations . as argued in ref .
@xcite , for a range of evolutionary scenarios , phenotypes with large @xmath53 will be exponentially more easy to find than phenotypes with smaller @xmath53 .
these large differences in the rate at which novel variation arrives will strongly modulate the evolvability .
* other measures of genetic correlations : * while we have built upon and introduced a number of metrics for genetic correlations in gp maps , the framework of correlations has other avenues for future computational work .
for example , the central focus here has been on the way phenotypes are neutrally connected and non - neutrally connected without any broader concern for the properties of the phenotypes themselves .
phenotypes themselves have measurable properties such as symmetry , size and modularity and one could take the analysis further by considering whether there is a relationship between the mutational connections between two phenotypes and their similarity based on such properties . making use of the null model again , where there is no predisposition for which phenotypes are mutationally close together , such _ phenotype similarity correlations _ could be studied in the biological gp maps . * measuring correlations in experiment : * computational studies on theoretical systems ultimately need to be backed up with empirical evidence in real biological systems .
robustness to mutations in protein tertiary structure has been a well - studied area in this regard , with both mutagenesis and phylogenetic experiments being used to illustrate robustness @xcite .
it may be that non - neutral local correlations could be verified using mutagenesis experiments .
for example , for mutational neighbourhood correlations , two neighbouring genotypes both with a chosen structure could have their neighbourhoods examined for the range of phenotypes and compared to the neighbourhood of a more distant genotype with the same methodology used here computationally , potentially replicating the findings fig .
[ fig7 ] , but for real molecules .
it may also be possible to measure the neutral correlation length @xmath158 by doing multiple mutation experiments .. however , because it is hard in practice to extract full gp map properties such as @xmath53 for a given phenotype , the most challenging aspect of such an experiment would be in generating an appropriate effectively random expectation .
this same challenge holds for the other kinds of correlations we investigate in this paper .
a few final caveats are in order . in these models
it is natural to use a restricted definition of a neutral mutation leading to exactly the same phenotype , whereas a more complete theory would count all mutations that are not visible to selection as effectively neutral .
thus the full picture of how these correlation affect evolutionary dynamics is complex , and depends not just on the gp map itself , but more generally on the genotype to phenotype to fitness map , for which the environment plays a key role .
moreover , population genetic parameters such as the population size and the mutation rate must be taken into account .
but notwithstanding these complications , the important influence that structure in the gp map , in this case measured through the lens of genetic correlations , has on the manner in which variation arises ( the arrival of the fittest " @xcite ) , and so on evolutionary dynamics , should be evident , confirming maynard smith s suggestions from many years ago
. it may even be that without these correlations , darwinian evolution , and therefore life itself , would not have been possible .
the authors would like to thank edward rolls for his initial analysis of john maynard smith s word game .
this work was funded under ep / p504287/1 by the engineering and physical sciences research council ( https://www.epsrc.ac.uk ) .
sea is supported by the royal society ( https://royalsociety.org/ )
. 10 smith jm .
natural selection and the concept of a protein space . nature .
1970;255:563564 .
kimura m , et al . evolutionary rate at the molecular level . nature .
1968;217(5129):624626 .
king j , jukes t. non - darwinian evolution .
science ( new york , ny ) .
1969;164(881):788 .
lipman dj , wilbur wj .
modelling neutral and selective evolution of protein folding .
proceedings of the royal society of london series b : biological sciences .
1991;245(1312):711 .
theory for the folding and stability of globular proteins . biochemistry .
1985;24(6):15011509 .
lau kf , dill ka . a lattice statistical mechanics model of the conformational and sequence spaces of proteins .
1989;22(10):39863997 .
schuster p , fontana w , stadler pf , hofacker il . .
proceedings : biological sciences . 1994;255(1344):279284 .
hofacker il , fontana w , stadler pf , bonhoeffer ls , tacker m , schuster p. fast folding and comparison of rna secondary structures .
monatshefte fr chemie / chemical monthly
. 1994;125(2):167188 .
mathews dh , disney md , childs jl , schroeder sj , zuker m , turner dh .
incorporating chemical modification constraints into a dynamic programming algorithm for prediction of rna secondary structure .
proceedings of the national academy of sciences of the united states of america . 2004;101(19):72877292 .
reuter js , mathews dh .
rnastructure : software for rna secondary structure prediction and analysis .
bmc bioinformatics .
2010;11(1):129 .
fontana w. . bioessays
. 2002;24(12):11641177 .
cowperthwaite mc , economo ep , harcombe wr , miller el , meyers la . the ascent of the abundant : how mutational networks constrain evolution .
plos computational biology . 2008;4(7):e1000110 .
aguirre j , buld jm , stich m , manrubia sc .
topological structure of the space of phenotypes : the case of rna neutral networks .
plos one . 2011;6(10):e26324 .
schaper s , louis aa .
the arrival of the frequent : how bias in genotype - phenotype maps can steer populations to local optima .
plos one . 2014;9(2):e86635 .
wagner a. .
princeton university press princeton , nj : ; 2005 .
wagner a. the origins of evolutionary innovations : a theory of transformative change in living systems .
oxford university press ; 2011 .
ciliberti s , martin oc , wagner a. robustness can evolve gradually in complex regulatory gene networks with varying topology .
plos computational biology .
2007;3(2):e15 .
raman k , wagner a. .
mol biosyst .
2011;. samal a , rodrigues jfm , jost j , martin oc , wagner a. genotype networks in metabolic reaction spaces .
bmc systems biology .
2010;4(1):30 .
greenbury sf , johnston ig , louis aa , ahnert se .
a tractable genotype phenotype map modelling the self - assembly of protein quaternary structure .
journal of the royal society interface . 2014;11(95):20140249 .
ohta t. slightly deleterious mutant substitutions in evolution . nature .
1973;246(5428):9698 .
wagner a. .
proceedings of the royal society b. 2008;275(1630):91 .
wright sg . .
proc 6th int cong genet .
1932;1:356366 .
schultes ea , bartel dp .
one sequence , two ribozymes : implications for the emergence of new ribozyme folds . science .
2000;289(5478):448452 .
hayden ej , ferrada e , wagner a. cryptic genetic variation promotes rapid evolutionary adaptation in an rna enzyme .
2011;474(7349):9295 .
payne jl , wagner a. the robustness and evolvability of transcription factor binding sites .
2014;343(6173):875877 .
ferrada e , wagner a. a comparison of genotype - phenotype maps for rna and proteins .
biophysical journal .
2012;102(8):19161925 .
wagner a. the molecular origins of evolutionary innovations . trends in genetics
. 2011;27(10):397410 .
ahnert s , johnston i , fink t , doye j , louis a. .
physical review e. 2010;82(2):026117 .
johnston ig , ahnert sa , doye jpk , louis aa .
evolutionary dynamics in a simple model of self - assembly .
physical review e. 2011;83:066105 .
irbck a , troein c. enumerating designing sequences in the hp model .
journal of biological physics .
2002;28(1):115 .
li h , helling r , tang c , wingreen n. .
1996;273(5275):666 .
dingle k , schaper s , louis aa . .
2015;. schram rd , schiessel h. exact enumeration of hamiltonian walks on the 4@xmath184 4@xmath184 4 cube and applications to protein folding .
journal of physics a : mathematical and theoretical . 2013;46(48):485001 .
reidys c , stadler pf , schuster p. generic properties of combinatory maps : neutral networks of rna secondary structures
. bulletin of mathematical biology .
1997;59(2):339397 .
schaper s , johnston ig , louis aa .
epistasis can lead to fragmented neutral spaces and contingency in evolution .
proceedings of the royal society b : biological sciences .
rspb20112183 .
schuster p. mathematical modeling of evolution .
solved and open problems .
theory in biosciences .
newman m. networks : an introduction .
oxford university press ; 2010 .
sumedha , martin oc , peliti l. population size effects in evolutionary dynamics on neutral networks and toy landscapes .
j stat mech : theory and experiment .
2007;2007(05):p05011 .
bhattacharyya a. on a measure of divergence between two multinomial populations .
sankhy : the indian journal of statistics ( 1933 - 1960 ) .
1946;7(4):pp . 401406 .
jorg t , martin oc , wagner a. .
bmc bioinformatics . 2008;9(1):464 . van nimwegen e , crutchfield jp , huynen m. neutral evolution of mutational robustness .
proceedings of the national academy of sciences . 1999;96(17):97169720 .
masel j , siegal ml .
robustness : mechanisms and consequences . trends in genetics .
2009;25(9):395403 .
draghi ja , parsons tl , wagner gp , plotkin jb . .
2010;463(7279):353355 .
wilke co , wang j , ofria c , lenski re , adami c. . nature .
2001;412:331333 .
goldstein ra .
the structure of protein evolution and the evolution of protein structure .
current opinion in structural biology . 2008;18(2):170177 .
conrad m. the geometry of evolution .
biosystems . 1990;24(1):6181 .
wagner gp , altenberg l. perspective : complex adaptations and the evolution of evolvability . evolution .
1996;p . 967976 .
kirschner m , gerhart j. evolvability .
proc natl acad sci usa .
1998 july;95(15):84208427 .
dawkins r. the evolution of evolvability . artificial life . 1989;p
. 201220 .
pigliucci m. nature reviews genetics .
2008;9(1):7582 . huynen ma , stadler pf , fontana w. smoothness within ruggedness : the role of neutrality in adaptation .
proc nat acad sci usa .
1996;93(1):397 .
wagner a. arrival of the fittest : solving evolution s greatest puzzle .
penguin ; 2014 .
schuster p , fontana w , stadler pf , hofacker il . from sequences to shapes and back : a case study in rna secondary structures .
proceedings of the royal society of london b : biological sciences . 1994;255(1344):279284 .
we begin with an analysis of the sampling threshold @xmath118 from average phenotypes in a given gp map .
we showed in eq .
that for effective sampling , we require @xmath185 which may be expressed alternatively as @xmath186 the average phenotype frequency may be written down as @xmath187 or with respect to the genotype sampling distribution as @xmath188 substituting in the smaller of the two , the average phenotype frequency , and then considering the required threshold for effectively sampling phenotype @xmath64 from the average phenotype @xmath54 , we find @xmath189 for rna , where empirical scaling values are known ( @xmath190 @xcite ) , we can further write @xmath191 for a phenotype @xmath64 to be effectively sampled . we can change the question of effective sampling to ask the conditions on @xmath52 for the average phenotype @xmath64 to be accessed from the average phenotype @xmath54 . in these circumstances , we can see that @xmath192 for which we can see rna satisfies to an increasing extent for increasing @xmath4 , as @xmath193 monotonically increases with increasing @xmath4 .
finally , we can also write down the approximate fraction less than the average phenotype frequency that is accessible from an average phenotype , through expressing @xmath194 , which leads to a threshold fraction @xmath195 again , using rna as an example system this leads to @xmath196 for a given length of rna , showing that an increasing fraction of phenotypes with frequencies below the average may be effectively sampled from the average phenotype . [ s1_fig ]
in this section , in contrast to the analysis in the main body considering the frequency of phenotypes @xmath64 around phenotype @xmath54 where @xmath198 , we consider different @xmath54 across a range of phenotype frequencies @xmath53 in the gp map . in the random null model , at values of @xmath199 , we expect phenotypes to almost exactly follow @xmath111 .
when @xmath200 , there are two likely possibilities for a given phenotype : either the phenotype is not found at all ( @xmath121 ) or it is found a single time ( @xmath201 ) .
the latter is an over - representation of the local prevalence of @xmath64 for the gp map , while the former is clearly no local representation at all . in fig .
10 , we display three pairs of plots for the rna12 , @xmath28 and hp24 gp maps and a randomised null model counterpart .
the null models are displayed on the left , with the actual gp maps on the right . in each plot , we show the values of @xmath105 against @xmath116 for three different phenotypes @xmath54 ( coloured by data point as red , blue and green , with red the largest frequency phenotype and blue the smallest ) .
upward triangular data points represent values for @xmath202 , downward triangular data points @xmath121 ( shown at 1e-8 for visualisation purposes only ) and the circular data points are all other phenotypes .
vertical and horizontal dotted coloured lines represent @xmath117 and @xmath201 respectively .
the diagonal dashed black line is @xmath203 , the null expectation for phenotypes with @xmath204 .
we begin by discussing the behaviour of the null model .
the @xmath28 and hp24 null models provide the extreme cases . for @xmath28 ,
all phenotypes are highly frequent and have @xmath205 .
consequently , we see that each of the three phenotypes follows the expected trend of @xmath203 to a very high degree of accuracy ( spearman rank correlation coefficient and p - value in the top left ) . for the hp24 null model ,
all frequencies are such that @xmath206 . as such , phenotypes that are found
locally are found only once ( @xmath207 ) and most are not found at all ( the many downward triangular points ) .
for the rna12 null model , the frequency of phenotypes used for phenotype @xmath54 span the range of all @xmath205 ( red and green ) and to some phenotypes having @xmath208 ( blue ) .
as a result , we see the red and green phenotypes follow @xmath111 strongly , while the tail of the blue phenotype has fluctuations between the three behaviours .
the results from the null models demonstrate the accuracy of the above outlined intuition for a null relationship between the local connectivities of phenotypes with respect to the global abundance . with this in mind
, we can now draw direct comparisons between each phenotype in the null model and actual behaviour exhibited in the biological gp maps .
for each of the gp maps , we plot the same phenotype as in the null model case . for rna12 , positive correlations are found for each phenotype , with deviations from @xmath203 being more pronounced for lower frequency phenotypes ( blue is subject to much greater fluctuations than red ) .
the fluctuations are approximately up to an order of magnitude either side of the @xmath203 .
we see a similar behaviour for @xmath28 , with the largest fluctuations exhibited for the low frequency blue phenotype . finally , we consider the biological hp24 gp map . as was the case in the null model version ,
every phenotype ( apart from the deleterious phenotype ) lies in the region where @xmath206 .
the notable difference in the actual gp map is the tendency for phenotypes with a larger @xmath116 to also be more likely to be locally present ( @xmath209 )
. we can understand this with the following rationale : due to the neutral correlations present , if a single genotype with phenotype @xmath64 is found locally , then it is also likely that other genotypes with phenotype @xmath64 will be local to genotypes with phenotype @xmath54 . and due to this effect being more pronounced for phenotypes with a greater frequency ( @xmath72 , c.f .
[ fig2 ] ) , we also see this effect locally with increased @xmath116 resulting in a greater @xmath105 , leading to the positive proportionality between @xmath210 and @xmath211 in the actual gp maps when compared to the null models . | mutational neighbourhoods in genotype - phenotype ( gp ) maps are widely believed to be more likely to share characteristics than expected from random chance .
such _ genetic correlations _ should strongly influence evolutionary dynamics .
we explore and quantify these intuitions by comparing three gp maps a model for rna secondary structure , the hp model for protein tertiary structure , and the polyomino model for protein quaternary structure to a simple random null model that maintains the number of genotypes mapping to each phenotype , but assigns genotypes randomly .
the mutational neighbourhood of a genotype in these gp maps is much more likely to contain genotypes mapping to the same phenotype than in the random null model .
such _
neutral correlations _ can be quantified by the robustness to mutations , which can be many orders of magnitude larger than that of the null model , and crucially , above the critical threshold for the formation of large neutral networks of mutationally connected genotypes which enhance the capacity for the exploration of phenotypic novelty .
thus neutral correlations increase evolvability .
we also study _ non - neutral correlations _ : compared to the null model , i ) if a particular ( non - neutral ) phenotype is found once in the 1-mutation neighbourhood of a genotype , then the chance of finding that phenotype multiple times in this neighbourhood is larger than expected ; ii ) if two genotypes are connected by a single neutral mutation , then their respective non - neutral 1-mutation neighbourhoods are more likely to be similar ; iii ) if a genotype maps to a folding or self - assembling phenotype , then its non - neutral neighbours are less likely to be a potentially deleterious non - folding or non - assembling phenotype .
non - neutral correlations of type i ) and ii ) reduce the rate at which new phenotypes can be found by neutral exploration , and so may diminish evolvability , while non - neutral correlations of type iii ) may instead facilitate evolutionary exploration and so increase evolvability .
* keywords * : genotype - phenotype map , neutral correlations , neutral networks , rna secondary structure , protein quaternary structure , polyomino , hp lattice model |
in the standard model , the @xmath0 matrix @xcite describes the couplings , through charged weak currents , of up - type quarks with down - type quarks .
the @xmath4 unitary matrix is determined by four parameters : three of those can be interpreted as mixing angles between the three pairs of generations , while the fourth parameter is a non trivial complex phase which is the only known source of @xmath1 violation in the standard model .
the following among the unitarity constraints of the @xmath0 matrix : @xmath5 can be used to construct a triangle in the complex plane , the so - called unitarity triangle .
one of the sides of the triangle has unitary length by construction , whereas the others have lengths : @xmath6 the angles are defined as : @xmath7 \ , , & \displaystyle \beta \equiv \arg \left [ -\frac{v_{cd}v^*_{cb}}{v_{td}v^*_{tb } } \right ] \ , , & \displaystyle \gamma \equiv \arg \left [ -\frac{v_{ud}v^*_{ub}}{v_{cd}v^*_{cb } } \right ] \ , .
\end{array}\ ] ] in the following , a few of the most recent measurements performed by the collaboration , relevant for the determination of the elements of the @xmath0 matrix , will be presented .
most of those are based on the full @xmath8 dataset available to the experiment , consisting of 467 @xmath3 pairs .
the element @xmath9 can be extracted from the measurement of the branching fraction of @xmath10 decays @xcite .
these decays are searched for on a sample where one of the two @xmath2 s is fully reconstructed in one of many hadronic final states .
the measurement of the branching fractions is performed in bins of @xmath11 , where @xmath11 is the product of the four - velocities of the @xmath2 and @xmath12 mesons .
the signal yield is extracted from a maximum likelihood fit to the missing mass squared of the unreconstructed @xmath2 candidate , which peaks at zero for signal decays ( see fig .
[ fig : sides ] ) .
the measured branching fractions are : @xmath13 where the first error is statistical and the second systematic . from this , using the calculations from unquenched lattice qcd @xcite , the value of @xmath9 is extracted : @xmath14 where the first error is statistical , the second is the experimental systematic , and the third is the error from the theory . [ cols="^,^ " , ] the @xmath15 angle can be extracted , within an 8-fold ambiguity , from an isospin analysis of @xmath16 decays , as proposed in @xcite .
this analysis was performed after obtained the result on the measurement of the branching fraction and direct @xmath1-violation of @xmath17 , based on the full dataset @xcite .
the measured branching fraction : @xmath18 _ flattens _ the isospin triangles , thus allowing the removal of some of the ambiguities with respect to the previous analysis .
the result of the isospin analysis is ( discarding the solution close to zero ) : @xmath19
at the end of the extensive experimental campaign carried on by the @xmath2-factories and belle , the @xmath0 mechanism has proven to be successful in explaining all the flavour physics phenomena .
the global fits combining all the measurement relevant for the determination of the parameters of the @xmath0 matrix show no significant discrepancy between sets of measurements @xcite .
there are some tensions at the 2@xmath20 level , for example between the measured value of @xmath21 and the one predicted using the ratio @xmath22 .
although not yet significant , the investigation of these discrepancies constitute one of the motivations for further pursuing the precision measurements on flavour physics at the hadronic colliders ( tevatron and lhc ) and at the next generation of @xmath23 colliders . | the @xmath0 paradigm has been proven to be successful in explaining the flavour structure of the standard model and the non - trivial imaginary phase of the @xmath0 matrix is the only know source of @xmath1-violation .
@xmath2-meson decays allow us to precisely determine the fundamental parameters of the @xmath0 matrix and put stringent constraints on the models of new physics .
i present some of the most recent measurements related to the @xmath0 unitarity triangle performed by the experiment , located at the slac national accelerator laboratory .
most results are based on the final dataset , consisting of 467 @xmath3 pairs . |
gallager in @xcite showed that for @xmath5 and @xmath6 , there exist @xmath7 regular low - density parity - check ( ldpc ) codes for which the bit error probability tends to zero asymptotically whenever we operate below the threshold .
richardson and urbanke in @xcite derived the capacity of ldpc codes for various message passing algorithms and described density evolution , a deterministic algorithm to compute thresholds .
zyablov and pinsker @xcite analyzed ldpc codes under a simpler decoding algorithm known as the bit flipping algorithm and showed that almost all the codes in the regular ensemble with @xmath8 can correct a constant fraction of worst case errors .
sipser and spielman in @xcite used expander graph arguments to analyze bit flipping algorithm .
burshtein and miller in @xcite applied expander based arguments to show that message passing algorithms can also correct a fixed fraction of worst case errors when the degree of each variable node is at least five .
@xcite showed that linear programming decoder @xcite is also capable of correcting a fraction of errors .
recently , burshtein in @xcite showed that regular codes with variable nodes of degree four are capable of correcting a linear number of errors under bit flipping algorithm .
he also showed tremendous improvement in the fraction of correctable errors when the variable node degree is at least five . in this paper
, we consider the error correction capability of the ensemble @xmath9 of @xmath10 regular ldpc codes as defined in @xcite when decoded using the gallager a algorithm .
we analyze decoding failures using the notion of trapping sets and prove that a code with girth @xmath11 can not correct all @xmath12 or fewer errors . using this result , we prove that for any @xmath13 , for sufficiently large block length @xmath14 , no code in the @xmath15 ensemble can correct @xmath16 fraction of errors .
this result settles the problem of error correction capability of column - weight - three codes .
the rest of the paper is organized as follows . in section [ section2 ]
we establish the notation and describe the gallager a algorithm .
we then characterize the failures of the gallager a decoder with the help of fixed points .
we also introduce the notions of trapping sets , failure sets and critical number . in section [ section3 ]
we investigate the relation between error correction capacity and girth of the code .
we extend the results to bit flipping algorithm in section [ section4 ] and conclude in section [ section5 ] .
ldpc codes @xcite are a class of linear block codes which can be defined by sparse bipartite graphs @xcite .
let @xmath17 be a bipartite graph with two sets of nodes : @xmath14 variable nodes and @xmath18 check nodes .
the check nodes ( variable nodes ) connected to a variable node ( check node ) are referred to as its neighbors .
the degree of a node is the number of its neighbors .
this graph defines a linear block code @xmath19 of length @xmath14 and dimension at least @xmath20 in the following way : the @xmath14 variable nodes are associated to the @xmath14 coordinates of codewords .
a vector @xmath21 is a codeword if and only if for each check node , the sum of its neighbors is zero .
such a graphical representation of an ldpc code is called the tanner graph @xcite of the code .
the adjacency matrix of @xmath17 gives @xmath22 , a parity check matrix of @xmath23 .
an @xmath7 regular ldpc code has a tanner graph with @xmath14 variable nodes each of degree @xmath24 ( column weight ) and @xmath25 check nodes each of degree @xmath26 ( row weight ) .
this code has length @xmath14 and rate @xmath27 @xcite . in the rest of the paper we consider codes in the @xmath10 , @xmath28 , regular ldpc code ensemble .
note that the column weight and row weight are also referred to as left degree and right degree in literature .
it should also be noted that the tanner graph is not uniquely defined by the code and when we say the tanner graph of an ldpc code , we only mean one possible graphical representation .
the girth @xmath29 is the length of the shortest cycle in @xmath17 . in this paper
, @xmath30 represents a variable node , @xmath31 represents an even degree check node and @xmath32 represents an odd degree check node .
gallager in @xcite proposed two simple binary message passing algorithms for decoding over the binary symmetric channel ( bsc ) ; gallager a and gallager b. see @xcite for a detailed description of gallager b algorithm . for column - weight - three codes , which are the main focus of this paper , these two algorithms are the same
. every round of message passing ( iteration ) starts with sending messages from variable nodes ( first half of the iteration ) and ends by sending messages from check nodes to variable nodes ( second half of the iteration ) .
let @xmath33 , a binary @xmath14-tuple be the input to the decoder .
let @xmath34 denote the message passed by a variable node @xmath35 to its neighboring check node @xmath36 in @xmath37 iteration and @xmath38 denote the message passed by a check node @xmath36 to its neighboring variable node @xmath35 .
additionally , let @xmath39 denote the set of all messages from @xmath35 , @xmath40 denote the set of all messages from @xmath35 except to @xmath36 , @xmath41 denote the set of all messages to @xmath36 .
@xmath42 and @xmath43 are defined similarly .
the gallager a algorithm can be defined as follows .
@xmath44 at the end of each iteration , an estimate of each variable node is made based on the incoming messages and possibly the received value .
the decoded word at the end of @xmath37 iteration is denoted as @xmath45 .
the decoder is run until a valid codeword is found or a maximum number of iterations @xmath46 is reached , whichever is earlier .
the output of the decoder is either a codeword or @xmath47 . _ a note on the decision rule : _ different rules to estimate a variable node after each iteration are possible and it is likely that changing the rule after certain iterations may be beneficial .
however , the analysis of various scenarios is beyond the scope of this paper . for column - weight - three codes
only two rules are possible .
* decision rule a : if all incoming messages to a variable node from neighboring checks are equal , set the variable node to that value ; else set it to received value * decision rule b : set the value of a variable node to the majority of the incoming messages ; majority always exists since the column - weight is three we adopt decision rule a throughout this paper .
we now characterize failures of the gallager a decoder using fixed points and trapping sets @xcite .
consider an ldpc code of length @xmath14 and let @xmath48 be the binary vector which is the input to the gallager a decoder .
let @xmath49 be the support of @xmath48 .
the support of @xmath48 is defined as the set of all positions @xmath50 where @xmath51 . without loss of generality ,
we assume that the all zero codeword is sent over bsc and that the input to the decoder is the error vector .
hence , throughout this paper a message of @xmath52 is alternatively referred to as an incorrect message , a received value of @xmath52 is referred to as an initial error .
@xcite a decoder failure is said to have occurred if the output of the decoder is not equal to the transmitted codeword .
@xmath48 is called a _ fixed point _ if @xmath53 that is , the message passed from variable nodes to check nodes along the edges are the same in every iteration .
since the outgoing messages from variable nodes are same in every iteration , it follows that the incoming messages from check nodes to variable nodes are also same in every iteration and so is the estimate of a variable after each iteration . in fact , the estimate after each iteration coincides with the received value .
it is clear from above definition that if the input to the decoder is a fixed point , then the output of the decoder is the same fixed point .
@xcite let @xmath48 be a fixed point
. then @xmath49 is known as a trapping set .
a @xmath54 trapping set @xmath55 is a set of @xmath56 variable nodes whose induced subgraph has @xmath57 odd degree checks . [ thm1 ]
let @xmath19 be a code in the ensemble of @xmath10 regular ldpc codes .
let @xmath55 be a set consisting of @xmath56 variable nodes with induced subgraph @xmath58 .
let the checks in @xmath58 be partitioned into two disjoint subsets ; @xmath59 consisting of checks with odd degree and @xmath60 consisting of checks with even degree .
let @xmath61 and @xmath62 .
@xmath55 is a trapping set iff : ( a ) every variable node in @xmath58 is connected to at least two checks in @xmath60 and ( b ) no two checks of @xmath59 are connected to a variable node outside @xmath58 .
see appendix [ appendix1 ] we note that theorem [ thm1 ] is a consequence of fact 3 in @xcite .
we also remark that theorem [ thm1 ] can be extended to higher column weight codes but in this paper we restrict our attention to column - weight - three codes .
if the variable nodes corresponding to a trapping set are in error , then a decoder failure occurs .
however , not all variable nodes corresponding to trapping set need to be in error for a decoder failure to occur .
@xcite the minimal number of variable nodes that have to be initially in error for the decoder to end up in the trapping set @xmath55 will be referred to as _
critical number _ for that trapping set .
a set of variable nodes which if in error lead to a decoding failure is known as a _ failure set_. _ remarks _ 1 .
to `` end up '' in a trapping set @xmath55 means that , after a possible finite number of iterations , the decoder will be in error , on at least one variable node from @xmath63 at every iteration @xcite .
the notion of a failure set is more fundamental than a trapping set .
however , from the definition , we can not derive necessary and sufficient conditions for a set of variable nodes to form a failure set .
3 . a trapping set is a failure set .
subsets of trapping sets can be failure sets . more specifically ,
for a trapping set of size @xmath56 , there exists at least one subset of size equal to the critical number which is a failure set .
4 . the critical number of a trapping set is not fixed .
it depends on the outside connections of checks in @xmath60 .
however , the maximum value of critical number of a @xmath54 trapping set is @xmath56 .
fig.[sixcycle ] shows a subgraph induced by a set of three variable nodes @xmath64 .
if no two odd degree check nodes from @xmath65 are connected to a variable outside the subgraph , then by theorem [ thm1 ] , fig.[sixcycle ] is a @xmath66 trapping set . on the other hand ,
if two odd degree checks , say @xmath67 and @xmath68 , are connected to another variable node , say @xmath69 , the subgraph resembles fig .
[ 42trappingset ] .
assuming no other connections , fig.[42trappingset ] is a @xmath70 trapping set .
we make the following observations : 1 . the three variable nodes in a @xmath66 trapping set form a six cycle
. however , not all six cycles are @xmath66 trapping sets .
apart from the subgraph induced by variable nodes , the outside connections should be known to determine whether a given subgraph is a trapping set or not .
the @xmath70 trapping set in fig.[42trappingset ] illustrates this point .
the critical number of a @xmath66 trapping set is three .
there exist @xmath70 trapping sets with critical number three and it is highly unlikely that a @xmath70 trapping set does not contain a failure set of size three .
however , we can show by a counterexample that this is indeed possible .
a @xmath54 trapping set is not unique i.e. , two trapping sets with same @xmath56 and @xmath57 can have different underlying topological structures ( induced subgraphs ) .
so , when we talk of a trapping set , we refer to a specific topological structure . in this paper ,
the induced subgraph is assumed to be known from the context .
4 . to avoid a trapping set in a code , it is sufficient to avoid topological structures isomorphic to the subgraph induced by the trapping set . for example to avoid @xmath66 trapping sets of fig.[sixcycle ] , it is sufficient to avoid six cycles .
it is possible that a code has six cycles but no @xmath66 trapping sets . in this case
all six cycles are part of @xmath70 or other trapping sets .
[ sixcycle ] [ 42trappingset ]
burshtein and miller in @xcite applied expander based arguments to message passing algorithms .
they analyzed ensembles of irregular graphs and showed that if the degree of each variable node is at least five , then message passing algorithms can correct a fraction of errors .
codes with column weight three and four can not achieve the expansion required for these arguments .
recently , burshtein in @xcite developed a new technique to investigate the error correction capability of regular ldpc codes and showed that at sufficiently large block lengths , almost all codes with column weight four are also capable of correcting a fraction of errors under bit flipping algorithm .
for column - weight - three codes he notes that such a result can not be proved .
this is because a non negligible fraction of codes have parallel edges in their tanner graphs and such codes can not correct a single worst case error . in this paper
we prove a stronger result by showing that for any given @xmath71 , at sufficiently large block lengths @xmath14 , no code in the @xmath15 ensemble can correct all @xmath4 or fewer errors under gallager a algorithm and show that this holds for the bit flipping algorithm also .
@xcite [ lemma1 ] a code whose tanner graph has parallel edges can not correct a single worst case error .
see @xcite .
the proof is for bit flipping algorithm , but also applies to gallager a algorithm .
[ lemma2 ] let @xmath23 be an @xmath72 regular ldpc code with girth @xmath73 .
then @xmath23 has at least one failure set of size two or three
. see appendix [ appendix2 ] .
[ lemma3 ] let @xmath23 be an @xmath72 regular ldpc code with girth @xmath74 .
then @xmath23 has least one failure set of size three or four . since @xmath74
, there is at least one six cycle . without loss of generality , we assume that @xmath64 together with the three even degree checks @xmath75 and the three odd degree checks @xmath65 form a six cycle as in fig.[sixcycle ] . if no two checks from @xmath65 are connected to a variable node , then @xmath64 is a @xmath66 trapping set and hence a failure set of size three . on the contrary ,
assume that @xmath64 do not form a @xmath66 trapping set
. then there exists @xmath69 which is connected to at least two checks from @xmath65 .
if @xmath69 is connected to all the three checks , @xmath76 is a codeword of weight four and it is easy to see that @xmath64 is a failure set .
now assume that @xmath69 is connected to only two checks from @xmath65 . without loss of generality ,
let the two checks be @xmath67 and @xmath68 .
let the third check connected to @xmath69 be @xmath77 as shown in fig.[42trappingset ] .
if @xmath78 and @xmath77 are not connected to a common variable node then @xmath76 is a @xmath70 trapping set and hence a failure set of size four .
if @xmath78 and @xmath77 are connected to say @xmath79 , we have two possibilities : ( a ) the third check is @xmath80 and ( b ) the third check of @xmath79 is @xmath81 ( the third check can not be @xmath82 or @xmath83 as this would introduce a four cycle ) .
we claim that in both cases @xmath76 is a failure set .
the two cases are discussed below .
case ( a ) : let @xmath84 .
@xmath85 the messages in the second half of first iteration are , @xmath86 similar equations hold for @xmath87 . for @xmath78
we have @xmath88 similar equations hold for @xmath77 . at the end of first iteration
, we note that @xmath89 and @xmath90 receive all incorrect messages , @xmath91 and @xmath79 receive two incorrect messages and all other variable nodes receive at most one incorrect message .
we therefore have @xmath92 and @xmath93 .
the messages sent by variable nodes in the second iteration are , @xmath94 the messages passed in second half of second iteration are same as in second half of first iteration , except that @xmath95 . at the end of second iteration , we note that @xmath89 and @xmath90 receive all incorrect messages , @xmath91 and @xmath79 receive two incorrect messages and all other variable nodes receive at most one incorrect message .
the situation is same as at the end of first iteration .
the algorithm runs for _ m _ iterations and the decoder outputs @xmath96 which implies that @xmath76 is a failure set .
case ( b ) : the proof is along the same lines as for case ( a ) .
the messages for first iteration are the same .
the messages in the first half of second iteration are , @xmath97 the messages passed in second half of second iteration are same as in second half of first iteration , except that @xmath98 and @xmath99 . at the end of second iteration , @xmath100 and
@xmath79 receive two incorrect messages and all other variable nodes receive at most one incorrect message and hence @xmath101 .
the messages passed in first half of third iteration ( and therefore subsequent iterations ) are same as the messages passed in first half of second iteration .
the algorithm runs for _ m _ iterations and the decoder outputs @xmath96 which implies that @xmath76 is a failure set .
let @xmath23 be an @xmath72 regular ldpc code with girth @xmath102 .
then @xmath23 has at least one failure set of size four or five
. see appendix [ appendix2 ] .
_ remark : _ it might be possible that lemmas [ lemma2][lemma4 ] can be made stronger by further analysis , i.e. , it might be possible to show that a code with girth four has a failure set of size two , a code with girth six has failure set of size three and a code with girth eight has a failure set of size four . however , these weaker lemmas are sufficient to establish the main theorem .
[ lemma5 ] let @xmath23 be an @xmath72 regular ldpc code with girth @xmath11 .
then the set of variable nodes @xmath103 involved in the shortest cycle is a trapping set of size @xmath12 .
since @xmath23 has girth @xmath29 , there is at least one cycle of length @xmath29 . without loss of generality , assume that @xmath103 form a cycle of minimum length as shown in fig.[mincycle ] .
let the even degree checks be @xmath104 and the odd degree checks be @xmath105 .
note that each variable node is connected to two checks from @xmath60 and one check from @xmath59 and @xmath106 is connected to @xmath107 .
we claim that no two checks from @xmath59 can be connected to a common variable node .
the proof is by contradiction .
assume @xmath108 and @xmath109 ( @xmath110 are connected to a variable node @xmath111 . then @xmath112 form a cycle of length @xmath113 and @xmath114 form a cycle of length @xmath115 .
since @xmath11 , @xmath116 this implies that there is a cycle of length less than @xmath29 , which is a contradiction as the girth of the graph is @xmath29 . by theorem [ thm1 ]
, @xmath103 is a trapping set .
[ corollary1 ] for a code to correct all @xmath0 or fewer errors , it is necessary to avoid all cycles up to length @xmath1 .
we now state and prove the main theorem .
[ thm2 ] consider the standard @xmath10 regular ldpc code ensemble .
let @xmath71 .
let @xmath117 be the smallest integer satisfying @xmath118 then , for @xmath119 , no code in the @xmath120@xmath10 ensemble can correct all @xmath4 or fewer errors .
first observe that for any @xmath121 , we have @xmath122 from [ theorem c.1 @xcite ] and [ lemma c.1 @xcite ] , we have the girth @xmath29 of any code in @xmath15 is bounded by @xmath123 for @xmath121 , equations ( [ girtheq1 ] ) and ( [ girtheq2 ] ) imply that for any code in the @xmath15 ensemble , the girth is bounded by @xmath124 the result now follows from corollary [ corollary1 ] .
the bit flipping algorithm does not belong to the class of message passing algorithms .
however , the definitions from section [ section2 ] and the results from section [ section3 ] can be generalized to the parallel bit flipping algorithm @xcite . without loss of generality
we assume that the all zero codeword is sent .
we begin with a few definitions .
@xcite a variable node is said to be corrupt if it is different from its original sent value . in our case , a variable node is corrupt if it is @xmath52 .
a check node is said to be satisfied if it is connected to even number of corrupt variables and unsatisfied otherwise .
let @xmath33 be the input to the parallel bit flipping decoder .
@xmath125 is a trapping set for bit flipping algorithm if the set of corrupt variables after every iteration is @xmath125 .
[ thm3 ] let @xmath126 be a set of variable nodes satisfying the conditions of theorem [ thm1 ] .
then @xmath126 is a trapping set for the bit flipping algorithm .
let @xmath127 .
then @xmath126 is the set of corrupt variable nodes .
observe that a variable flips if it is connected to at least two unsatisfied checks .
since no variable is connected to two unsatisfied checks , the set of corrupt variable nodes is unchanged and by definition @xmath126 is a trapping set .
we note that theorem [ thm3 ] is also a consequence of fact 3 from @xcite .
a trapping set for gallager a is also a trapping set for bit flipping algorithm .
it can be shown that lemmas [ lemma1][lemma5 ] and theorem [ thm2 ] also hold for the bit flipping algorithm .
in this paper we have investigated the error correction capability of column - weight - three codes under gallager a and extended the results to bit flipping algorithm .
future work includes investigation of sufficient conditions to correct a given number of errors for column - weight - three as well as higher column weight codes .
[ [ appendix1 ] ] _ proof of theorem [ thm1 ] : _ let @xmath33 be the input to the decoder with @xmath128 . then , @xmath129 let a check node @xmath130 . then , @xmath131 let a check node @xmath132 . then , @xmath133 for any other check node @xmath36 , @xmath134 . by the conditions of the theorem , at the end of first iteration , any @xmath135 receives at least two @xmath52 s and any @xmath136 receives at most one @xmath52
so , we have @xmath137 by definition , @xmath126 is a trapping set . to see that the conditions stated are necessary observe that for a variable node to send the same messages as in the first iteration , it should receive at least two messages which coincide with the received value . @xmath32 [ [ appendix2 ] ] _ proof of lemma [ lemma2 ] : _ let @xmath138 be the variable nodes that form a four cycle with even degree checks @xmath139 and odd degree checks @xmath140 .
if @xmath83 and @xmath78 are not connected to a common variable node , then @xmath141 is a @xmath142 trapping set and hence a failure set of size two .
now assume that @xmath83 and @xmath78 are connected to a common variable node @xmath90 .
then , @xmath64 is a @xmath143 trapping set and therefore a failure set of size three .
@xmath32 _ proof of lemma [ lemma4 ] : _ let @xmath144 be the variable nodes that form an eight cycle ( see fig.[44trappingset ] ) . if no two checks from @xmath145 are connected to a common variable node , then @xmath146 is a @xmath147 trapping set and hence a failure set of size four . on the other hand ,
if @xmath146 is not a trapping set , then there must be at least one variable node which is connected to two checks from @xmath145 .
assume that @xmath67 and @xmath77 are connected to @xmath79 and the third check of @xmath79 is @xmath148 ( see fig.[53trappingset ] ) .
we claim that @xmath149 is a failure set .
let @xmath150 and @xmath151 be as defined in theorem [ thm1 ] .
[ 44trappingset ] [ 53trappingset ] * case 1 : * no two checks from @xmath152 are connected to a common variable node .
then @xmath153 is a @xmath154 trapping set and hence a failure set of size five .
* case 2 : * all the three checks in @xmath151 are connected to a common variable node , say @xmath155 .
then @xmath156 is a codeword of weight six and it is easy to see that @xmath153 is a failure set .
* case 3 : * there are variable nodes connected to two checks from @xmath151 .
there can be at most two such variable nodes ( if there are three such variable nodes , they will form a cycle of length less than or equal to six violating the condition that the graph has girth eight ) . note that if @xmath157 , the decoder has a chance of correcting only if a check node in @xmath150 receives an incorrect message from a variable node outside @xmath153 in some @xmath37 iteration .
we now prove that this is not possible .
indeed in the first iteration @xmath158 by similar arguments as in the proof for theorem [ thm1 ] , it can be seen that the only check nodes which send incorrect messages to variable nodes outside @xmath153 are @xmath159 and @xmath148 .
there are now two subcases .
* subcase 1 : * there is one variable node connected to two checks from @xmath151 .
let @xmath155 be connected to @xmath68 and @xmath80 .
it can be seen that the third check connected to @xmath155 can not belong to @xmath150 as this would violate the girth condition .
so , let the third check be @xmath160 .
in the first half of second iteration , we have @xmath161 the only check nodes which send incorrect messages to variable nodes outside @xmath153 , are @xmath162 and @xmath160 .
the variable node @xmath155 is connected to @xmath68 and @xmath80 . if @xmath148 and @xmath160 are not connected to any common variable node , we are done . on the other hand ,
let @xmath148 and @xmath160 be connected to a variable node , say @xmath163 .
the third check of @xmath163 can not be in @xmath150 .
proceeding as in the case of proof for lemma [ lemma3 ] , we can prove that @xmath153 is a failure set by observing that there can not be a variable node outside @xmath153 which sends an incorrect message to a check in @xmath150 . * subcase 2 : * there are two variable nodes connected to two checks from @xmath151 .
let @xmath68 and @xmath80 be connected to @xmath155 and @xmath68 and @xmath148 connected to @xmath163 .
proceeding as above , we can conclude that @xmath153 is a failure set .
this work is funded by nsf under grant ccf-0634969 , itr-0325979 and insic - ehdr program .
the authors would like to thank anantharaman krishnan for illustrations .
a. shokrollahi , `` an introduction to low - density parity - check codes , '' in _ theoretical aspects of computer science : advanced lectures_.1em plus 0.5em minus 0.4emnew york , ny , usa : springer - verlag new york , inc . , 2002 , pp .
175197 .
s. k. chilappagari , s. sankaranarayanan , and b. vasic , `` error floors of ldpc codes on the binary symmetric channel , '' in _ international conference on communications _ ,
vol . 3 , june 11 - 15 2006 , pp . 10891094 . | in this paper , we investigate the error correction capability of column - weight - three ldpc codes when decoded using the gallager a algorithm .
we prove that the necessary condition for a code to correct @xmath0 errors is to avoid cycles of length up to @xmath1 in its tanner graph . as a consequence of this result ,
we show that given any @xmath2 such that @xmath3 , no code in the ensemble of column - weight - three codes can correct all @xmath4 or fewer errors .
we extend these results to the bit flipping algorithm . * index terms * low - density parity - check codes , gallager a algorithm , trapping sets , error correction capability |
let @xmath0 be a field , @xmath1 be a finite group and let @xmath2 be a fully @xmath1-graded @xmath0-algebra . by definition , @xmath2 has a decomposition @xmath3 where each @xmath4 is a a @xmath0-vector space , for any @xmath5 , such that for all @xmath6 we have @xmath7 in this paper we work only with symmetric fully @xmath1-graded algebras @xmath2 with the property that for any subgroup @xmath8 of @xmath1 we have that @xmath9 is a parabolic subalgebra of @xmath2 , for further details see ( * ? ? ?
* definition 2.3 , definition 5.1 ) .
we begin by giving an example of a group - graded algebra which lies in the above mentioned class of symmetric algebras and which includes the group algebra case .
[ coconet - todea - ex1 ] let @xmath2 be a fully @xmath1-graded algebra such that @xmath10 is a full matrix ring . by ( * ?
* theorem 4 ) we know that @xmath2 is a symmetric @xmath0-algebra . moreover by ( * ? ? ?
* lemma 1.1 ) we obtain that @xmath2 is a crossed product .
furthermore , these statements assure us that for any subgroup @xmath8 of @xmath1 , @xmath9 is a parabolic subalgebra of @xmath2 .
we use the language of transfer maps between hochschild cohomology algebras of symmetric algebras , defined originally in @xcite , to give a structure of mackey functor for hochschild cohomology of such fully @xmath1-graded algebras .
recall from @xcite that if @xmath11 is an @xmath12-bimodule , projective as a left @xmath13-module and as a right @xmath14-module ( where @xmath15 are two symmetric @xmath0-algebras ) there is a graded @xmath0-linear map , called transfer map @xmath16 these transfer maps are also analyzed by explicit definitions in @xcite .
let @xmath8 be a subgroup of @xmath1 and let @xmath5 .
we denote by @xmath17 the @xmath18-bimodule structure on @xmath2 given by multiplication in @xmath2 ; similarly @xmath19 as an @xmath20-bimodule .
also we consider @xmath21}$ ] to be the @xmath22}-r_h$]-bimodule given by multiplication in @xmath2 . by lemma [ coconet - todea - lema ] , a ) ( which is proved in the next section ) we can define @xmath23 hence we have the following graded @xmath0-linear maps @xmath24 @xmath25 @xmath26});\ ] ] which can be viewed as : `` restriction '' , `` transfer '' and `` conjugation '' maps . in the main theorem of this short note
we verify that the quadruple @xmath27 is a mackey functor , see @xcite .
[ coconet - todea - thmmain ] let @xmath28 and @xmath6 .
the following statements hold . * @xmath29 * @xmath30 * @xmath31 * @xmath32 if @xmath33 ; * @xmath34 and @xmath35 * @xmath36}t_{k\cap { ^g}h}^k\circ r^{{}^gh}_{k\cap { } ^gh}\circ c_{g , h,~~}$ ] where @xmath37 $ ] is a system of representatives of double cosets @xmath38 with @xmath5 .
in order to prove the main result , we need the following lemma . although the results from the next lemma are easily checkable , for completeness we give the entire proof .
we are inspired by the methods used in ( * ? ? ?
* lemma 2.1 ) .
* @xmath11 , @xmath40 and @xmath41 are projective as left and as right modules .
* @xmath42}}r_{[gh]}\cong r_{[kgh]}$ ] as @xmath43-bimodules . * @xmath44}\otimes_{r_{[^hh]}}r_{[hh]}\cong r_{[ghh]}$ ] as @xmath45}-r_h$]-bimodules . for the proof of @xmath46
we easily see that the first two bimodules are projective left , respectively right modules since @xmath9 is a parabolic subalgebra of @xmath2 , see ( * ? ? ?
* definition 5.1 , ( pa2 ) , ( pa1 ) ) .
since @xmath47}$ ] is a direct summand of @xmath2 as a right @xmath9-module , and @xmath2 is a projective right @xmath9-module ( we use again the argument that @xmath9 is parabolic ) we obtain that @xmath47}$ ] is a projective right @xmath9-module .
similar arguments show that @xmath47}$ ] is a projective left @xmath22}$]-module .
we prove assertion @xmath48 the map @xmath49}\otimes_{r_{[{}^hh]}}r_{[hh]}\rightarrow r_{[ghh]}\ ] ] defined by @xmath50 for any @xmath51}\otimes_{r_{[{}^hh]}}r_{[hh]}$ ] is a well - defined @xmath45}-r_h$]-bimodule homomorphism .
the inverse of this map is @xmath52}\rightarrow r_{[g(^hh)]}\otimes_{r_{[^hh]}}r_{[hh]},\ ] ] given by @xmath53 for any @xmath54}$ ] , where @xmath55 and all @xmath56 @xmath57 since @xmath58 ; here @xmath59 is a finite set of indices .
consider another decomposition @xmath60 where @xmath61 @xmath62 for any @xmath63 and @xmath64 is a finite set of indices .
then the equalities @xmath65 show that the last mentioned map is well - defined . at last , for the assertion
b ) we fix @xmath66 and @xmath67 and we observe that @xmath68 as @xmath69-bimodules . indeed , this can similarly be shown as statement @xmath70 the map given by multiplication has an inverse that is defined using the relation @xmath55 where for all @xmath71 we have @xmath72 and @xmath73 one can observe that , by linearity , the multiplication map extends to a @xmath43-bimodule isomorphism between @xmath42}}r_{[gh]}$ ] and @xmath74}.$ ] \ii ) and iv ) are an easy exercise if we use ( * ? ? ?
* proposition 2.7 ( 4 ) ) .
statement v ) is similar to iii ) and is left for the reader .
also , the second part of ( i ) is analogous to the first part and is left as an exercise .
the rest of the proof is a consequence of ( * ? ? ?
* proposition 2.7 ( 1 ) ) and of the bimodule isomorphisms from lemma [ coconet - todea - lema ] .
let @xmath75 viewed as an @xmath43-bimodule , @xmath76}$ ] viewed as an @xmath45}-r_h$]-bimodule , @xmath77}$ ] viewed as an @xmath45}-r_{[^hh]}$]-bimodule and @xmath78}$ ] viewed as an @xmath79}-r_h$]-bimodule .
for ( i ) we have @xmath80 for iii ) we have @xmath81}}u};\end{aligned}\ ] ] from lemma [ coconet - todea - lema ] , c ) we know that @xmath82}}u$ ] as @xmath45}-r_h$]-bimodules , hence iii ) .
consider the @xmath83-bimodule @xmath84 , the @xmath43-bimodule @xmath85 , the @xmath86}$]-bimodule @xmath87 and the @xmath43-bimodule @xmath88}$ ] . for vi )
we have @xmath89 since @xmath90 . by (
* proposition 2.11 ( i ) , 2.12 ( iii ) ) and since @xmath91}}p\cong d$ ] as @xmath43-bimodules ( see lemma [ coconet - todea - lema ] , b ) ) the following equalities hold @xmath92}t_{k\cap^gh}^k\circ r_{k\cap^gh}^{^gh}\circ c_{g , h } & = \sum_{g\in[k\backslash g / h]}t_{c\otimes_{r_{[k\cap^gh]}}r_{[{}^gh]}\otimes_{r[{}^gh]}p}\\&=\sum_{g\in[k\backslash g / h]}t_{c\otimes_{r_{[k\cap^gh]}}p}\\&=\sum_{g\in[k\backslash g / h]}t_{d}\\&=t_{b}.\end{aligned}\ ] ] 99 m. brou , _ higman s criterion revisited _ , michigan math .
j. volume , 58(1 ) 2009 , 125179 .
p. r. boisen , _ the representation theory of fully group - graded algebras _ , j. algebra 151 ( 1992 ) , 160179 .
m. linckelmann , _ transfer in hochschild cohomology of blocks of finite groups _ ,
theory 2 ( 1999 ) , 107135 .
m. e. harris , _ a note on symmetrizing form of fully group - graded algebras _ , j. alg . 237 ( 2001 ) , 789797 . s. knig , y. liu , g. zhou , _
transfer maps in hochschild ( co)homology and applications to stable and derived invariants and to the auslander - reiten conjecture _ , trans .
( 2012 ) , 195232 .
p. schmidt , _ clifford theory for simple modules _ , j. alg .
119 ( 1998 ) , 185212 . | we prove that hochschild cohomology of a certain class of fully group - graded algebras is a mackey functor .
we use the machinery of transfer maps between the hochschild cohomology of symmetric algebras .
hochschild cohomology , group , graded algebra , mackey functor 20c20 |
it is known that dynamic vortex shedding can lead to losses in lift , sharp increases in drag , and destructive pitching moments and buffeting , which all limit an aircraft flight envelope .
therefore , in order to improve aerodynamic characteristics , flow separation control would be highly desirable . the classical approach
an open - loop control achieved either by mechanical or fluidic actuation has demonstrated robustness , but its efficiency is still far from optimal .
this standard control scheme is based on actuator operating schedules , which are usually constructed using extensive and costly experimental studies .
alternatively , from a theoretical point of view , should one be able to construct an accurate solution of the navier - stokes description ( nses ) for a given airfoil shape and flow conditions , it would suggest control strategies .
however , in view of the impossibility of solving the nses in real time and in view of noisy and unpredictable real conditions , this approach is difficult to implement . at the same time , in reality one can use sensors on the boundary of lifting surfaces , which in turn read off a certain amount of extra information from the physical system and therefore should allow one to weaken the requirements on the accuracy of theoretical prediction of the flow behavior .
thus , one is naturally led to consider _ coarse _ models .
however , it should be kept in mind that the dynamical behavior of the original and coarsened ( reduced ) system will never be identical , and thus one has to decide , based on the application objectives , which aspects of the dynamics should be modelled accurately . in this work
we identify the crucial elements of the dynamics of separation bubble , namely _ bifurcation _ and _ hysteresis _ , which need to be reflected in the model and thus result in the applicability of the model to a wide range of physical parameters .
this procedure is targeted to produce a model , upon which an observer in a closed - loop control scheme can be based .
being more efficient and reliable @xcite , feedback control also naturally allows one to address the optimization issue .
while the above is a transparent justification to appeal to coarse modeling , the main challenge is that the resulting model should be both _ low - dimensional _ , for real time computational efficiency , and _
physically motivated _ , in order to reflect the actual behavior for a wide range of flight and control parameters .
since separation phenomena are clearly nonlinear , the model should also be _
these are certainly _ key requirements _ on a model .
while these key _ requirements _ are readily appreciated , the methods available to formulate such models are very limited and the connection of known models to physics is rather far from what is desired .
a commonly used approach is to first generate experimental data and , then , to extract the model by a projection onto proper orthogonal decomposition ( pod ) modes using , for example , balanced truncation or similar methods , is not reliable in view of the open flow nature of the problem and the wide range of governing parameters .
because of this , we shall make use of _ phenomenological _ modeling , which has been successful in many other problems , such as the use of duffing s equation for the buckling of elastic beams @xcite , simple maps to describe a dripping faucet @xcite , which even captures the observed chaotic behavior to a great extent , and bubble dynamics in time periodic straining flows @xcite , to name a few .
the phenomenological approach was already used in the construction of the first few models for separation phenomena , _
e.g. _ @xcite and the onera model @xcite , after the recent understanding of the importance of low - dimensional models for controlling separation .
the state - of - the - art low dimensional model used in a closed - loop control of the dynamic stall with pulsed vortex generator jets is due to magill _
its key feature is a choice of the governing physical parameters , such as lift @xmath0 and separation state @xmath1 with @xmath2 corresponding to fully attached flow and @xmath3 to fully separated flow .
steady states , @xmath4 and @xmath5 , represent the baseline case and the measured steady lift , respectively , as functions of the angle of attack @xmath6 .
the experimentally measured function @xmath5 which may contain a hysteretic behavior and thus is an empirical way of accounting for a hysteresis , as suggested by magill _
_ @xcite . exploiting the physical arguments :
( i ) lift @xmath0 @xmath7 circulation @xmath8 ; ( ii ) relaxation to a baseline state @xmath9 ; ( iii ) rise in lift when a dynamic vortex is shed @xmath10 , one arrives at the simplest low - order model with adjustable parameters , [ model_magill ] @xmath11 , \\
z_{t } & = k_{3 } b_{tt } + k_{4 } \left[z_{s}(\alpha)-z\right ] + \gamma_{\alpha } \alpha_{t}.\end{aligned}\ ] ] it should be stressed that this and many earlier attempts to develop dynamical models are based on the anzatz that this nonlinear phenomenon behaves linearly for small variations of the parameters involved @xcite , which clearly has many limitations , in particular can not account for bifurcations and hysteresis .
thus , only with an alternative approach the subject of this work can one construct a model that meets the above _ requirements_. as it will be clear from the text later , while we appeal to phenomenological analysis of empirical facts , we provide the dynamical systems grounds for it .
a symbiosis of these two methodologies yield a complete picture of the phenomena . a central notion and object ,
whose dynamics we study , is _ a separation bubble _ , whose main features are as follows .
first of all , separation of the boundary layer develops due to an adverse pressure gradient @xcite which occurs when the angle of attack of an airfoil is sufficiently large , cf .
figure [ separated_flow ] , and may be followed by re - attachment as in figure [ reattached_flow ] , thus forming a typical flow around an airfoil .
the region encompassed by the boundary layer is termed a _ a separation bubble _ after the work of jones @xcite and , as shown in figure [ notion_bubble ] , it can be closed or open .
classification of separation bubbles concerns their laminar or turbulent nature , but topologically they do not differ and thus we will not be distinguishing between various cases , but rather treat a _
generic _ case .
it should be noted that , in certain physical situations , a bubble needs to be understood in a time - averaged sense @xcite .
given the notion of a separation bubble , our dynamical systems model will aim to capture its characteristics , which are important for controlling separation phenomena .
a central idea of this work is to approach the modeling of separation bubble phenomena by identifying the key crucial elements of the bubble dynamics , namely bifurcations and hysteresis , in the appropriate portion of parameter space , as sketched in figure [ parametric_space ] . in this figure
we show the minimal dimension of the parameter space , defined by the bubble size @xmath12 , the angle of attack @xmath6 , and the actuation amplitude @xmath13 ; that is , we will be looking for the minimal model determined by the dependence of the bubble size on the angle of attack and actuation amplitude . , the angle of attack @xmath6 , and the actuation amplitude @xmath13 .
] this minimal approach is motivated by the fact that while generally there are other parameters involved , such as the reynolds number @xmath14 , the critical angle of attack @xmath15 , and the airfoil thickness @xmath16 , the resulting model will still have wide applicability .
this can be understood based on the aerodynamic properties of airfoils . to explain this
, we draw critical curves , i.e. , when separation takes place depending upon @xmath14 , @xmath15 , and @xmath16 in figure [ thick_airfoils ] .
as illustrated by figure [ re_vs_h ] , in the case of real airfoils , separation occurs at finite reynolds numbers even at zero critical angle of attack ; the higher @xmath15 the lower the critical reynolds number @xmath17 ; also , the thicker the airfoil , the lower @xmath17 . figure [ alpha_vs_h ] demonstrates the fact that the thinner an airfoil , the larger the critical angle of attack is required to achieve separation at a given reynolds number @xmath18 .
finally , in the @xmath15-@xmath14 plane in figure [ alpha_vs_re ] one can observe that for fixed airfoil thickness @xmath19 separation can occur at zero @xmath15 , which requires high enough reynolds numbers .
since in reality the reynolds numbers are huge ( e.g. for real aircraft @xmath14 varies between @xmath20 and @xmath21 ) , one concludes that limiting ourselves to `` thick airfoils '' , which can , in fact , be regarded as real airfoils since they have to carry structural load and fuel , is not a serious restriction in this first step towards low - dimensional modeling of separation phenomena . to achieve the above objectives of our modeling identified above , we will appeal to the tools of the bifurcation and catastrophe theory @xcite , as will be made precise in [ sec : bifurcation_bubble ] .
the outline of the paper is as follows . in [ sec : bifurcation_bubble ] , we discuss the first nonlinear aspect of separation bubbles , namely bifurcation phenomenon and the way to model it . in [ sec : hysteresis_bubble ] , we explore the basic physics of hysteresis phenomena , and suggest its mathematical model and how to construct a single model capable of capturing both bifurcation and hysteresis .
as was noted in the introduction , bubbles can be either in a closed or open state .
this allows us to introduce a key element of the low - dimensional modeling , namely it must capture this _
basic bifurcation _ from an _ open _ to a _ closed _ state , as shown schematically in figure [ basic_bifurcation ] , which is also known as _ bursting _ @xcite .
notably , the fact that this is the primary bifurcation was realized just recently @xcite . the vast literature on separation bubbles behavior is still at a descriptive level and suggests that one separated flow is not like any other . here
we take a different point of view , i.e. we treat the coarse behavior of separation bubbles as ( generic ) phenomena that can be modelled by a single low dimensional dynamical system . to quantify the behavior of a separation bubble ,
consider the coordinate @xmath12 , measuring the distance along the airfoil from the bubble onset to the bubble reattachment , as shown in figure [ basic_bifurcation ] .
the bubble dynamics in the first approximation can be described by two parameters : the location of separation , @xmath22 , and of reattachment , @xmath23 , which can move under the change of flight and control parameters . in some cases , e.g. the glauert glas ii airfoils , the separation point @xmath22 remains fixed for all practical purposes .
therefore , we will start by considering only the behavior of the reattachment point , which experiences a primary bifurcation in the above sense ; extending the model to include variation of @xmath22 will require the addition of a reliable separation criterion .
as an alternative to @xmath23 , one could also utilize the bubble area . from now on we will use @xmath12 as a variable representing the bubble state . the mechanism by which the excitation affects the flow lies in the generation of instabilities , and thus of large coherent structures transferring high momentum fluid towards the surface , and therefore leading to reattachment , as indicated in figure [ mechanism_actuation ] .
since actuation exploits the instabilities of the shear layer @xcite , the response to actuation depends on both @xmath13 and @xmath24 and therefore is nonlinear .
the latter again indicates , now from the point of view of actuation control mechanisms , that the low - dimensional model must be nonlinear .
as follows from experiments , the critical phenomena are as sketched in figure [ criticality ] , where the shaded region corresponds to a reattached flow ( that is , a closed bubble ) and the arrow indicates a change in location of the transition curve with an increase in @xmath6 .
the size of the bubble , @xmath12 , has a specific dependence on the amplitude @xmath13 and frequency @xmath24 of actuation , _
i.e. _ @xmath25 , when moving away from the origin @xmath26 in figure [ criticality ] . in this work
we focus on the case of time - invariant actuation , @xmath27 , although the time - varying case will be commented on later in this section .
finally , it is notable that the criticality and hysteresis phenomena depend on the connectedness of the flow domain : the bubble experiences bifurcation only in the case of flow around an airfoil , as in figure [ airfoil_model ] , while in the case of a hump model in figure [ hump_model ] , which is frequently used in experiments , there is no bifurcation .
thus , there are two basic configurations in which the behavior of the separation bubble differs : the _ hump model _ and the _ airfoil model_. namely , in the hump case @xmath28 is smooth , while in the case of an airfoil @xmath28 is discontinuous .
also , as will be important in 3 , the hysteresis phenomena are present only in the airfoil case . here , in view of its practical importance , we naturally focus on the airfoil case . in developing a model , we are guided by the principle of a minimal complexity together with the physical requirements one has to meet . at the methodological level , there are two basic ways to account for the form of @xmath28 , which has both the saturation and criticality shown in figure [ experimental_configurations ] : ( a ) to design @xmath29 as an _ algebraic _ relation , or ( b ) to introduce a _ dynamic _ description @xmath30 .
the latter approach is better suited for dynamics and control purposes , because in the case of active feedback control one would need to deal with a few characteristic times and transient effects , and thus the model should be time - dependent .
the simplest possible way of introducing time - dependent dynamics is a second - order oscillator model , @xmath31 , where @xmath32 is a damping parameter .
the justification for the latter may serve the fact that both separation and reattachment points may oscillate @xcite .
in what follows , we first formulate mathematical requirements on a model in [ subsubsection : math_req ] , then by appealing to the ideas of a potential function in [ subsubsection : pot_fun ] and a dynamic bifurcation in [ subsubsection : dyn_bif ] , we construct the model in [ subsubsection : model ] . naturally , the bubble size @xmath12 also depends on a _ flight parameter _ , in our case the angle of attack @xmath6 , which needs to be incorporated in the model ; thus , @xmath33 .
since we want to minimize the functional complexity , but to retain the nonlinear features of the phenomena , the simplest form is a quadratic nonlinearity , @xmath34 , which possesses a takens - bogdanov bifurcation , as shown in figure [ takens_bogdanov_bifurcation ] , when @xmath35 changes sign .
indeed , equilibria points are given by @xmath36 , so that @xmath37 can be represented as @xmath38 .
the eigenvalues of the linearization around @xmath39 are given by @xmath40 , while the eigenvalues of the linearization around @xmath41 are @xmath42 .
thus when @xmath43 changes sign , one observes the transition from the picture in figure [ open ] to the one in figure [ closed ] .
the requirements on the parameters in this model are dictated by the physics : ( a ) : : the stability of equilibria points should obey @xmath44 where stability implies that one equilibrium is stable ( @xmath45 is imaginary ) , and the second one is unstable ( @xmath45 is real ) .
the above inequalities indicate that the physical behavior of the model is also governed by the critical angle of attack @xmath15 , when the flow separates at @xmath46 , and the critical control amplitude @xmath47 , when flow reattaches at @xmath6 fixed .
( b ) : : the critical actuation amplitude @xmath48 should grow with @xmath49 , since the higher the angle of attack , the more control input is required to make the flow reattached .
( c ) : : the bubble size @xmath12 , which is a stable equilibrium , should shrink , @xmath50 as @xmath13 increases . at the same time , the domain of attraction of this equilibrium should shrink too , so that the bubble becomes susceptible to the finite - amplitude instabilities , as it is known from experiments , cf .
the upper part of figure [ criticality ] . in order to get better insight in the model construction ,
let us assign a potential function @xmath51 such that @xmath52 : @xmath53 which is physically determined by the elastic properties of a bubble and its interaction with the outer flow .
then we can observe that a finite bubble corresponds to @xmath51 as in figure [ potential_stable_bifurcation ] , and an infinite bubble corresponds to figure [ potential_unstable_bifurcation ] . without loss of generality
, we can assume that @xmath54 .
considering @xmath13 , as a control parameter , the requirements on the coefficients in @xmath51 are such that the equilibria , @xmath55 , obey @xmath56 the stability conditions can also be reformulated in terms of eigenvalues , as indicated in figure [ bifurcation ] . in this particular case ,
the equilibria points @xmath57 are easily computable : @xmath58 and @xmath59 . the stability criterion ( second variation ) for these
equilibria is given by the sign of the second derivative , @xmath60 , which at the equilibria points assumes the values @xmath61 and @xmath62 , respectively .
besides the stability conditions , one needs to impose @xmath63 , since the bubble shrinks when the control amplitude increases . thus
, the bifurcation from the state in figure [ potential_stable_bifurcation ] to the one in figure [ potential_unstable_bifurcation ] is obviously associated with the condition when @xmath64 .
as one can further infer , in the space of curves in @xmath65 there is an infinite number of solutions @xmath66 , @xmath67 . in practice ,
a _ systematic procedure _ would be as follows : depending on the particularities of the experimental data , one expands @xmath62 and @xmath68 in terms of some basis functions of @xmath13 , @xmath48 , etc . , and then determines coefficients in that expansion through the calibration procedure .
-plane : on the dynamic bifurcation ; solid black line represents stable equilibria , solid red line is a dynamic bifurcation when bubble grows indefinitely with time .
phase portraits in rectangles correspond to the ones in figure [ takens_bogdanov_bifurcation ] . ]
the transition from one potential to another is controlled by a bifurcation parameter , such as angle of attack @xmath6 or actuation amplitude @xmath13 .
in fact , the latter two parameters are interchangeable to a certain extent as argued in @xcite , since a change in @xmath6 or in @xmath13 leads to a change in circulation around an airfoil , and thus to a change in the flow structure .
apparently , this transition of @xmath12 from finite to infinite is _ dynamic _ in a sense that the bubble becomes infinite in figure [ potential_unstable_bifurcation ] as time @xmath69 .
this _ dynamic bifurcation _ can be clarified using phase portraits in figure [ bifurcation ] , and should be opposed to the standard static bifurcation , which is of algebraic nature as resulting from the condition of vanishing vector fields . as one can learn from figure [ bifurcation ] , at the critical value of the actuation amplitude @xmath48 both equilibria coincide and are unstable ( marginally stable ) , so that the bubble grows with time and becomes unbounded for @xmath69 . for @xmath70
there are two equilibria points , one is unstable and one is stable .
the latter corresponds to the situation when bubble is of finite size , and this state has a finite domain of attraction .
note that the potential energy shape , as in figure [ potential_bifurcation ] , is crucial in allowing the `` dynamic '' bifurcation : a @xmath71-shaped potential function apparently would not allow this type of bifurcation , as well as the domain of attraction would be modelled inconsistently with physics .
similar type of argument will be applied to the hysteresis phenomena in [ sec : hysteresis_bubble ] . in conclusion ,
having identified , based on the physical argumentation , that the potential should be of the shape as in figure [ potential_bifurcation ] in order to allow a dynamic bifurcation , the problem reduces to determination of the coefficients in .
this general procedure is the subject of the catastrophe theory @xcite and , at the technical level , is in the realm of calculus @xcite . for simplicity , restricting ourselves to the case of thick airfoils when separation occurs at @xmath72 without actuation , with one of infinitely many admissible choices of @xmath62 and @xmath68 we get : @xmath73 here @xmath74 represents a _ nonlinear response _ of the separation region to actuator excitations @xmath13 .
the product @xmath75 implies that the effect of actuation depends upon the bubble size @xmath12 .
as required , @xmath76 and the bubble shrinks as @xmath77 for @xmath78 . while this is the simplest possibility , from the above description it is clear that there is enough flexibility to calibrate the model through the fitting functions , @xmath79 , and parameters , @xmath80 within the given above bounds . by construction , the model reflects the basic generic dynamic behavior of separation bubbles . in
the _ conservative _ time - invariant case the parameter space is just @xmath81 .
when control is absent , @xmath82 , the bubble is open , which corresponds to an unstable phase portrait in figure [ open ] , that is any initial conditions lead to an unbounded bubble size @xmath12 .
when sufficient control is applied ( consider first @xmath6 fixed ) , the bubble closes , which is reflected in the change of the phase portrait as shown in figure [ closed ] . in this case
there are two equilibrium points , one is a saddle , which is unstable and thus not physically observable , and another one is a stable center .
therefore , there exists a non - zero basin of attraction which leads to a finite bubble size , @xmath83 .
figure [ closed ] also suggests that the system is susceptible to finite - amplitude instability for @xmath70 , the fact which is conceivable but has never been studied in experiments systematically .
nevertheless , it is known empirically that the bubble opens if the actuation amplitude becomes large enough , as in figure [ criticality ] ; see also @xcite .
also , the fact that the boundary layer is susceptible to finite - amplitude instabilities @xcite suggests that the separation bubble formed out of it may also be finite - amplitude unstable .
the inclusion of dissipation in the model does not change the nature of the phase portrait ; however , it does change the basin of attraction .
finally , the inclusion of time - varying effects in the control , @xmath84 with @xmath85 , also demonstrates that the bubble transforms from an open to a closed state .
thus , as required , the model captures the primary bifurcation and dynamic behavior of the separation bubble , except for the hysteresis . in the rest of this paper
we will explain how the model can be _ enhanced _ to account for the hysteresis shown in figure [ conjecture ] .
while the model is given for one of infinitely many choices of parameters , it is clear that a variety of other admissible choices can produce the same type of bifurcation and dynamics .
this freedom to choose parameters is important , however , in order to fit the model to a particular application via calibration .
it is notable that a model of a similar form was deduced _
ad hoc _ for a real bubble deforming in a straining flow studied by kang & leal @xcite , shown in figure [ fourroll ] , which experiences a bifurcation from a deformed but closed state to an open tip - streaming state , when bubble forms pointed open ends emitting tiny bubbles .
namely , the model is @xmath86 , where @xmath13 is the control parameter ( weber number ) .
the dynamics of this problem is illustrated in figure [ codim2_bifurcation ] for the conservative case , @xmath87 , and reflects the fact that for the same @xmath13 there are two equilibria , one of which is a stable center and another one is a saddle ; the latter is not observed physically in view of its unstable character .
this problem also illustrates the analogy of the dynamics of real and separation bubbles .
the basic effects of time - varying control were discussed in [ subsec : bubble_bifurcation_nature ] and reflected in figure [ criticality ] . however , the effect of changing amplitude and frequency is not trivial in view of the presence of a hysteresis @xcite in all variables @xmath88 , as illustrated in figure [ conjecture ] for the dependence of the bubble size on the actuation amplitude , @xmath28 .
experiments demonstrate that hysteresis is present no matter how slowly the actuation amplitude @xmath13 is changed , which suggests that the model should depend only on the sign of the rate @xmath89 . .
] therefore , the model should reflect the fact that there are two stable steady state solutions for the range of the control parameter @xmath90 , as in figure [ conjecture ] , which is an experimental fact .
mathematically , this means that the selection between these two solutions is due to the placement of initial conditions in the corresponding domain of attraction .
also , for @xmath91 , there should be only on stable solution , while for @xmath92 the bubble should `` bifurcate '' to infinity in a dynamical manner as described in
[ subsec : bubble_bifurcation_model ] .
the challenge of modeling the hysteresis comes from the fact that the behavior of the bubble is known only from experimental observations , while there are no analytical results .
clearly , the domain of attraction of stable solutions is not well - characterized from the existing empirical data .
therefore , in order to model hysteresis , one first needs to understand its physical origin , which is addressed below , in [ subsec : hysteresis_physics ] , where we suggest the physical mechanisms of the hysteresis .
this together with the dynamical systems and catastrophe theory allow us to modify the model to account for hysteresis , which is the subject of [ subsec : hysteresis_model ] .
physically , the separation bubble is caused by a strong adverse pressure gradient , which makes the boundary layer separate from the curved airfoil surface .
actuation with @xmath93 effectively reduces the adverse pressure gradient , i.e. , the larger the angle of attack @xmath6 , the stronger the adverse pressure gradient : this `` interchangeability '' of the effects of the actuation amplitude and the angle of attack is well - known @xcite and is reflected in the dependence @xmath47 . ] and makes the bubble closed , as in figure [ closed_bubble ] .
this can be seen from bernoulli s equation , since the velocity drop is related to the pressure rise , @xmath94 , where @xmath95 is a dynamic pressure , and @xmath96 is the fluid pressure at rest . from bernoulli s equation and figure [ basic_bifurcation ]
we can conclude that pressure rise @xmath97 and bubble length @xmath98 correlate @xmath99 and @xmath100 , respectively .
for the current purposes we neglect by the second order effects of vorticity and thus assume constant pressure inside the bubble , @xmath101 .
let us compare the above behavior of a separation bubble with that of a real static bubble ( see figure [ contrasting ] ) , which is governed by @xmath102 , where @xmath103 is the pressure inside the bubble , @xmath96 is the pressure outside the bubble , @xmath104 is the interfacial tension , and @xmath105 is the radius of the bubble . apparently , if the pressure @xmath96 outside the bubble decreases while the pressure inside is maintained constant , the bubble shrinks .
this behavior of a real bubble , when its pressure inside is maintained constant , contrasts with that of a separation bubble , which grows if the pressure outside the bubble , @xmath95 , reduces .
the underlying physics of these two problems differs : in the first case , the phenomena are dominated by static forces , while separation phenomena are dynamic .
this suggests that the separation bubble boundary possesses elastic properties , which for the current purposes can be modelled with negative interfacial tension .
note that real ( positive ) tension tends to minimize the interfacial area , while the effective tension of the shear layer tends to maximize the bubble boundary and only the external energy input ( excitation ) counterparts this effect and makes the bubble closed : this justifies a negative sign of the tension .
alternatively , one can use a nontrivial state equation for the pressure inside the bubble , which can be measured experimentally .
the elasticity of the separation bubble is evidenced by introducing disturbances outside the bubble and observing the changes in the bubble characteristics , i.e. shape and pressure inside . with the above physical background
, we can provide a simple mechanistic model explaining the origin of the hysteresis . for simplicity
, consider the two - dimensional situation depicted in figure [ mechanical_analog ] : a hemispherical bubble having variable size with the left end fixed and with its right end free to move thus modeling a separation bubble with moving reattachment point .
the bubble size changes depending upon the free - stream velocity @xmath106 , which is chosen to be the control parameter . when @xmath106 increases and the right end of the bubble reaches the trailing edge at @xmath107 at critical @xmath108 , the pressure inside the bubble increases by a finite amount , @xmath109 , which is due to suction of a high pressure fluid from the lower side of the airfoil .
hence the bubble size increases abruptly by some amount .
conversely , when @xmath106 decreases and bubble reaches the trailing edge at @xmath107 at a different critical @xmath110 , the pressure inside the bubble relaxes to its original value , @xmath111 .
the jump in pressure at the critical point when reattachment is at the trailing edge has the following physical explanation .
it is known that the lift drops when the bubble opens , which effectively means that the pressure balance between the lower and upper surfaces of an airfoil has changed : some amount of pressure at the lower surface has leaked into the upper surface , namely into the bubble . the latter is allowed by unsteadiness of the process , i.e. , the unsteady kutta - joukowsky condition . therefore , the mechanical analog of a separation bubble is @xmath112 , so that the bubble grows when the ambient pressure dictated by bernoulli s equation , @xmath113 , decreases : [ model_mechanistic_hysteresis ] @xmath114 which produces a hysteretic behavior .
obviously , @xmath115 is consistent with the physical observations . in the light of the above
, one can account for the hysteresis in figure [ conjecture ] in model as follows .
when @xmath116 and @xmath13 passes through @xmath48 the transformation @xmath117 with @xmath118 is applied , since physically the effectiveness of control drops by @xmath119 . when @xmath120 and @xmath121 then @xmath122 , since too conservative amount of control has been applied before reaching @xmath121 .
these altogether lead to the desired hysteretic behavior .
same can be done to account for the hysteretic dependence on the angle of attack @xmath6 .
the above mechanistic model of the hysteresis captures the physics and proves that the separation bubble has a nontrivial potential function associated with it .
it should be noted that such type of discontinuous modeling of hysteresis based on the rate @xmath123 is still widely used in applications and known as _ play _ and _ stop _ ( classical prandtl model ) models , cf .
visintin @xcite .
however , for the purpose of deriving a universal model which combines both the bifurcation and the hysteresis in a dynamic manner , i.e. suitable for control purposes , it makes sense to follow another way of modeling hysteresis phenomena , based on the choice of an appropriate potential function @xmath51 , similar to what was done in [ subsec : bubble_bifurcation_model ] .
we will enforce this point of view in [ subsec : analogies_hysteresis ] , where we will illustrate the analogy to other physical phenomena .
the grounding thesis is that the true curve of states in figure [ conjecture ] is not the solid discontinuous one , but rather the `` true '' picture for separation bubbles corresponds to the smooth curve ( including the dashed line ) in figure [ conjecture ] , the fact which has not been realized in the literature before .
this smooth curve corresponds to the equilibria states of an appropriate potential function @xmath51 ; the dashed curve is not physically observable in view of instability of the corresponding equilibria states . from [ subsec : bubble_bifurcation_model ] we know that the potential function should be of special shape , i.e. when @xmath124 , then @xmath125 , that is the highest order terms in @xmath6 should be odd .
then , as a natural generalization of the picture in figure [ bifurcation ] , we arrive at figure [ hysteresis ] .
-plane and corresponding potential functions ; solid black lines represent stable equilibria , while dashed lines are unstable equilibria ; solid red line represents a dynamic bifurcation ( bubble size grows with time unboundedly ) .
plots @xmath51 in rectangles show the shape of the potential for @xmath126 , @xmath127 $ ] , @xmath91 , respectively . ] at the technical level , the lowest order potential suitable for achieving the picture in figure [ hysteresis ] is of the fifth order , so that model becomes @xmath128 with @xmath51 of the fifth order .
the existence of such potential is apparent , and its coefficients in the polynomial representation can be found with the help of linear programming given a set of inequalities and equalities based on the calibration requirements .
the fact that the hysteresis originates from the particularity of the potential function is well - known from other physical systems , e.g. a ferromagnetic drop deforming in a magnetic field @xcite and cavitating hydrofoils @xcite .
consider the _ deformation of ferrofluid drop _ of permeability @xmath129 , placed in a fluid of permeability @xmath130 , in a magnetic field @xcite .
the surface energy of the drop is given by @xmath131 , \
\epsilon = \left(1-e^{2}\right)^{1/2},\end{aligned}\ ] ] where @xmath132 is the aspect ratio , @xmath133 and @xmath62 are semi major and semi minor axes respectively , and @xmath134 is the interfacial tension .
the magnetic energy is of the form @xmath135 where @xmath136\right\}/2 \epsilon^{3}$ ] is the demagnetization factor , @xmath71 is the volume of the drop , and @xmath137 is the applied magnetic field .
minimization of the total energy , @xmath138 , with respect to the aspect ratio @xmath139 produces @xmath140 the behavior of which is depicted in figure [ ferro ] .
the bubble shape is a simple counterplay between magnetic and interfacial energy of the drop : the former tends to elongate the drop , while the latter tends to make the drop spherical . for certain values of @xmath141
there are three solutions , but not all of them are stable .
when the solution reaches point a it jumps to the point b , and similarly for the points c and d. ac portion corresponds to a maximum of @xmath142 and thus is unstable , while the rest of the curve is minima of @xmath142 and thus is stable . finally , a _ cavitation bubble _ on a _ hydrofoil _ , where hysteresis can be explained with the help of inviscid free - streamline theory @xcite , is another example , where it has been done analytically . in this physical problem
the boundary of the bubble is well - defined physically and thus the problem is reliably treated with a free - streamline theory , that is its predictions @xcite agree well with experiments @xcite .
for the theoretical treatments of cavitation flows with free - streamline theory we refer to tulin @xcite , yeung & parkinson @xcite , birkhoff & zarantonello @xcite , and on the physics of cavitation flows to wu @xcite , brennen @xcite . on the physical side of cavitation phenomena , it is known from the general equations of fluid dynamics that the pressure depends on the velocity distribution ( in the steady case ) and on the acceleration ( in the unsteady case ) . more importantly , the pressure might become negative at points where the velocity is large . in the majority of cases
, fluids can not sustain a negative pressure and the continuity of the flow breaks down . as a result , a region filled with fluid vapor is formed this is a cavitation phenomena ( see , for instance , @xcite ) . in continuous
incompressible flows the maximum velocity occurs at the boundary and hence cavitation first appears on the body surface : @xmath143 which is a cavitation number .
deviations from this law are due to vortex shedding and other unsteady effects .
the behavior of the cavitation bubble is given by for partially cavitating , @xmath144 @xcite , and supercavitating , @xmath145 @xcite , foils respectively , [ cavity_length ] @xmath146 where @xmath6 is the angle of attack .
the expressions are basically the solution of the equilibrium condition , @xmath147 , and schematically shown in figure [ cavitation_hysteresis ] for fixed cavitation number ( see also sychev @xcite ) .
this work has focused on the fundamental aspects the most important physics and dynamic behavior of a generic separation bubble using thick airfoils as a paradigm . given an incomplete experimental knowledge of the complex phenomena of separation bubble
, we applied the deduction based on bifurcation and catastrophe theory and thus ( 1 ) filled in incomplete pieces in the dynamical picture of the phenomena , ( 2 ) advocated that this dynamical picture is finite - dimensional at the coarse level , ( 3 ) developed a constructive way of building a model , and ( 4 ) produced a model .
the model can be enhanced in particular by ( a ) incorporating a non - trivial state equation of a bubble , ( b ) accounting for separation at non - zero angle of attack @xmath148 , and ( c ) calibrating the model for a given airfoil .
these are the future directions of this study and will require considerable theoretical and experimental efforts .
we also expect that this approach to low dimensional modeling will be helpful in real time flow control . | this work uncovers the low - dimensional nature the complex dynamics of actuated separated flows .
namely , motivated by the problem of model - based predictive control of separated flows , we identify the requirements on a model - based observer and the key variables and propose a prototype model in the case of thick airfoils as motivated by practical applications .
the approach in this paper differs fundamentally from the logic behind known models , which are either linear or based on pod - truncations and are unable to reflect even the crucial bifurcation and hysteresis inherent in separation phenomena .
this new look at the problem naturally leads to several important implications , such as , firstly , uncovering the physical mechanisms for hysteresis , secondly , predicting a finite amplitude instability of the bubble , and thirdly to new issues to be studied theoretically and tested experimentally .
more importantly , by employing systematic reasoning , the low - dimensional nature of these complex phenomena at the coarse level is revealed . separation bubble , separation control , low - dimensional modeling , phenomenology , catastrophe theory |
quantum chromodynamics ( qcd ) is widely believed to be the theory of strong interactions . in this non abelian theory , the gauge bosons , the gluons , carry a color charge .
we are therefore led to observable color singlet configurations made of quarks but also gluons .
glueballs , bound states of only gluons , and hybrid mesons are consequences and predictions of qcd . besides glueballs ,
hybrid mesons deserve much interest .
the gluonic excitation leads to low - lying states with quantum numbers not allowed for usual mesons ( note that the low - lying gluon states are non exotic but glueballs with exotic quantum number also exits and are called oddballs , for a review see ref .
their observation would be another great confirmation of qcd .
however , the detection of exotic hadrons remains a challenging task ( for a review see ref .
@xcite ) .
the properties of hybrid mesons were investigated in various approaches , see for examples ref .
@xcite , and in lattice qcd @xcite .
the former have two common interpretations . in the first one ,
the flux tube linking the quark to the antiquark is in an excited state , allowing quantum numbers that can not be reached by the usual quark - antiquark picture . in the other scenario
, the excitation is modelled by a constituent gluon , leading to a three - body system .
we know that both picture are closely linked : the constituent gluon creates an equivalent potential compatible with the energy of an excited string @xcite .
all these approaches rely on quasiparticle interpretation . for heavy particles ,
the spectrum and decay properties are extracted from a hamiltonian formalism .
one has then to resort to a numerical procedure to find eigenvalues and eigenfunctions . for two - body systems , like heavy charmonia , different methods exist .
the most efficient one is the so called lagrange mesh method .
this technique allows one to compute energies and wave functions straightforwardly since the method is not variational @xcite .
matrix elements for semi - relativistic kinetic energy are easily computed .
however , it does not admit a three - body generalization in the semi - relativistic case . for three - body systems , like hybrid mesons ,
one generally computes the mass within a finite dimensional basis .
we expand the unknown wave function on a set of trial functions forming a basis of the hilbert space in the limit where the dimension of the basis goes to infinity . in practice
, we only deal with a finite number of basis functions .
fortunately , we know the approximation of the mass to be always an upper of the true eigenvalue , allowing to minimize the mass with respect to parameters @xcite .
the question is then : how many basis functions should we consider to have an acceptable accuracy ? in this work , we answer to this question for the three - body description of hybrid mesons .
recently , kalashnikova and nefediev investigated the spectrum of the charmonium hybrid @xmath1 within the constituent gluon model @xcite .
they derived a hamiltonian thanks to the field correlator method and introduced einbein ( or auxiliary ) fields to deal with relativistic kinematics .
the resulting hamiltonian was diagonalized with one trial function that was taken as a gaussian depending of a hyperradius . finally , the authors computed correction coming from the string , self - energy and spin - dependent operators .
the authors studied the case of the lowest @xmath2 hybrid meson in with the gluon quantum numbers are @xmath3 and @xmath4 ( magnetic gluon ) . in this work ,
we investigate how accurate are those approximations for the wave function ( truncation of the basis ) and for the hamiltonian ( introduction of auxiliary fields ) and greatly improve the accuracy on the mass thanks to correlated gaussian functions . in sec .
[ sec : def ] , we present the model of ref.@xcite and recall its main properties . the authors used a particular ansatz for the wave wave function which is described in sec .
[ sec : hypers ] .
we discuss some improvements for the wave function in sec .
[ sec : gauss ] . in this section , which consist in the main part of this work
, we perform numerical calculations with various approximations for the model and make some comparisons .
having identified a good approximation for the wave function , we test in sec . [
sec : param ] the stability of the error under parameter evolution . finally , we draw our conclusion in sec .
[ sec : conclu ]
the starting point for the hybrid model of @xcite is a hamiltonian derived from qcd thanks to the field correlator method . for a hybrid meson seen as a @xmath6 system , this hamiltonian reads @xmath7 where @xmath8 @xmath9 the first terms of @xmath10 represent the kinetic energy of two quarks of the same mass and of the massless constituent gluon .
the long - range confining force consists in two strings , two fundamental flux tubes of energy density @xmath11 , linking the gluon to the quarks .
such a confining potential is in agreement with other phenomenological approaches @xcite as well as with recent lattice qcd computations @xcite . for other three - body systems like baryons ,
the confinement is generally assumed to be a y - shape . in the case of the hybrid , however , under casimir scaling hypothesis for the string tension , the v - configuration is resolved to be energetically favorable @xcite .
the torricelli point of the y - shape merges with the gluon position . imposing a y - shape would obviously raise the whole spectrum .
the casimir scaling hypothesis is supported by recent lattice studies , see for instance ref .
@xcite .
the confining potential is augmented by the short - range coulomb potential @xmath12 , where @xmath13 is the strong coupling constant , arising from one gluon exchange effects between the constituent particles .
in particular , the coulomb potential between the quark and the antiquark is repulsive since the @xmath14 pair is in a color octet .
we refer the reader to ref .
@xcite for a detailed discussion about the short - range interaction potentials in qcd .
spin - dependent short - range interactions could be added in perturbation as it is done in ref .
however , it is enough for our purpose to consider only the dominant order hamiltonian ( [ eq : mainh ] ) . to be complete
, we recall that the parity and charge conjugation of a @xmath6 system are given respectively by @xmath15 where @xmath16 and @xmath17 are the orbital angular momenta of the @xmath14 pair and of the gluon , and where @xmath18 is the intrinsic spin of the @xmath14 pair .
compared to usual mesons , extra phases give rise to exotic quantum numbers such as @xmath19 for example .
two cases have to be distinguished following the total spin @xmath20 of the gluon : the magnetic gluon , for which @xmath21 , and the electric gluon for which @xmath22 @xcite .
although it may gain a dynamical mass induced by confinig forces , the gluon is a massless particle that remains transverse , with only two polarisations @xcite .
therefore , the minimal value for @xmath20 is 1 .
exotic quantum numbers require at least one @xmath23-wave . in the following
, we shall refer to magnetic ( electric ) , the @xmath6 states with @xmath3 ( @xmath24 ) .
solving the eigenequation associated to a three - body semirelativistic hamiltonian such as ( [ eq : mainh ] ) is a difficult numerical problem .
the intrinsic complexity of three - body systems obviously comes into play , but another problem is the determination of the matrix elements for kinetic operators of the form @xmath25 , which are not commonly found in quantum mechanics .
this last difficulty can be avoided by introducing auxiliary ( or einbein ) fields to get rid of the square roots appearing in hamiltonian ( [ eq : mainh ] ) .
one obtains @xmath26 where @xmath27 , @xmath28 , and @xmath29 are the so - called auxiliary fields .
being formally defined as operators , they can be eliminated through the following equations @xmath30 it is then readily checked that @xmath31 ; both hamiltonians are equivalent up to the elimination of the auxiliary fields as operators . however , the calculations are considerably simplified if one considers them are c - numbers variational parameters .
the eigenvalues of the spinless hamiltonian are more easily found since only nonrelativistic kinetic operators are present .
these eigenvalues , denoted as @xmath32 , are finally minimized with respect to the einbeins .
the optimal values of the auxiliary fields seen as variational parameters are logically close to the average values of the corresponding operators @xcite .
for example , the optimal value of @xmath29 , denoted as @xmath33 , is such that @xmath34 .
it can be interpreted as a dynamical gluon mass .
the same arguments hold for the other auxiliary fields .
an important point , that has been shown in ref .
@xcite , is that the eigenvalues of hamiltonian ( [ eq : mainh_af ] ) are upper bounds of the eigenvalues of hamiltonian ( [ eq : mainh ] ) .
the more auxiliary fields are introduced , the less this bound is accurate . in particular , the accuracy of the auxiliary field method decreases when light or massless particles are present in the system under study . in this work we focus on the case where the quark and the antiquark have the same mass , _
i.e. _ @xmath35 .
then , by symmetry , @xmath36 and , using the jacobi coordinates @xmath37 hamiltonian ( [ eq : mainh_af ] ) becomes @xmath38 where @xmath39 and @xmath40 are the momenta associated to @xmath41 and @xmath42 respectively , and where @xmath43 the center of mass , defined as @xmath44 is decoupled and its conjugate momentum is set equal to zero since we work in the rest frame of the system .
eigenvalues of hamiltonian @xmath45 with auxiliary fields ( [ af3 ] ) in the case of @xmath46 systems have been numerically computed in ref .
@xcite by using the hyperspherical formalism .
we recall in this section the procedure that has been used in this last reference .
the authors of @xcite state that since the @xmath14 pair is heavy , the assumption @xmath47 can be made . moreover ,
the particular case of a magnetic gluon with @xmath3 is considered in this reference . for this particular choice
, the spin of the quark pair @xmath18 is a good quantum number and leads to the following states : @xmath48 for @xmath49 and @xmath50 for @xmath51 .
all these states are degenerate since @xmath45 does not have any spin - dependent term .
this leads to wave functions of the form @xmath52^j,\ ] ] with the hyperradius @xmath53 @xmath54 is a shorthand notation for the coupling @xmath55^\ell$ ] , with the solid spherical harmonics @xmath56 .
the following states can consequently be described within this approach @xmath57 ^ 1\\ \left|j^{-+}\right\rangle&=&\left|\chi(\bm x ) \right\rangle\otimes\left[\left|0,1\right\rangle^1\otimes \left|1\right\rangle\right]^{j},\end{aligned}\ ] ] with @xmath58 .
the trial wave function is chosen to be a gaussian depending on one variational parameter @xmath59 , _
i.e. _ @xmath60 = \exp\left(-\frac{1}{2}\beta^2\bm x^2\right){\cal y}_{1m}(\bm \rho).\ ] ] this solid harmonic , @xmath61 , determines the angular momentum in order to treat exotic mesons with a @xmath23-wave gluon and a @xmath62-wave @xmath63 pair .
the spin function for the quark pair is irrelevant since our hamiltonian is spin - independent .
we will therefore get the same mass for the the four states @xmath64 under consideration .
a convenient method to fit the parameters is to reproduce the charmonium spectrum within the same flux tube model , that is with the hamiltonian @xmath65 .
the parameters obtained in this way are compatible with typical values used in potential models @xcite : @xmath66 we , as the authors of ref .
@xcite , then implicitly assumed the same value for the parameters for both systems ( charmonium and hybrid charmonium ) .
it is also natural to keep the same parameters as in ref .
@xcite since the present paper focuses on the improvement of the numerical resolution . in sec .
[ sec : param ] we vary the parameters and investigate this influence of the one gaussian approximation . a numerical resolution of @xmath45 with the ansatz leads to @xcite @xmath67 for the ground state .
this mass corresponds in a first approximation to @xmath68 hybrid mesons with quantum numbers @xmath48 and @xmath50 .
spin - dependent corrections have also been computed to be equal to @xcite [ splittings ] @xmath69 the splittings induced by the spin - dependent operators in perturbation theory are clearly not negligible .
however , we know from the rayleigh - ritz method @xcite that the mass @xmath70 gev can only be an upper bound of the true eigenvalue of @xmath45 . indeed , when truncating the basis with a finite number of basis function , one restrict the operator to the subspace spanned by the trial functions .
it is then natural to investigate how far is @xmath71 from the true eigenvalue and compare the difference with respect to the additional corrections .
the real eigenvalue @xmath72 of a schrdinger equation @xmath73 is generally found by expanding the wave function @xmath74 in a basis @xmath75 the real eigenvalue is reached in the limit @xmath76 . since for fixed @xmath77 , @xmath78 is an upper bound , we can then minimize the mass with respect to the parameters on which depend @xmath79 . a appropriate ansatz for @xmath79 will lead to an accurate mass for low values of @xmath77 . for the trial functions
, we use a generalisation of the hyperspherical function @xmath80 the wave function used in @xcite is recovered for the particular values @xmath81 , @xmath82 and the condition on the variational parameters @xmath83 et @xmath84 .
the correlated gaussian shared many advantages @xcite .
they admit a generalisation for an arbitrary number of interacting particles thanks to matrix notation ; matrix elements for usual power laws potential are expressed in close forms ; the fourier transform of a gaussian is also a gaussian . in particular , we have shown that matrix elements for semirelativistic kinematics @xmath25 can be easily derived @xcite .
this form is very convenient for hybrid mesons since we deal easily with angular excitations for the gluon but also for the quarks .
imposing some restriction on the parameters we get two ansatz for the lowest @xmath23-wave hybrid mesons @xmath85 the function @xmath86 has the angular part @xmath87 and correspond to a hybrid meson with a magnetic gluon , and the function @xmath88 has the angular behavior @xmath89 and corresponds to a hybrid meson with an electric gluon .
let us now see how changes the mass when relaxing the condition on @xmath83 and @xmath84 . in order to have a relevant comparison point
, we consider @xmath45 with the parameters given in and .
if we use the hyperspherical formalism with only one trial function ( and hence only one variational parameter ) in our basis , we find a mass of @xmath90 gev ( obviously the same value as in ref .
the optimisation procedure leads to @xmath91 gev@xmath92 which correspond to @xmath93 gev and @xmath94 gev .
the relation with the parameters implies a nearly symmetric gaussian shape , _
i.e _ @xmath95 here we perform a numerical optimisation on the two parameters of our single wave function @xmath86 . we expect to find a mass slightly lower than in @xcite .
indeed , the resulting mass for the four states @xmath64 in this approximation reads @xmath96 we gain 109 mev by simply choosing a more general ansatz ( with two variational parameters instead of one parameter ) for the trial wave function .
the shift is of order of the spin - splittings .
let us look at the wave in order to understand this difference .
the optimisation procedure leads to @xmath97 gev and @xmath98 gev .
their ratio strongly differs from . as an illustration
, we plot the spatial part of the wave functions in figs [ fig : a : hyper ] and [ fig : a : one ] ( remind the factor @xmath99 in the solid spherical harmonic ) .
[ h ] .wave function coefficients for magnetic gluon .
[ cols="^,^,^,^",options="header " , ] [ tab : wfelecsemirelat ] in order to compare qualitatively the different wave functions , we display in fig . [
fig : lambda0 ] the wave functions for a magnetic gluon at @xmath100 and in fig .
[ fig : rho0 ] the wave functions for an electric gluon at @xmath101 . for the sake of completeness and in order to compare all the wave functions quantitatively , we add the contour lines of all the eight wave functions described in the text .
those plots are displayed in figs .
[ fig : a : hyperplan ] , [ fig : a : oneplan ] , [ fig : a : tenplan ] , [ fig : a : sr_plan ] , [ fig : b : hyperplan ] , [ fig : b : oneplan ] , [ fig : b : tenplan ] and [ fig : b : sr_plan ] .
the mass of the hybrid with an electric gluon is resolved to be lighter than the magnetic one .
our simple hamiltonian mixes the two states and the mass of the lightest states has clearly a bigger electric component .
however , the decay modes of these two states are different .
an electric gluon allows the hybrid to decay into a s - wave @xmath102 .
this may induce difficulty to single out a hybrid interpretation with respect to conventional charmonium .
i would be interesting to investigate more deeply the decay properties of those states which could guide experimentalist in the search for hybrid charmonia .
the previous sections emphasized on the fact that various approximations may lead to overestimation of hybrid masses .
however , the single gaussian approximation was resolved to be an acceptable approximation .
we now turn our attention to this specific approximation and test its evolution for different values of the parameters for the semi - relativistic hamiltonian .
we then change one parameter and keep the others to their `` optimal '' values .
the results are displayed in figs [ fig : mass_evo ] and [ fig : sigma_evo ] for respectively the evolution of the gluon mass and the string tension .
varying the gluon mass is interesting since effective approaches used a non - vanishing value for the gluon mass in the kinetic energy , see for instance ref .
@xcite . for wide ranges of the parameters ,
the single gaussian approximation is resolved always to give an overestimation of @xmath103 mev constant in the intervals . for a wide range of variation of @xmath104 , the single gaussian approximation lie around 20 - 50 mev above the true mass . for hybrid meson systems , the approximative wave function is then robust under parameter evolution .
we checked that we obtain similar results for electric gluon wave functions .
during the main part of this paper , we were concerned by kalashnikova and nefediev s model @xcite in which the starting point was hamiltonian .
this model shares the constituent model s typical features .
the kinetic energy is the semi - relativistic expression valid for massless particles ( the constituent gluon ) . a linear ( in fact a v - junction ) plus coulomb term is used for the potential .
the parameters ( charm quark mass , string tension and strong coupling ) were determined on a similar model for charmonium .
the parameters reproduce a quarkonium spectrum in agreement with experimental data and are expected to predict an acceptable value for hybrid masses .
however , solving a three - body systems may require hypothesis : on one side in the hamiltonian by introducing auxiliary fields to get of rid of square roots ; on the other side on the wave function by imposing particular values for internal quantum numbers and/or restricting the space spanned by the wave function to a restricted subspace of the total hilbert space .
we demonstrated in this work that one should be very careful with the approximations made to solve the eigenvalue problems numerically .
indeed , auxiliary fields , hyperspherical formalism with only one function , fock space reduction ( magnetic / electric gluons ) may causes an overestimation of the mass by amount of 100 - 600 mev .
however , in each cases of study ( magnetic or electric gluon ) , the less worst approximation was the only gaussian function .
for the same hamiltonian , the single gaussian overestimated the mass by only @xmath105 mev where the hyperspherical formalism with only one function gave always a discrepancy around @xmath106 mev . in their conclusion
, the authors of @xcite pointed out a numerical vector hybrid meson mass of 4.397 gev , substantially higher than the experimental candidate @xmath0 .
the improvements of the numerical method developed in the present paper clearly favor the hybrid meson interpretation of the candidate .
indeed , the real eigenvalue of the hamiltonian used in @xcite is 130 mev below the approximation with the hyperspherical formalism , the exact discrepancy with the vector x(4260 ) .
we stressed , nevertheless , that the correction to the bare mass were computed with the approximative wave function and should slightly differ .
it would be interesting to investigate how strong is the difference .
moreover , we have shown that the mass remains more or less stable under the approximations considered in this paper but we think that the decays properties , depending strongly of the wave functions , should change more that the mass .
it is also worth mentioning the two other vector states y(4325 ) from babar @xcite and y(4360 ) from belle @xcite .
their mass lie close to the x(4260 ) and mixing may cause a mass shift with respect to pure @xmath2 hybrid states . however , the purpose of this paper was the investigation of the numerical procedure for a simple hamiltonian for pure @xmath2 .
a detail investigation of the possible mixing with other states would require a more complicated description of the system and is beyond this study .
the author thanks f. buisseret and c. semay for valuable suggestions and comments about this work and the i.i.s.n . for financial support . f. buisseret and v. mathieu , eur .
phys . j. a * 29 * , 343 ( 2006 ) [ arxiv : hep - ph/0607083 ] ; f. buisseret and c. semay , phys . rev .
d * 74 * , 114018 ( 2006 ) [ arxiv : hep - ph/0610132 ] ; f. buisseret , v. mathieu , c. semay and b. silvestre - brac , eur .
phys . j. a * 32 * , 123 ( 2007 ) [ arxiv : hep - ph/0703020 ] .
f. buisseret and c. semay , phys .
e * 71 * , 026705 ( 2005 ) .
b. silvestre - brac , c. semay , and f. buisseret , j. phys .
a : math . theor . * 41 * , 275301 ( 2008 ) .
f. buisseret and v. mathieu , eur .
j. a * 29 * , 343 ( 2006 ) . | we revisited a model for charmonium hybrid meson with a magnetic gluon [ yu . s. kalashnikova and a. v. nefediev , phys .
rev .
d * 77 * , 054025 ( 2008 ) ] and improved the numerical calculations .
these improvements support the hybrid meson interpretation of @xmath0 . within the same model , we computed the hybrid meson mass with an electric gluon which is resolved to be lighter .
relativistic effects and coupling channels decreased also the mass . |
relativistic nucleus - nucleus ( a+a ) collisions provide us with experimental information about the phase diagram of quantum chromodynamics ( qcd ) and the strongly interacting matter properties . the last stage of such collisions is traditionally analyzed within the statistical approach which gives us an excellent opportunity to reveal the parameters of chemical freeze - out .
this approach is based on the assumption of the thermal equilibrium existence during the last stage of a+a reaction .
such an equilibrium can be reached due to intensive particle scattering .
the stage of the system evolution when the inelastic reactions between hadrons stop is referred to as a chemical freeze - out ( fo ) .
particle yields are determined by the parameters of fo , namely by chemical potentials and temperature .
this general picture is a basis of the hadron resonance gas model ( hrgm ) @xcite which is the most successful one in describing the hadronic yields measured in heavy - ion experiments for energies from ags to lhc . despite a significant success of the hrgm ,
in the experimental data analysis there are a few unresolved problems . in general
they are related to the description of hadron yields which contain ( anti)strange quarks .
thus , the energy dependence of @xmath4 and @xmath5 ratios remained out of high quality description for almost a decade .
excess of strange hadrons yields within the hrgm led physical community to ponder over strangeness suppression in heavy ion collisions .
the first receipt to resolve this problem was to introduce the strangeness suppression factor @xmath0 which should be fitted in order to describe the experimental data @xcite .
however , such an approach is not supported by any underlying physical model and the physical meaning of @xmath0 remains unclear @xcite .
in addition the strangeness suppression approach in its original form does not contain a hard - core repulsion between hadrons , while the latter is an important feature of the hrgm .
a significant role of the hard - core repulsion was demonstrated once more in ref .
@xcite where the global fit of hadron yield ratios was essentially improved ( to @xmath6 1.16 ) compared to all previous analyses .
the most advanced way to account for the hard - core repulsion between hadrons is to consider a hadron gas as a multi - component mixture of particles with different hard - core radii @xcite . within this approach all baryons and mesons except for the kaons and the pions
are endowed by the common hard - core radii @xmath7 and @xmath8 , respectively . at the same time
the kaon and the pion radii @xmath9 and @xmath10 are fitted independently in order to provide the best description of @xmath4 ratio @xcite .
this is an important finding since the non - monotonic energy dependence of @xmath4 ratio may indicate some qualitative changes of the system properties and may serve as a signal of the deconfinement onset .
this is a reason why such a ratio known as the strangeness horn is of a special interest .
note , that the multi - component approach substantially increased the strangeness horn description quality , without spoiling the other ratios including @xmath5 one .
however , even this advanced approach does not reproduce the topmost point of the strangeness horn indicating that the data description is still not ideal . in order to resolve this problem in ref .
@xcite the @xmath0 factor was considered as a free parameter within the hrgm with multi - component repulsion .
although the @xmath0 data fit sizably improves the quality of strangeness horn description , it does not seem to be useful for the description of other hadron multiplicities @xcite .
furthermore , in contrast to the claims established on the low - quality fit @xcite , at low energies it was found @xcite that within the error bars in heavy ion collisions there is an enhancement of strangeness , i.e. @xmath11 , and not a suppression .
the strangeness enhancement was confirmed very recently @xcite by the high quality fit of the available data within the multicomponent hrgm in which the hard - core radius of @xmath12 ( anti)hyperon was considered as a global fitting parameter in addition to the set of hard - core radii used in @xcite .
however , the effect of apparent strangeness non - equilibration can be more successfully explained by the hypothesis of separate chemical fo for all strange hadrons . since all the hadrons made of @xmath13 and @xmath14 quarks are under thermal equilibration whereas the hadrons containing @xmath15 quark are not , then it is reasonable to assume two different fos for these two kinds of particles . following this conclusion in ref .
@xcite a separate strangeness fo ( sfo ) was introduced .
note , that according to @xcite both fo and sfo parameters are connected by the conservation laws of entropy , baryonic charge and isospin projection , while the net strangeness is explicitly set to zero at fo and at sfo .
these conservation laws are crucial elements of the concept of separate sfo developed in @xcite which allows one to essentially reduce the number of independent fitting parameters .
another principal element that differs the hrgm of @xcite from the ideal gas treatment used in @xcite is the presence of multi - component hard - core repulsion . using the hrgm of @xcite it was possible to successfully describe all hadron multiplicities measured in a+a collisions at ags , sps and rhic energies with @xmath6 1.06 .
the concept of separate sfo led to a systematic improvement of all experimental data description .
however , the topmost point of the strangeness horn again was not fitted even within the experimental error , although the general description of @xmath16 ratio energy dependence was rather good except for the upper point .
since an introduction of the @xmath17 factor demonstrated a remarkable description of all points of the strangeness horn , whereas the separate sfo led to a systematic improvement of all hadron yields description , we decided to combine these elements in order to describe an experimental data with the highest possible quality . in this way
we would like to examine the problem whether the concept of separate sfo is able to completely explain a possible non - equilibrium of strange charge and whether on top of the sfo there exist a necessity to employ the @xmath0 factor in statistical approach .
note that from the academic point of view the problem of residual strangeness non - equilibration , i.e. the question whether the strange charge is or is not in full chemical equilibrium , is of principal importance .
this ambitious task is the main goal of the present paper .
evidently , the best tool for such a purpose is the most successful version of the hrgm , i.e. the hrgm with the multi - component hadronic repulsion and a separate sfo . as it will be shown below , such an approach allows us to describe 111 independent hadron yield ratios measured for 14 values of the center of mass collision energy @xmath18 in the interval from 2.7 gev to 200 gev with very high quality .
the paper is organized as follows .
the basic features of the developed model are outlined in section 2 , while the fitting procedure of the present model is outlined in section 3 .
the main obtained results are compared with the other models in section 4 . in section 5
we discuss in detail the new fit of hadronic multiplicity ratios with two chemical freeze - outs and @xmath0 factor , while section 6 contains our conclusions .
in what follows we treat a hadronic system as a multi - component boltzmann gas of hard spheres . the effects of quantum statistics are negligible for typical temperatures of the hadronic gas whereas the hard - core repulsion between the particles significantly affects a corresponding equation of state @xcite .
the present model is dealing with the grand canonical treatment .
hence a thermodynamic state of system under consideration is fixed by the volume @xmath19 , the temperature @xmath20 , the baryonic chemical potential @xmath21 , the strange chemical potential @xmath22 and the chemical potential of the isospin third component @xmath23 .
these parameters control the pressure @xmath24 of the system .
in addition they define the densities @xmath25 of corresponding charges @xmath26 ( @xmath27 ) . introducing the symmetric matrix of the second virial coefficients @xmath28
, we can obtain the parametric equation of state of the present model in a compact form @xmath29 using the partial pressure @xmath30 of @xmath31-th sort of particles . the equation of state is written in terms of the solutions @xmath30 of the following system @xmath32 \ , , \quad \quad \\ & & \hspace*{-4mm}\phi_i ( t ) = \frac{g_i}{(2\pi)^3}\int \exp\left(-\frac{\sqrt{k^2+m_i^2}}{t } \right)d^3k \ , .
% \label{eqiv } % \end{aligned}\ ] ] each @xmath33 sort is characterized by its full chemical potential @xmath34 , mass @xmath35 and degeneracy @xmath36 . the function @xmath37 denotes the corresponding particle thermal density in case of ideal gas .
the obtained model parameters for two freeze - outs and their dependence on the collision energy are discussed in the next section .
they are obtained for the following values of hard - core radii which were determined earlier in @xcite : @xmath38 fm for pions , @xmath39 fm for kaons , @xmath40 for all other mesons and @xmath41 fm for all baryons . in order to account for the possible strangeness non - equilibration
we introduce the @xmath17 factor in a conventional way by replacing @xmath42 in eq .
( [ eqii ] ) as @xmath43 where @xmath44 is a number of strange valence quarks plus number of strange valence anti - quarks .
the principal difference of the present model from the traditional approaches is that we employ an independent chemical fo of strange particles .
let us consider this in some detail .
the independent freeze - out of strangeness means that inelastic reactions ( except for the decays ) with hadrons made of s quarks are switched off at the temperature @xmath45 , the baryonic chemical potential @xmath46 , the strange chemical potential @xmath47 , the isospin third projection chemical potential @xmath48 and the three - dimensional emission volume @xmath49 . in general case
these parameters of sfo do not coincide with the temperature @xmath50 , the chemical potentials @xmath51 , @xmath52 , @xmath53 and the volume @xmath54 which characterize the freeze - out of non - strange hadrons .
the particle yields are given by the charge density @xmath55 in ( 1 ) and the corresponding volume at fo and at sfo .
at the first glance a model with independent sfo contains four extra fitting parameters for each energy value compared to the traditional approach ( temperature , three chemical potentials and the volume at sfo instead of strangeness suppression / enhancement factor @xmath17 ) .
however , this is not the case due to the conservation laws . indeed , since the entropy , the baryonic charge and the isospin third projection are conserved , then the parameters of fo and sfo are connected by the following equations @xmath56 where the entropy density @xmath57 , the density of baryonic charge @xmath58 and the density of the isospin third projection @xmath59 are found from the usual thermodynamic identities at the sfo ( a = sfo ) or at fo ( a = fo ) . the effective volumes can be excluded , if these equations are rewritten as @xmath60 thus , the baryonic @xmath46 and the isospin third projection @xmath48 chemical potentials at sfo are now defined by eqs .
( [ eq : fo_sfo1 ] ) .
note , that the strange chemical potentials @xmath52 and @xmath47 are found from the condition of vanishing net strangeness at fo and sfo , respectively .
therefore , the concept of independent sfo leads to an appearance of one independently fitting parameter @xmath45 .
hence , the independent fitting parameters are the following : the baryonic chemical potential @xmath21 , the chemical potential of the third projection of isospin @xmath23 , the chemical freeze - out temperature for strange hadrons @xmath45 , the chemical freeze - out temperature for all non - strange hadrons @xmath50 and the @xmath17 factor ( i.e. 5 fitting parameters for each collision energy ) . an inclusion of the width @xmath61 of hadronic states is an important element of the present model .
it is due to the fact that the thermodynamic properties of the hadronic system are sensitive to the width @xcite . in order to account for the finite width of resonances we perform the usual modification of the thermal particle density @xmath42 .
namely , we convolute the boltzmann exponent under the integral over momentum with the normalized breit - wigner mass distribution . as a result , the modified thermal particle density of @xmath33 sort hadron acquires the form @xmath62 here @xmath35 denotes the mean mass of hadron and @xmath63 stands for the threshold in the dominant decay channel .
the main advantages of this approximation is a simplicity of its realization and a clear way to account for the finite width of hadrons .
it is appropriate here to mention that one could use other prescriptions to account for the width of resonances .
however , in a recent work @xcite it was shown the breit - wigner prescription ( [ eqiii ] ) can provide somewhat better ( about 20% ) quality of the fit than the gaussian attenuation of resonance mass and essentially better one compared to the case without accounting for the width . at the same time in @xcite
it was found that within the error bars such fo parameters as temperature , chemical potentials and @xmath17 factor are the same for different prescriptions of resonance width accounting .
clearly , it is reasonable to expect that other physically motivated ways to account for the resonance width should give similar results .
therefore , in this work we employed the prescription ( [ eqiii ] ) , which provides the better description of the data .
the observed hadronic multiplicities contain the thermal and decay contributions .
for example , a large part of pions is produced by the decays of heavier hadrons .
therefore , the total multiplicity is obtained as a sum of thermal and decay multiplicities , exactly as it is done in a conventional model . however , writing the formula for final particle densities , we have to take into account that the fo and sfo volumes can be different : @xmath64 here the first term on the right hand side is due to decays after fo whereas the second one accounts for the strange resonances decayed after sfo .
the factor @xmath65 can be replaced by @xmath66 due to baryonic charge conservation .
@xmath67 denotes the branching ratio of the y - th hadron decay into the x - th hadron , with the definition @xmath68 = 1 used for the sake of convenience .
the input parameters of the present model ( masses @xmath35 , widths @xmath61 , degeneracies @xmath36 and branching ratios of all strong decays ) were taken from the particle tables of the thermodynamical code thermus @xcite .
* data sets*. the present model is applied to fit the data .
we take the ratios of particle multiplicities at midrapidity as the data points . in contrast to fitting multiplicities themselves
such an approach allows us to cancel the possible experimental biases . in this paper
we use the data set almost identical to ref . @xcite . at the ags energies ( @xmath69 agev or @xmath70 agev ) the data are available with a good energy resolution above 2 agev .
however , for the beam energies 2 , 4 , 6 and 8 agev only a few data points are available .
they corresponds to the yields for pions @xcite , for protons @xcite , for kaons @xcite ( except for 2 agev ) .
the integrated over @xmath71 data are also available for @xmath12 hyperons @xcite and for @xmath72 hyperons ( for 6 agev only ) @xcite . however , as was argued in ref .
@xcite , the data for @xmath12 and @xmath72 should be recalculated for midrapidity .
therefore , instead of raw experimental data we used the corrected values from @xcite . next comes the data set at the highest ags energy ( @xmath73 agev or @xmath74 agev ) .
similarly to @xcite , here we analyzed only the na49 mid - rapidity data @xcite .
since the rhic high energy data of different collaborations agree with each other , we analyzed the star results for @xmath75 gev @xcite , @xmath76 gev @xcite , @xmath77 gev @xcite and @xmath78 200 gev @xcite .
.the fit results of different versions of the hrgm are compared for 14 values of the center of mass collision energies : the column @xmath79 corresponds to a single fo model of @xcite ; the column @xmath80 is found for the sfo with @xmath81 @xcite ; the column @xmath82 corresponds to the sfo+@xmath17 fit with added data points for @xmath83 while the column @xmath84 is obtained by the direct sfo@xmath85 fit .
@xmath86 indicates the available number of independent hadronic ratios at given center of mass collision energy @xmath18 . in the row sum we list the sum of @xmath31-th column ,
while in the bottom row the number of degrees of freedom of each hrgm version is shown ( for more details see the text ) . [ cols="^,^,^,^,^,^ " , ] in fig .
[ fig : saguniiib ] we show the ratio of baryonic charge densities at sfo and at fo within the sfo@xmath87 model , which coincides with the inverse ratio of corresponding freeze - out volumes @xmath88 due to the baryonic charge conservation . from this figure
one can see that visually small difference of temperatures and baryonic chemical potentials at sfo and fo leads , nevertheless , to quite sizable difference of other thermodynamic quantities . as one
can see from the upper panel of fig .
[ fig : saguniv ] , this is also true for the pressure existing at sfo and at fo . also we confirm the existence of irregularities in the fo pressure found earlier @xcite .
similar irregularities we find for the sfo pressure and for the effective number of degrees of freedom for sfo , @xmath89 , i.e. for the ratio of sfo pressure to the fourth power of the sfo temperature . from the upper panel of fig .
[ fig : saguniv ] one can see that the largest increase of sfo pressure per increase of the center of mass energy of collision occurs at @xmath90 gev .
in other words , for about 14% increase of @xmath18 the sfo pressure increases in 5.3 times and the ratio @xmath89 increases on about 65 % . according to the recent work @xcite these and other irregularities observed at chemical
fo @xcite are signaling about the formation of the mixed quark - gluon - hadron phase and such an explanation can be experimentally verified in a few years at fair and nica .
one more important finding of the present work can be seen from a comparison of the upper and lower panels of fig .
[ fig : saguniv ] . note that in contrast to the temperature or baryonic chemical potential , the pressure allows one easier to distinguish the sfo from fo .
moreover , comparing the squares in the upper and lower panels of fig .
[ fig : saguniv ] one immediately concludes that the model of a single chemical fo @xcite with the same value of hard - core radius for all hadrons describes just the sfo for all values of @xmath18 below 62.4 gev .
the same is true for a single fo model @xcite with different hard - core radii discussed above .
this peculiar result can be easily understood , if one recalls that for the most values of collision energy the number of ratios involving strange hadrons is essentially larger than the number of ratios with non - strange particles . at the same time a single chemical fo model reproduces the fo pressure of the sfo+@xmath17 model only at @xmath91 gev .
the corresponding reasons we will discuss in the next section , while here we mention that at high rhic energies the fit quality of all models , including the single fo one , is rather high as one can see from table i. the main part of @xmath92 at these energies is formed by poor description of @xmath93-mesons at @xmath94 gev , @xmath12 hyperon at @xmath95 gev and @xmath96 hyperons at @xmath97 gev .
such a comparison of the single fo model and the sfo+@xmath17 model pressures allows us to explain a cause of why in previous thorough analysis of the particle ratios within the realistic single fo model with the same hard - core radius of hadrons the main conclusion was that there is no deviation of strange particles from chemical equilibrium for the mid - rapidity data . in other words , it is possible to naturally explain the reason of why in @xcite it was found that within the error bars @xmath98 .
our direct comparison of the sfo+@xmath17 model for the mid - rapidity data shows that the single fo models with the same or different hard - core hadronic radii reproduce the sfo pressure with @xmath98 almost at all collision energies , except for @xmath99 gev and for @xmath100 gev . in the former case one
finds that @xmath101 , while in the latter case @xmath98 , but , as we discussed above , for high collision energies the pressure at fo found within the sfo+@xmath17 model reproduces the single fo model pressure .
therefore , the main reason of why the chemical non - equilibrium of strange charge was not found within the single fo models is that the fitting procedure mainly described the ratios involving the strange particles and , hence , it would have been more appropriate to consider the chemical non - equilibrium of non - strange hadrons .
the findings discussed above motivate us to study in some details what ratios and at what energies are improved .
the most significant improvements correspond to the collision energies @xmath102 6.3 , 7.6 , and 12 gev , that are plotted in figs .
[ fig : sagunv ] and [ fig : sagunvb ] . figs .
[ fig : sagunv ] and [ fig : sagunvb ] demonstrate very high fit quality , especially for such traditionally problematic ratios as @xmath16 , @xmath103 , @xmath104 and @xmath105 , which is achieved within the sfo@xmath87 model compared to the single fo model and the sfo one .
for instance , for @xmath102 7.6 gev seven ratios out of ten are improved , while for other energies the improvements are less significant . on the contrary , the particle ratios measured at @xmath102 17 gev ( see fig . [
fig : sagunvb ] ) are improved within the sfo model , while the sfo@xmath87 fit practically does not lead to any significant improvement compared to the sfo model .
also we found that the sfo@xmath87 fit leads to a selective improvement and to a certain degradation of the fit quality of various ratios for different collision energies .
for instance , the @xmath103 ratio is slightly increased for @xmath102 6.3 and 7.6 gev , but the situation drastically changes for @xmath102 12 gev .
the same tendency is typical for @xmath106 . on the contrary , for @xmath107 ratio
there is a noticeably worse data description within sfo+@xmath17 approach at @xmath102 6.3 , 7.6 gev , but for @xmath102 12 gev the fit quality is sizably better compared to all previous approaches .
thus , within the present model we reveal a noticeable change in the trend of some ratios at @xmath102 7.6 - 12 gev , while at @xmath102 12 gev we do not observe any sizable improvement compared to the sfo model .
a special attention in our consideration was paid to the strangeness horn , i.e. to the @xmath16 ratio , because such a ratio is traditionally the most problematic one for the hrgm to fit it .
as one can see from fig .
[ fig : sagunvi ] , the remarkable @xmath16 fit improvement for @xmath102 2.7 , 3.3 , 4.3 , 4.9 , 6.3 , 7.6 , 12 gev justifies the usage of the present model .
quantitatively , we found that the @xmath108 improvement for the sfo+@xmath17 model is @xmath108=1.5/14 , i.e. even better than it was achieved in @xcite with @xmath108=3.3/14 for the @xmath17 fitting approach and with @xmath108=6.3/14 for the sfo model with @xmath81 . from fig .
[ fig : sagunvii ] one can see that at two highest rhic energies the description of ratios within the single fo model is very good for all ratios except for @xmath104 at @xmath95 gev and for @xmath109 at @xmath97 gev .
the main reason is that at these energies all chemical potentials except for the strange one are almost zero and hence the number of particles and antiparticles is almost the same .
as one can see from the upper panel of fig .
[ fig : sagunvii ] , the sfo@xmath87 fit significantly improves only the @xmath110 and @xmath111 ratios compared to the single fo model , i.e. only two ratios of strange hadrons responded to the variation of two additional parameters . for @xmath97
gev the sfo@xmath87 fit significantly improves only the @xmath109 ratio and worsens the @xmath112 ratio and @xmath113 ( less ) , i.e. only three ratios of strange hadrons responded to such a sophisticated fit . for @xmath94 gev
only two ratios out of five include kaons and , hence , using @xmath45 and @xmath17 one can perfectly reproduce the strange particle ratios without affecting the non - strange ones . treating the @xmath45 values found within the sfo model as an additional datum for the sfo@xmath87 fit ,
we obtain the same result .
therefore , in contrast to low collision energies at high collision energies only a few ratios with strange particles can be improved by simultaneous variation of @xmath45 and @xmath17 and , hence , the sfo can not represent the majority of fitted ratios which are well reproduced even within the single fo model with a single or with several hard - core radii .
the lower panel of fig .
[ fig : saguniv ] evidently supports such a conclusion for the hrgm with the same value of hadronic hard - core radius . within the sfo@xmath87 model the @xmath114 and @xmath115 ratios
demonstrate some worsening compared to less sophisticated models . in fig .
[ fig : sagunviii ] we show that the sfo@xmath87 model still does not improve these ratios .
the @xmath114 fit quality , for instance , is @xmath108=10/8 .
hence , up to now the best fit of the @xmath114 ratio was obtained within the sfo approach with @xmath81 . as it was mentioned in @xcite
a too slow decrease of model results for @xmath114 ratio compared to the experimental data is typical for almost all statistical models . evidently , the too steep rise in @xmath114 behavior is a consequence of the @xmath116 anomaly @xcite .
similar results are reported in refs .
@xcite as the @xmath117 , @xmath2 and @xmath3 selective suppression .
since even an introduction of the separate strangeness freeze - out with the strangeness enhancement factor does not allow us to better describe these ratios , we believe that there is a corresponding physical reason which is responsible for it .
one of them could be a necessity to introduce the different hard - core radius @xmath118 for the @xmath119 ( anti)hyperons @xcite .
we present a thorough investigation of the data measured at ags , sps and rhic energies within different versions of the multi - component hadron resonance gas model . the suggested approach
to separately treat the freeze - outs of strange and non - strange hadrons with the simultaneous @xmath17 fitting gives rise for the top - notch strangeness horn description with @xmath108=1.5/14 .
the developed model clearly demonstrates that the successful fit of hadronic multiplicities includes all the advantages of these two approaches discussed in @xcite . as a result for @xmath102 6.3 , 7.6 , 12 ,
130 gev we found a significant data fit quality improvement . at the same time
the lack of available data at @xmath18 = 2.7 , 3.3 , 3.8 , 4.3 , 9.2 , 62.4 gev forced us to redefine the fitting procedure at these collision energies in order to avoid the mathematical inconsistency which in combination with the large experimental error bars led to rather large uncertainties of the fitting parameters .
the suggested redefinition of the fitting procedure by including the @xmath45 temperatures obtained for these energies within the sfo model allowed us to avoid the mathematical problems and to get the safe answers on the values of the residual chemical non - equilibrium of strange particles .
the developed sophisticated hrgm , i.e. the sfo+@xmath17 model , allowed us to describe the hadron multiplicity ratios with rather high quality @xmath120 0.93 .
this very fact demonstrates that the suggested approach is a precise tool to elucidate the thermodynamics properties of hadron matter at two chemical freeze - outs .
the fresh illustrations to this statement can be found in @xcite .
the achieved total value of @xmath121 for the sfo+@xmath17 model is almost 50% lower than the @xmath92 value found for the single fo model and 30% lower than the sfo model @xmath92 value .
the obtained @xmath17 values are consistent with the conclusion @xmath98 ( within the error bars ) .
an evident exception is the topmost point of the strangeness horn ( located at @xmath102 7.6 gev ) , at which the mean value of the strangeness enhancement factor is @xmath122 . to reveal the physical reason for such a deviation we need more experimental data with an essentially higher accuracy .
one of the main results obtained here is that the idea of separate chemical fo of strange hadrons provides a very high quality of the data description .
further inclusion of the chemical non - equilibrium on top of the sfo is consistent with the result @xmath98 for all energies except for @xmath102 7.6 gev .
thus , the found residual chemical non - equilibrium of strange particles is weak and , hence , it can be safely ignored for all energies except for @xmath102 7.6 gev .
the strange charge enhancement of about 20% found at this collision energy allowed us to perfectly describe the topmost point of the strangeness horn , but at the expense of the worsening of @xmath114 and @xmath123 ratios .
in addition , the description of ratios containing the non - strange particles , especially such as @xmath103 and @xmath106 , gets better compared to previously reported results @xcite .
the remaining problem of the ratios involving the @xmath12 and @xmath124 ( anti)hyperons can be resolved by an inclusion of the different hard - core radius for @xmath12 ( anti)particles @xcite , but such a treatment is out of scope of the present work .
the performed analysis of the sfo+@xmath17 model hadronic pressures existing at fo and at sfo allowed us to elucidate an important conclusion that the single fo models with the same hard - core radius @xcite or with different hard - core radii @xcite for all hadrons reproduce the sfo pressure for all collision energies below @xmath94 gev .
the main reason for such a behavior is that the number of ratios involving strange hadrons is larger than the number of ratios with non - strange hadrons .
also we report an existence of strong jumps in the sfo pressure , the sfo temperature and the corresponding effective number of degrees of freedom , when the center of mass collision energy changes from 4.3 to 4.9 gev . based on the concept of non - smooth chemical freeze - out introduced recently in @xcite , we parameterized the dependencies @xmath125 and @xmath126 which can be verified in the future experiments planned at fair ( gsi ) and nica ( jinr ) .
we hope that the high precision data measured in these experiments will allow us to finally answer the question whether the residual non - equilibrium of strange charge is necessary to describe the topmost point of the strangeness horn or the concept of two separate chemical freeze - out for strange and non - strange hadrons can do this without introducing the @xmath17 factor .
* acknowledgments .
* we would like to thank a. andronic for providing an access to well - structured experimental data .
the authors are thankful to i. n. mishustin , d. h. rischke and l. m. satarov for valuable comments .
, a.i.i . and g.m.z .
acknowledge a support of the fundamental research state fund of ukraine , project no f58/04 .
k.a.b . acknowledges also a partial support provided by the helmholtz international center for fair within the framework of the loewe program launched by the state of hesse . | we present an elaborate version of the hadron resonance gas model with the combined treatment of separate chemical freeze - outs for strange and non - strange hadrons and with an additional @xmath0 factor which accounts for the remaining strange particle non - equilibration . within
suggested approach the parameters of two chemical freeze - outs are connected by the conservation laws of entropy , baryonic charge , third isospin projection and strangeness .
the developed model enables us to perform a high - quality fit of the hadron multiplicity ratios measured at ags , sps and rhic with @xmath1 .
a special attention is paid to a successful description of the strangeness horn .
the well - known problem of selective suppression of @xmath2 and @xmath3 hyperons is also discussed .
the main result is that for all collision energies the @xmath0 factor is about 1 within the error bars , except for the center of mass collision energy 7.6 gev at which we find about 20% enhancement of strangeness .
also we confirm an existence of strong jumps in pressure , temperature and effective number of degrees of freedom at the stage of strange particle chemical freeze - out , when the center of mass collision energy changes from 4.3 to 4.9 gev .
we argue that these irregularities may signal about the quark - hadron phase transition . |
the existence of high - mass , x - ray luminous clusters of galaxies at @xmath12 provides a strong constraint on cosmological models ( eke 1996 ) .
the presence of any @xmath13 clusters at @xmath14 invalidates most @xmath15 cdm models ( bahcall _ et al .
_ 1997 , henry 1997 ) .
further , the surface densities , amounts of sub - clustering , and mass to light ratios for these high @xmath16 clusters can place even stronger constraints on cosmological models and the nature of dark matter ( wilson 1997 , crone 1996 ) . currently , x - ray observations provide the most effective way of finding massive clusters at high redshifts .
the einstein extended medium sensitivity survey ( emss ) and the rosat north ecliptic pole ( nep ) survey have , so far , found three @xmath14 clusters of galaxies ( gioia and luppino 1994 , henry 1997 ) .
one of these clusters , ms 1054@xmath1703 , has had its mass confirmed by weak gravitational lensing measurements ( luppino and kaiser 1997 , hereafter lk ) . in this paper
we report on the detection of weak gravitational lensing signals in two more @xmath18 clusters of galaxies , ms 1137@xmath066 and rxj 1716@xmath067 .
ms1137@xmath066 is one of the high redshift clusters from the emss ( luppino and gioia 1995 ) .
it has a redshift of @xmath19 and an x - ray luminosity of @xmath20 erg s@xmath21 .
rxj 1716@xmath067 is a recently discovered high - redshift cluster from the rosat nep survey .
it has a redshift of @xmath2 and an x - ray luminosity of @xmath22 erg s@xmath21 ( converted from the 0.5 - 2.0 kev rosat data assuming a @xmath23 6 kev ) .
the x - ray luminosities of these clusters are comparable to ms 1054@xmath1703 and suggest they may also be massive systems .
our observations , data reduction , and the weak lensing analysis are outlined in 2 . in
3 we discuss the properties of the cluster and the weak lensing results .
section 4 contains our conclusions .
we assume an @xmath24 cosmology throughout this paper , and all errors are @xmath25 .
very deep optical @xmath26-band images of rxj 1716@xmath067 and ms 1137@xmath066 were obtained with the keck ii 10 m telescope on the nights of august 17 - 18 1996 and january 10 - 11 , 1997 .
the images were obtained using lris ( oke 1995 ) in imaging mode , resulting in a scale of 0215 pixel@xmath21 and a field of view of 6@xmath27 8 .
total exposure times were 7500 s and 8700 s respectively .
optical @xmath28-band images were obtained with the uh 2.2 m telescope on the nights of july 20 - 22 , 1996 for rxj 1716@xmath067 and april 6 - 7 , 1995 and april 29 - may 1 , 1997 for ms 1137@xmath066 .
a thinned tek 2048@xmath29 ccd was mounted at the f/10 rc focus resulting in a scale of 022 pixel@xmath21 and a field of view of 75 @xmath27 75 .
total exposure times were 26100 s and 21000 s respectively .
the individual images in each filter were first debiased and then flattened using a median of all the images taken in that filter during each set of observations .
the @xmath28-band 88 images were shifted into registration and summed with cosmic - ray rejection .
the @xmath26-band keck images were re - mapped with a bi - cubic polynomial to correct for the focal plane curvature and to map the images onto the @xmath28-band plate scale .
the sky was removed by subtracting a smoothed fit to the minima in the sky .
the images were then summed with cosmic - ray rejection , with images showing high extinction after correction for airmass being rejected from the sum .
the resulting images had seeing of @xmath30 in @xmath26-band and @xmath31 in @xmath28-band for rxj 1716@xmath067 and @xmath30 in @xmath26-band and @xmath32 in @xmath28-band for ms 1137@xmath066 .
these images , along with some shallower @xmath33-band images obtained from the uh 88 telescope , have been combined into the three color images shown in figure 1 ( plate xx : ms 1137@xmath066 ) and figure 2 ( plate xx : rxj 1716@xmath067 ) .
photometric calibration was performed using the standard stars of landolt ( 1992 ) .
the 1@xmath34 surface brightness limits of the summed @xmath26-band and @xmath28-band images for a seeing - sized disc are 29.4 and 26.6 for ms 1137@xmath066 and 29.0 and 26.8 for rxj 1716@xmath067 .
the first step in the weak lensing analysis was detecting and measuring the positions and luminosities of all the objects in the field .
we used a hierarchical peak finding algorithm to detect objects ( kaiser , squires , and broadhurst 1995 ) , which provided both centroid positions and a rough measure of the object size .
the luminosity of each object was measured using a circular aperture with radius equal to three times the smoothing radius at which the object was first detected .
the catalog of objects was then used to mask the summed image .
small , residual bias around each object left by the sky subtraction was then measured in the masked image and subtracted .
noise peaks and non - splitable groups of objects were removed by rejecting extremely small and large objects , as well as objects with abnormally high eccentricities . with very few exceptions , all objects detected in the i - band images
were also detected in the much deeper r - band images .
the i - band magnitudes were calculated using an aperture @xmath35 , where @xmath36 is the smoothing scale at which the object is first detected in the r - band image .
the resulting i - band magnitudes agree with the magnitudes from the initial detection in the i - band images to within 0.1 mag rms . for purposes of color selection
, it is assumed that all objects present in the r - band images but not detected in the i - band images are bluer than objects of the same r - band magnitude with a detection in the i - band images .
the ellipticities of each object were measured using optical polarizations @xmath37 formed from the quadrupole moments @xmath38 where @xmath39 is the flux density and @xmath40 is a gaussian weighting function of a scale equal to the smoothing radius at which the object was first detected . a low order polynomial
fit to a sample of bright but unsaturated stars ( @xmath41 distributed across the field ) was used to correct the ellipticities of the objects for psf variations across the field ( spherical aberrations and coma in the focal plane , wind shake on the telescope , etc . ) .
the resulting ellipticities should average to the object ellipticity , possibly sheared by the gravitational lens , smeared by a circular psf .
stars were removed from the catalog by using half - light radii of the objects . given that the smeared ellipticities reduce the strength of any lensing signal present , the next step is to boost the ellipticities of the galaxies to remove the effects of the psf smearing .
this boost factor was calculated using the shear polarizability @xmath42 ( lk ) , where @xmath43 and @xmath44 are as defined in kaiser , squires , and broadhurst ( 1995 ) , and @xmath45 denotes the values for stellar objects .
the individual values of @xmath46 are rather noisy for small galaxies , but the mean value , @xmath47 , varies smoothly as a function of galaxy size .
thus by binning the galaxies by size an estimate of the shear @xmath48 was calculated .
this shear is then related to the dimensionless surface density @xmath49 by @xmath50 where @xmath51 and where the critical density @xmath52 .
because of the @xmath53 factor in @xmath54 , the mass obtained from @xmath49 is dependent on the redshift of the source galaxies used in the weak lensing analysis . in this paper
we will be assuming that the background galaxies lie on a sheet at @xmath55 .
if , instead , the background galaxies were to lie on sheets at @xmath56 , @xmath57 , or @xmath58 , the measured masses would be @xmath59 , @xmath60 , or @xmath61 the quoted mass .
it should be noted that the intrinsic ellipticities of the galaxies can not be removed from @xmath62 measurements , and are the dominant source of noise in the analysis .
we used two different methods to derive @xmath49 from the measured shears of the background galaxies .
the first was the unbiased inversion algorithm of squires and kaiser ( 1996 ) .
this technique , while producing two dimensional maps of the surface density , is only able to determine @xmath49 to within an unknown constant .
the second technique is that of aperture mass densitometry , which measures the mass interior to a given radius ( fahlmen 1994 ) .
the traditional statistic of aperture mass densitometry @xmath63 provides a lower bound on @xmath64 interior to radius @xmath65 .
however , because this statistic subtracts @xmath64 of the annulus outside @xmath65 , the final measured mass @xmath66 depends not only on the strength of the detected lensing but also on the mass profile of the cluster . because of the large numbers of background galaxies in our deep images , we have been able to use a slightly altered statistic @xmath67 which allows the subtraction of a constant @xmath68 for a fixed inner annulus radius @xmath69 for all apertures in the measurement , thereby eliminating the dependence of the statistic on the mass profile of the cluster
the measured mass profile @xmath70 is now the mass of the cluster minus an unknown , but presumably small , constant times @xmath71 . in both statistics
the tangential shear is @xmath72 , where @xmath73 and @xmath74 is the azimuthal angle with respect to some chosen center ( which we take to be the peak of the smoothed light distribution for each cluster ) .
both clusters catalogs of objects were then divided into several subcatalogs by magnitude and color of the galaxies . as was the case with ms 1054@xmath1703 in lk , both ms 1137@xmath066 and rxj 1716@xmath067 had the strongest weak lensing signal from the catalogs containing the faintest , bluest galaxies , specifically @xmath75 and @xmath76 .
the discussion of the results below come from the subcatalogs of the faint blue galaxies .
this cluster is easily identifiable as the compact group of red galaxies in the center of the field , seen in figure 1 ( plate xx ) . using a 500 @xmath4 kpc aperture and a galaxy color selection of @xmath77 to isolate the cluster galaxies , an over - density of @xmath78 galaxies above the background level was determined from the galaxy density in the field outside the aperture .
this corresponds to an abell class 3 cluster richness ( bahcall 1981 ) .
a smoothed image of the luminosities of the selected galaxies is shown in figure 3b .
the centroid of the cluster light peak is consistent with the position of the bcg .
the total magnitude for all selected galaxies inside a 500 @xmath4 kpc aperture centered on the bcg , minus a background determined from the galaxies outside the aperture , is @xmath79 , which corresponds to a cluster luminosity @xmath80 .
as can be seen in figure 3a , ms 1137@xmath066 has a mass distribution as compact as the light ( figure 3b ) and galaxy distribution .
the small peak just north of the cluster is centered around three galaxies with colors similar to the cluster galaxies , but of unknown redshift .
the minimum mass enclosed in a 500 @xmath4 kpc radius centered on the bcg ( coincident with the center of the mass peak in the 2-d reconstruction ) , calculated using aperture densitometry , is @xmath81 .
the small northern peak is included in this estimate , but contributes less than @xmath82 of the total mass detected . combining this mass estimate with the cluster luminosity estimate from 3
gives @xmath83 , which is similar to that found for ms 1054@xmath1703 ( lk ) .
the ultra - deep keck image has also revealed a system of giant arcs present in the core of ms 1137@xmath066 .
the brightest arc is located 55 ( 22 @xmath4 kpc ) south - west of the bcg and has a surface brightness of @xmath84 mag@xmath85arcsec@xmath29 .
a second arc is located 180 ( 73.6 @xmath4 kpc ) north - east of the bcg and has a surface brightness of @xmath86 mag@xmath85arcsec@xmath29 .
there is a potential counter arc with the same surface brightness located 185 ( 75.7 kpc @xmath4 ) south of the bcg .
there are also two more arc candidates , all of which are marked in figure 1 ( plate xx ) .
no redshifts have yet been obtained for any of these arcs .
it is interesting to note , however , that the outer giant arc is located presumably on the circle at which @xmath87 for the redshift of the arc .
the weak lensing result , when extrapolated into this radius , has a @xmath88 .
this implies that the source galaxy redshift of the outer giant arc is higher than the typical redshifts of the faint background galaxy ( fbg ) population over the magnitudes which we have measured .
the outer giant arc can be used to measure the mass of ms 1137@xmath066 interior to its radius unambiguously , and , because of the large radius of the arc from the cluster core , serve as an anchor point to the masses measured by the weak lensing .
unlike ms 1137@xmath066 , rxj 1716@xmath067 does not appear to be a well formed cluster .
as can be seen in figure 2 ( plate xx ) , the bcg is located on the north - western edge of the main group of galaxies , and there is a smaller clump of galaxies to the north - east of the main group .
a filament of galaxies extends north - east of the cluster , disappearing off the edge of the image . a smoothed image of the luminosities of galaxies with @xmath77 , shown in figure 3d ,
has the main peak in the cluster luminosity located 8southeast of the bcg . in a 500 @xmath4 kpc aperture centered on the peak of the cluster luminosity ,
rxj 1716@xmath067 has an over - density of galaxies of @xmath89 , corresponding to abell class 3 ( bahcall 1981 ) , and a cluster magnitude of @xmath90 which corresponds to @xmath91 .
while much of the fine structure seen in the smooth cluster light image ( figure 3d ) for rxj 1716@xmath067 is lost in the noise of the mass reconstruction ( figure 3c ) , the mass of the northeastern group of galaxies is clearly separated from the mass of the main cluster .
further , while the mass peak of the northeastern group is consistent with the center of light and the galaxy distribution of the group , the center of mass of the main cluster is located 27(111 @xmath4 kpc ) east and 11(45 @xmath4 kpc ) north of the bcg .
this offset is in the same direction but far larger than the offset seen in the center of the light peak of the cluster .
the minimum mass enclosed in a 500 @xmath4 kpc radius aperture centered on the center of mass for the main cluster , calculated using aperture densitometry , is @xmath92 .
this includes both the mass of the main cluster and the smaller northeastern group of galaxies .
combining this with the cluster luminosity estimate of 3 gives @xmath93 .
the mass profiles of both clusters are plotted in figure 4 , along with `` universal '' cdm profiles of navarro , frenk , and white ( 1996 ) . while the cdm profiles are not robust fits ( both parameters in the model can be adjusted to a small extent without affecting the goodness of fit ) , they can be used as a general comparison with mass profiles of lower redshift clusters . in both clusters ,
the observed masses are similar to those measured in lower redshift clusters ( squires 1996 , fischer 1997 ) , but ms 1137@xmath066 is much more compact than usual ( navarro 1996 ) .
we believe that this compactness may be an artifact of projecting 3d filamentary structures onto a 2d plane and that the compact profile of ms 1137@xmath066 can be explained if we are looking down the length of a filamentary structure .
this would cause all the mass of the central filament to be projected as a large , steeply sloped 2d surface potential , and would not seriously affect the strength of the weak lensing provided the tube has a length smaller than the typical background galaxy
2d cluster center distance .
we do not , however , have enough redshifts on cluster galaxies to determine if this is indeed the case .
finally , these results show that a substantial part of the fbg population must lie at @xmath94 . because of the dependence of @xmath95 on both the redshift of the lensing cluster and the redshifts of the background galaxies , we believe that the redshift distribution of the fbg population can be measured by using a number of clusters at varying redshifts .
the outer giant arc in ms 1137@xmath066 will prove to be a valuable tool in this measurement .
the radius of the arc implies that the arc is being strongly lensed primarily by the core of the cluster , and any nearby galaxies will provide only a minor perturbation to the shape of the arc .
thus , once the redshift of the outer arc is known , it will be possible to obtain a mass measurement of the cluster interior to the arc with an uncertainty much lower than both weak lensing and x - ray observations can obtain .
we wish to thank gillian wilson , mark metzger , lev koffman , len cowie , dave sanders , john learned , phil fischer , james bauer , and neil trentham for their help and advice . we also wish to thank harald ebeling , chris mullis , and megan donahue for sharing their x - ray data with us before publication .
this work was supported by nsf grants ast-9529274 and ast-9500515 , nasa - stsci grant go-540201 - 93a , and asi grants ars-94 - 10 and ars-96 - 13 .
bahcall , n. a. , fan , x. , & cen , r. 1997 , , 485 , l53 .
bahcall , n. a. 1981 , , 246 , 122 .
crone , m. m. , evrard , a. e. , & richstone , d. o. 1996 , , 467 , 489 .
eke , v. r. , cole , s. , & frenk , c. s. 1996 , , 282 , 263 .
fahlman , g. , kaiser , n. , squires , g. , & woods , d. 1994 , , 437 , 56 .
fischer , p. , bernstein , g. , rhee , g. , & tyson , j. a. 1997 , , 113 , 521 .
gioia , i. m. & luppino , g. a. 1994 , , 94 , 583 .
henry , j. p. 1997
, , 489 , l1 .
henry , j. p. , gioia , i. m. , mullis , c. r. , clowe , d. i. , luppino , g. a. , boehringer , h. , briel , u. g. , voges , w. , & huchra , j. p. 1997
, , 114 , 1293 .
kaiser , n. , squires , g. , & broadhurst , t. 1995 , , 449 , 460 .
landolt , a. u. 1992 , , 104 , 340 .
luppino , g. a. & gioia , i. m. 1995 , , 445 , 77 .
luppino , g. a. & kaiser , n. 1997 , , 475 , 20 ( lk ) .
morgan , j. a. 1995 , paspcs , 77 , 129 .
navarro , j. f. , frenk , c. s. , & white , s. d. m. 1996 , , 462 , 563 .
oke , j. b. , cohen , j. g. , carr , m. , cromer , j. , dingizian , a. , harris , f. h. , labrecque , s. , lucinio , r. , schaal .
w. , epps , h. , & miller , j. 1995 , , 107 , 375 squires , g. & kaiser , n. 1996 , , 473 , 65 .
squires , g. , kaiser , n. , babul , a. , fahlman , g. , woods , d. , neumann , d. m. , & boehringer , h. 1996 , , 461 , 572 .
wilson , g. , cole , s. , & frenk , c. s. 1996 , , 282 , 501 . | we report the detection of weak gravitational lensing of distant background galaxies by two rich , x - ray luminous clusters of galaxies , ms 1137@xmath066 at @xmath1 and rxj 1716@xmath067 at @xmath2 .
we detect a shear signal in each cluster over the radial range @xmath3 700 @xmath4 kpc . assuming the background galaxies lie in a sheet at @xmath5 , we measure a mass of @xmath6 and @xmath7 at @xmath8 kpc radius for ms 1137@xmath066 , and a mass of @xmath9 and @xmath10 for rxj 1716@xmath067 at the same radius . both the optical lightmaps and weak lensing massmaps of rxj 1716@xmath067 show two spatially distinct sub - clusters , as well as a long filamentary structure extending out of the cluster to the northeast .
in contrast , ms 1137@xmath066 is an ultra - compact massive cluster in both massmaps and optical lightmaps , and contains a strongly - lensed arc system in the cluster core .
these data add to the growing number of massive clusters at @xmath11 . |
majority of stars reside in binary systems @xcite .
the mechanism of planet formation around binary systems would be different from that around single stars not only because the environment of the protoplanetary disk would be affected by the binary companion but also because the binary companion would affect the long - term stability of the planet orbit .
therefore , one of the most generic environments to be considered in the study of planet formation should be that of a binary . however , the major planet - formation scenarios that have been developed over the past decades , e.g. , core - accretion theory @xcite and disk instability theory @xcite , were mostly focused on the case of single stars .
although a considerable work has been done about the effect of binary companions on the planet formation ( e.g. , * ? ? ?
* ) and the long - term orbital stability ( e.g. , * ? ? ?
* ) , this work was restricted to mostly theoretical studies and thus many details about the planet formation scenario remain uncertain .
these details can be refined by the constraints provided by the sample of actually detected planetary systems .
unfortunately , there exists only a total of 19 known planets in 17 binary systems and most of these planets exist under a similar environment , i.e. circumbinary planets orbiting very close binaries . to give details about the planet formation in binary systems , therefore , it is important to detect more of such planets residing under various environments .
when a gravitational microlensing event is caused by a very wide binary object , the lensing behavior in the region around each lens component is approximated by chang - refsdal ( c - r ) lensing , which refers to the gravitational lensing of a point mass perturbed by a constant external shear @xmath0 @xcite . in the low shear regime ( @xmath1 )
, c - r lensing induces a small astroid - shape caustic around the lens .
the lensing behavior of a very close binary , on the other hand , can be approximated by a point - mass plus quadrupole lensing . in this case ,
an astroidal caustic similar to the c - r lensing caustic is produced around the center of mass of the binary .
for this reason , the event caused by a very wide or a very close binary lens is often referred to as the c - r lensing event . due to the existence of the caustic in the central region , c - r lensing events are identified by a short - term anomaly that appears near the peak of a very high - magnification event .
although c - r lensing events are routinely detected from microlensing surveys , not much attention is paid to them because they are thought to be simply one type of numerous binary - lens events and thus of little scientific importance . in this work , we point out that c - r lensing events provide an important channel to detect planets in binaries , both in close and wide binary systems . in order to demonstrate that dense and high - precision coverage of a c - r lensing - induced perturbation can provide a strong constraint on the existence of a planet in the wide range of the planet parameters , we present analysis of the actually observed c - r lensing event ogle-2015-blg-1319 .
the paper is organized as follows . in section 2 ,
we briefly describe the lensing properties in the cases where the lens is composed of a single , binary , and triple masses .
we also describe the c - r lensing behavior . in section 3
, we estimate the detection efficiency of planets in binaries by conducting analysis of the lensing event ogle-2015-blg-1319 .
we summarize the results and conclude in section 4 .
when a point - mass lensing event occurs the lensing behavior is described by the lens equation @xmath2 where @xmath3 and @xmath4 denote the complex notations of the source and image positions , respectively , and @xmath5 represents the complex conjugate of @xmath6 . here
all lengths are normalized to the angular einstein ring radius @xmath7 and the lens is positioned at the origin . solving the lens equation yields two solutions of image positions : one outside and the other inside the einstein ring .
the magnification @xmath8 of each image @xmath9 is given by @xmath10 where @xmath11 is the jacobian of the lens equation evaluated at the image positions @xmath12 and @xmath13 is the determinant of the jacobian . since the individual microlensing images can not be resolved , the observed lensing magnification is the sum of the magnifications of the individual images , i.e. @xmath14 . for a point mass ,
the magnification is represented analytically by @xmath15 where @xmath16 is the lens - source separation with the length normalized to @xmath7 , @xmath17 is the einstein time scale , @xmath18 is the time of the closest lens - source approach , and @xmath19 is the lens - source separation at @xmath18 . for a rectilinear relative lens - source motion ,
the lensing light curve is characterized by a smooth and symmetric shape @xcite .
when an event is produced by a lens composed multiple components , the lens equation is expressed as @xmath20 where @xmath21 and @xmath22 , and @xmath23 represent the location , mass fraction , and mass of each lens component , respectively .
the notation @xmath24 denotes the number of the lens components , and thus @xmath25 for a binary lens . here
all lengths ( @xmath26 and @xmath6 s ) are in units of the angular einstein radius corresponding to the total mass of the lens , @xmath27 .
one of the most important properties of a binary lens that differentiate from those of a single lens is the formation of caustics , which represent the sets of source positions at which @xmath28 and thus the magnification of a point source becomes infinite . as a result ,
a binary - lensing light curve can exhibit strong deviations when a source approaches close to or passes over the caustic .
caustics of a binary lens form a single or multiple sets of closed curves and each curve composed of concave curves that meet at cusps .
the topology of the binary - lens caustic is broadly classified into 3 categories @xcite . in the case of a binary where the binary separation is greater than @xmath29{\epsilon_1 } + \sqrt[3]{\epsilon_2})^{3/2}$ ] ( wide binary ) , there exist two sets of 4-cusp caustics that are located close to the individual lens components .
when the binary separation is smaller than @xmath29{\epsilon_1 } + \sqrt[3]{\epsilon_2})^{-3/4}$ ] ( close binary ) , the caustic is composed of 3 pieces , where the central caustic with 4 cusps is formed around the center of mass of the binary lens , and the other 2 triangular caustics are located away from the center of mass . in the intermediate separation region
, there exists a single big caustic with 6 cusps . for the visual presentation of the binary caustic topology , see figure 3 of @xcite . in the extreme case where the binary separation is much greater or less than @xmath7 ,
the 4-cusp central caustic has an astroid shape which is symmetric with respect to the binary axis and the line vertical to the binary axis . for the exact description of the lensing behavior produced by a lens system of a planet orbiting a binary object , one needs the triple lens equation , i.e. eq . ( [ eq4 ] ) with @xmath30 . with the addition of a third lens component , the complexity of the lensing behavior greatly increases and caustics can exhibit self - intersection and nesting @xcite .
( upper panel ) and the mass ratio between the planet and the host @xmath31 ( lower panel ) . both the caustic size and the binary separation are normalized to the angular einstein radius corresponding to the total lens mass . ]
a planetary lens corresponds to the extreme case of a binary lens where the mass of one of the lens components is much smaller than the other , i.e. @xmath32 and @xmath33 . in this case , the lens equation is approximated as @xmath34 where @xmath35 is the mass ratio between the planet and the host and @xmath36 represents the position of the planet with respect to the host .
we note that the lengths of @xmath26 and @xmath6 s in eq .
( 5 ) are normalized to the angular einstein radius corresponding to the mass of the primary lens , i.e. @xmath37 . for a planetary lens , however , the mass of the primary dominates , i.e. @xmath38 , and thus @xmath39 .
the planet induces two types of caustics , where one is located close to the host ( central caustic ) and the other is away from the host .
the central caustic has an arrowhead shape and its size is related to the star - planet separation @xmath40 and the mass ratio by @xcite @xmath41 due to the location of the central caustic close to the host , perturbations induced by the central caustic of a planet always appear near the peak of the lensing light curve produced by the host of the planet @xcite .
the lens equation of c - r lensing is expressed as @xmath42 which describes the lensing behavior around the primary lens with an external shear @xmath0 . here
lengths are given in units of the einstein radius corresponding to the primary mass .
the shear induces a caustic around the lens .
the shape of the caustic is similar to the central caustic of a very wide binary lens and thus the c - r lensing provides a good approximation in describing the binary lensing behavior in the region around the caustic . for a very close binary lens ,
the lensing behavior around the center of mass is described by quadrupole lensing @xcite , which also induces an astroidal caustic similar to the c - r lensing caustic . to the first order approximation , the caustic size in units of the angular einstein radius of the binary - lens mass , @xmath7 ,
is related to the separation and the mass ratio between the binary - lens components by @xmath43 for a wide binary and @xmath44 for a close binary .
a c - r lensing event can provide a channel to detect planets in binary systems .
this is because both of the c - r lensing caustic and the central caustic induced by the planet occur in the same region around the primary lens and the size of the planet - induced central caustic can be comparable to the size of the c - r lensing caustic @xcite . in figure
[ fig : one ] , we present the sizes of the c - r lensing caustic and the planet - induced central caustic as a function of the primary - companion separation . we note that a pair of caustics with separations @xmath45 and @xmath46 have the same size to a linear order and thus we present distributions for only wide binaries .
the plot shows that the central caustic induced by planets located in the `` lensing zone '' , although the range varies depending on how the lensing zone is defined . ]
is of considerable size compared to the size of the c - r lensing caustic .
this suggests that the c - r lensing anomaly can be additionally affected by the planetary perturbation , enabling one to identify the presence of a planet in the binary system . in figure [ fig : two ] , we present the variation of the c - r lensing caustic and the lensing light curve affected by the presence of planets with various separations and mass ratios between the planet and the primary of the binary .
see @xcite for more details about the caustic variation .
we note that the c - r lensing channel enables detections of planets not only in close binaries but also in wide binaries . to be dynamically stable
, a planet should be either in a circumbinary ( or p - type ) orbit , where the planet orbits the barycenter of the two stars of a close binary , or in a circumprimary ( s - type ) orbit , where the planet orbits just one star of a wide binary system .
this condition of the planet existence in the binary system matches the lens system configuration of the proposed c - r lensing channel of planet detections .
planets in binary systems can be detected by various methods such as transit ( e.g. , * ? ? ?
* ) , eclipsing binary timing ( e.g. , * ? ? ?
* ) , and radial - velocity methods ( e.g. , * ? ? ?
due to the intrinsic nature of the methods , however , it is difficult to detect planets in circumprimary orbits and thus all planets in binaries detected by these methods , 19 in total , reside in circumbinary orbits . on the other hand ,
c - r lensing events can be produced by both close and wide binary systems and thus the proposed c - r lensing channel provides a unique channel to detect planets both in circumbinary and circumprimary orbits .
in this section , we demonstrate the high efficiency of the proposed c - r lensing channel in detecting planets of binary systems . estimating the efficiency requires to consider various details of observational conditions such as photometric precision and cadence . in order to reflect realistic observational conditions
, we estimate the detection efficiency for an example c - r lensing event that was actually observed by lensing experiments .
the event used for our efficiency estimation is ogle-2015-blg-1319 .
the event was analyzed in detail by @xcite and turned out to be an exemplary c - r lensing event caused either by a close or a wide binary lens .
figure [ fig : three ] shows the light curve of the event reproduced based on the same data sets as those used in the previous analysis .
the model light curve superposed on the data points is obtained from binary - lensing modeling based on one of the 8 degenerate solutions ( `` + + wide '' solution ) presented in @xcite .
the inset in the lower panel shows the source trajectory with respect to the caustic , which has a very characteristic shape of a c - r lensing caustic .
the anomaly caused by the c - r lensing caustic in the peak of the light curve was densely and precisely observed .
this was possible because the event was predicted to be a high - magnification event before it reached the peak based on the light curve obtained from the survey observation conducted by the optical gravitational lens experiment ( ogle : * ? ? ? * ) and the microlensing observations in astrophysics ( moa : * ? ? ?
* ; * ? ? ?
* ) , enabling intensive observations of the peak by follow - up observation groups including the microlensing follow - up network ( @xmath47fun : * ? ? ? * ) and the robonet @xcite groups .
we note that the event was also observed in space using two space telescopes _
spitzer _ and _ swift_. this enabled the determinations of the lens mass and the distance , but we do not use the space - based data because these data are irrelevant to our scientific purpose .
the values of the binary separation and the mass ratio presented in @xcite are @xmath48 for the close binary solution and and @xmath49 for the wide solution . for other lensing parameters ,
see table 1 of @xcite . to show the constraint on the existence of a planet
, we construct an `` exclusion diagram '' , which shows the probability of excluding the existence of a planet as a function of the planet separation and the mass ratio @xcite .
we construct the exclusion diagram following the procedure of @xcite . in this procedure
, we first introduce a planet with the parameters @xmath50 to the binary lens with the parameters @xmath51 . here
@xmath51 represent the separation and the mass ratio between the binary components , while @xmath52 are the separation and the mass ratio between the primary of the binary and the planet .
the angle @xmath53 denotes the orientation angle of the planet with respect to the binary axis connecting the binary lens components .
we use @xmath51 that are determined from the binary - lensing modeling . for a given set of @xmath50 ,
we then search for other parameters that yield the best fit to the observed light curve and compute @xmath54 of the fit .
we repeat this process for many different orientation angles .
then , the probability of excluding a planet for a given @xmath52 is estimated as the fraction of the angles @xmath53 that result in fits with @xmath55 , where @xmath56 is the difference between the triple - lens ( i.e. binary + planet ) and the binary - lens models . as a criteria for the planet detection
, we adopt a threshold value of @xmath57 , which is a generally agreed value for planets detected through the high - magnification channel @xcite .
the probability of excluding a planet corresponds to the probability of detecting the planet , i.e. planet detection efficiency . and the mass ratio @xmath31 between the planet and the primary of the binary lens .
the values marked in the upper @xmath58 axis and the right of @xmath59 axis represent the physical primary - planet separation in au and the mass of the planet in jupiter masses , respectively .
the color coding represents the regions of different efficiencies that are marked in the legend .
the planetary separation is expressed in units of the angular einstein radius corresponding to the binary lens mass . ]
figure [ fig : four ] shows the constructed exclusion diagram which shows the planet detection efficiency as a function of the normalized separation @xmath60 and the mass ratio @xmath31 of the planet . since the microlens parallax of the event ogle-2015-blg-1319 was determined using the combined data taken from the ground and space , the physical sizes of the primary - planet separation and the mass of the planet were determined . by adopting the lens mass of @xmath61 , the angular einstein radius of @xmath62 mas , and the distance to the lens of @xmath63 kpc , that were determined by @xcite , i.e. ( + , + ) wide model , we convert @xmath60 and @xmath31 into the physical sizes of the primary - planet separation in au , i.e. @xmath64 , and the mass of the planet in jupiter masses ( @xmath65 ) , i.e. @xmath66 , and they are presented in the upper @xmath58 axis and the right of the @xmath59 axis , respectively .
lccc jupiter & 1.0 10 au & 0.5 11 au & 0.4 18 au + saturn & 2.0 5 au & 1.0 7 au & 0.9 9 au + neptune & & 2.0 3.5 au & 1.8 4.0 au in table [ table : one ] , we present the ranges of planet detection for 3 different planets with masses corresponding to those of the jupiter , saturn ( @xmath67 ) , and neptune ( @xmath68 ) of the solar system .
it is found that the detection efficiency is greater than @xmath69 for the jupiter- and saturn - mass planets located in the ranges of 1.0 10 au and 2.0 5 au from the host , respectively .
the ranges with @xmath70 probability are 0.4 18 au , 0.9 9 au , and 1.8 4 au for the jupiter- , saturn- , and neptune - mass planets , respecively .
these ranges encompass wide regions around snow lines where giants are believed to form . despite the high efficiency of c - r lensing events in detecting planets of binary systems
, there exists no report of planet detection yet . the reasons for this can be ( 1 ) the rarity of c - r lensing events , ( 2 ) the unoptimized observational strategy for planet detections , and ( 3 ) the rarity of planets in the region of sensitivity .
considering the existence of the already known microlensing planets detected through the repeating - event channel and those detected by other methods combined with the wide region of planet sensitivity of the c - r lensing channel , the reason ( 3 ) is unlikely to be the main reason of nondetection . to check the possibilities ( 1 ) and ( 2 ) , we investigate the lensing events reported by the ogle and moa surveys in 2015 season . from the systematic analyses of all high - magnification events with anomalies near the peaks based on the online data of the surveys
, we find 12 c - r lensing events including ogle-2015-blg-1319 , moa-2015-blg-040/ogle-2015-blg-0318 , ogle-2015-blg-0697/moa-2015-blg-148 , ogle-2015-blg-0797 , ogle-2015-blg-0812 , ogle-2015-blg-0189 moa-2015-blg-085/ogle-2015-blg-0472 , ogle-2015-blg-0313/moa-2015-blg-047 , ogle-2015-blg-0033/moa-2015-blg-017 , moa-2015-blg-047 , ogle-2015-blg-0919 , and ogle-2015-blg-0863 .
this indicates that the rarity of c - r lensing events is not the reason for the nondetection , either .
however , we find that the coverage of the peak regions for all of the c - r lensing events except ogle-2015-blg-1319 was not dense enough to constrain the existence of a planet .
therefore , it is likely that the main reason for the nondetection of planet is due to the unoptimized observational strategy . in other words
, this suggests that planets can be detected through the proposed channel in abundance with an aggressive strategy to densely cover the peak regions of high - magnification events , e.g. , vigilant monitoring of high - magnification events and timely alerts of anomalies followed by prompt and intensive coverage of the anomalies by follow - up observations
we pointed out that c - r lensing events could provide one with an important channel to detect planets in binary systems .
we also pointed out that while other methods could detect planets only in circumbinary orbits the proposed c - r lensing channel could provide a unique channel to detect planets both in circumbinary and circumprimary orbits .
we demonstrated the high sensitivity of the c - r lensing channel to planets in a wide range of planet parameter space by presenting the exclusion diagram for an actually observed c - r lensing event .
we mentioned that an increased number of microlensing planets in binary systems could be detected with an aggressive strategy to densely cover the peak regions of high - magnification events .
the sample of an increased number of microlensing planets in binaries will make it possible to provide important observational constraints that can give shape to the details of the formation scenario which has been restricted to the case of single stars . work by c. han was supported by the creative research initiative program ( 2009 - 0081561 ) of national research foundation of korea .
we acknowledge the high - speed internet service ( kreonet ) provided by korea institute of science and technology information ( kisti ) .
abt , h. a. 1983 , , 21 , 343 albrow , m. d. , beaulieu , j .- p . , caldwell , j. a. r. , et al .
2000 , , 535 , 176 bond , i. a. , abe , f. , dodd , r. j. , et al .
2001 , , 327 , 86 boss , a. p. 2012
, , 419 , 1930 cameron , a. g. w. 1978 , the moon and the planets , 18 , 5 cassan , a. , kubas , d. , beaulieu , j .-
, et al .
2012 , nature , 481 , 167 chang , k. , & refsdal , s. 1979 , nature , 282 , 561 chang , k. , & refsdal , s. 1984 , nature , 132 , 168 chung , s .-
han , c. , park , b .-
, et al .
2005 , , 630 , 535 correia , a. c. m. , udry , s. , mayor , m. , laskar , j. , naef , d. , pepe , f. , queloz , d. , & santos , n. 2005 , , 440 , 751 dank , k. , & heyrovsk , d. 2015 , , 806 , 99 dominik , m. 1999 , , 349 , 108 doyle , l. r. , carter , j. a. , fabrycky , d. c. , et al .
2011 , science , 333 , 1602 erdl h. , & schneider p. 1993
, , 268 , 453 gaudi , b. s. , naber , r. m. , & sackett , p. d. 1998 , , 502 , l33 gaudi , b. s. , & sackett , p. d. 2000 , , 528 , 5 goldreich , p. , & ward , w. 1973 , , 183 , 1051 gould , a. , & loeb , a. 1992 , , 396 , 104 gould , a. , dong , s. , gaudi , b. s. , et al . 2010 , , 720 , 1073 gould , a. , udalski , a. , an , d. , et al . 2006 , , 644 , 37 gould , a. , udalski , a. , shin , i .- g . , et al .
2014 , science , 345 , 46 griest , k. , & safizadeh , n. 1998 , , 500 , 37 hayashi , c. , nakazawa , k. , & nakagawa y. 1985 , in protostars and planets ii , eds .
d. c. black & m. s. matthew ( tucson : univ .
arizona press ) , 1100 kim , s .- l . , lee , c .- u . , park , b .-
, et al .
2015 , journal of the korean astronomical society , 49 , 37 kubas , d. , cassan , a. , dominik , m. , et al .
2008 , , 483 , 317 kuiper , g. p. 1951 proc .
, 37 , 1 lee , d. w. , lee , c .- u . ,
park , b .-
chung , s .-
kim , y .- s . ,
, & han , c. 2008 , , 672 , 623 luhn , j. k. , penny , m. t. , & gaudi , b. s. 2016 , , 827 , 61 paczyski , b. 1986 , , 304 , 1 poleski , r. , skowron , j. , udalski , a. , et al .
2014 , , 795 , 42 pollack , j. b. , hubickyj , o. , bodenheimer , p. , lissauer , j. j. , podolak , m. , & greenzweig , y. 1996 , icarus , 124 , 62 qian , s .- b .
, liu , l. , zhu , l .- y . , dai , z .- b . ,
lajs , e. f. , & baume , g. l. 2010 , , 401 , l34 raghavan , d. , mcalister , h. a. , henry , t. j. , et al .
2010 , , 190 , 1 rhie , s. h. 1997 , , 484 , 63 safronov , v. 1969 , evolution of the protoplanetary cloud and formation of the earth and planets ( moscow : nauka ) shin , i .-
g . , han , c. , choi , j .- y . , hwang , k .- h . , jung , y .- k . , & park , h. 2015 , , 802,108 shvartzvald , y. , li , z. , udalski , a. , et al . 2016 , , submitted sumi , t. , abe , f. , bond , i. a. , et al .
2003 , , 591 , 204 szebehely , v. 1980 , celest .
, 22 , 7 thebault , p. , & haghighipour , n. 2015 , planetary exploration and science : recent results and advances , eds . s. jin , n. haghighipour , w - h . ip ( berlin : springer ) , 309 tsapras , y. , street , r. , horne , k. , et al .
2009 , astron .
nachr . , 330 , 4 udalski , a. 2003 , acta astron . , 53 , 291 udalski , a. , szymaski , m. , kaluny , j. , kubiak , m. , & mateo , m. 1992 , acta astron . , 42 , 253 | chang - refsdal ( c - r ) lensing , which refers to the gravitational lensing of a point mass perturbed by a constant external shear , provides a good approximation in describing lensing behaviors of either a very wide or a very close binary lens .
c - r lensing events , which are identified by short - term anomalies near the peak of a high - magnification lensing light curves , are routinely detected from lensing surveys , but not much attention is paid to them . in this paper , we point out that c - r lensing events provide an important channel to detect planets in binaries , both in close and wide binary systems . detecting planets through the c - r lensing event channel is possible because the planet - induced perturbation occurs in the same region of the c - r lensing - induced anomaly and thus the existence of the planet can be identified by the additional deviation in the central perturbation . by presenting the analysis of the actually observed c - r lensing event ogle-2015-blg-1319
, we demonstrate that dense and high - precision coverage of a c - r lensing - induced perturbation can provide a strong constraint on the existence of a planet in the wide range of the planet parameters
. the sample of an increased number of microlensing planets in binary systems will provide important observational constraints in giving shape to the details of the planet formation scenario which has been restricted to the case of single stars . |
while a static magnetic field can not change the kinetic energy of charged particles , it affects the direction of their motion due to a lorentz force .
this property is widely used in physics and technology .
for example , magnetic fields play a crucial role in cyclotrons and particle accelerators , where they keep beams of charged particles on correct trajectories .
another application is the mass spectrometry , which utilizes magnetic fields in order to sort charged particles based on their mass - to - charge ratio .
magnetic fields can be also used as lenses for deflecting and focusing beams of charged particles .
such lenses are utilized , for example , in cathode ray tubes and electron microscopes @xcite .
the history of geometric electron optics started from a paper by hans busch in 1926 @xcite , where he suggested that the magnetic field of a short coil can act as a converging lens for electrons .
this proposal was instrumental for the construction of the first electron microscope by knoll and ruska @xcite . owing to the fact that the wavelengths of electrons can be much smaller than the wavelengths of the visible light
, the resolution of the electron microscope quickly surpassed that of the conventional optical microscope .
modern transmission electron microscopes with aberrations correctors achieve a few orders of magnitude better resolution than the usual light microscopes and , therefore , are used to investigate the structure of metals , crystals , large molecules , as well as biological specimens including microorganisms and cells . in geometrical particle optics ,
the motion of charged particles in macroscopic electromagnetic fields is described analogously to the propagation of the light rays .
mathematically , the possibility of such a description stems from the eikonal or wentzel kramers brillouin ( wkb ) approximation @xcite .
the eikonal approximation is an appropriate and efficient means to solve wave equations in the case when the de broglie wavelengths of particles are small compared to the characteristic inhomogeneities of the system or external fields . in mathematical physics ,
the wkb approximation is a well - established method for finding approximate solutions to linear differential equations with coefficients varying slowly in space @xcite .
recent experimental discovery of dirac and weyl materials made possible a condensed - matter realization of the systems whose low - energy quasiparticles are massless dirac or weyl fermions ( for reviews , see refs .
the corresponding quasiparticles have a well - defined chirality in the vicinity of dirac points or weyl nodes .
in this connection it should be noted that a dirac point in the brillouin zone could be viewed as an overlap of two weyl nodes of opposite chirality . according to the celebrated nielsen - ninomiya theorem @xcite , weyl nodes in condensed matter materials , in fact ,
always come in pairs of opposite chirality .
in general , however , the nodes in each pair need not coincide .
experimentally , the first three - dimensional dirac materials , i.e. , na@xmath0bi and cd@xmath0as@xmath1 , were discovered only a few years ago @xcite . since then a large number of weyl materials ( i.e. , @xmath2 , @xmath3 , @xmath4 , @xmath5 , @xmath6 , @xmath7 ) were discovered as well @xcite .
these discoveries opened a new chapter in studies of numerous effects associated with quantum anomalies and chirality by using simple table - top experiments , rather than accelerator techniques of high - energy physics .
in addition to quantum field theoretical methods , there exist a number of very powerful semiclassical techniques for studying condensed matter systems such as weyl and dirac materials in external electromagnetic fields .
one of them is the kinetic theory . because of the nontrivial topological properties of weyl fermions @xcite , the corresponding kinetic theory should account for the berry curvature effects @xcite .
amazingly , the resulting chiral kinetic theory @xcite can reproduce even the effects due to the quantum chiral anomaly @xcite .
note that the latter implies the nonconservation of the chiral charge in the presence of parallel background electric and magnetic fields .
the dirac and weyl materials make also possible the investigations of novel quantum phenomena that can not exist in high - energy physics . in particular
, a number of such phenomena can be associated with the quasiparticle response to background pseudoelectromagnetic ( axial ) fields . unlike the ordinary electromagnetic fields @xmath8 and @xmath9 , their pseudoelectromagnetic counterparts @xmath10 and @xmath11 couple to the left - handed and right - handed particles with opposite signs . indeed , as shown in refs .
@xcite , such fields can be induced by mechanical strains in dirac and weyl materials .
a pseudoelectric field @xmath12 , for instance , can be created by dynamically stretching or compressing the sample .
a nonzero pseudomagnetic field @xmath13 is generated , e.g. , by applying a static torsion @xcite or bending the sample @xcite .
a typical magnitude of the pseudomagnetic field @xmath14 is estimated to be about @xmath15 in the former case and about @xmath16 in the latter case .
for the purposes of this study , it will be sufficient to consider only static deformations so that @xmath17 .
in addition , we will also assume that the ordinary electric field is also absent @xmath18 .
in this study we suggest that the strain - induced pseudomagnetic fields can be utilized for creating pseudomagnetic lenses for deflecting and focusing beams of weyl quasiparticles , depending on the particle chiralities .
a general experimental setup that allows a maximum control of chiral beams is given by a combination of the magnetic and pseudomagnetic lenses as shown schematically in fig .
[ fig : illustration ] .
the system consists of a thin long crystal ( wire ) of a weyl material placed inside a solenoid . in this case ,
magnetic and pseudomagnetic fields are directed along the @xmath19 axis and are present in the region @xmath20 .
( generically , the sample itself can be longer than @xmath21 . )
the magnetic field is generated by an electric current in the solenoid and the pseudomagnetic one is produced by the torsion of the crystal .
since the characteristic scales of spatial variations of the background magnetic and pseudomagnetic fields in the ( pseudo-)magnetic lens are much larger than the de broglie wavelengths of weyl quasiparticles , it is justified to use the eikonal approximation for the description of their motion . because of a nontrivial topology of chiral particles , we will pay a special attention to the berry curvature effects on the corresponding chiral rays .
is produced by an electric current @xmath22 in the solenoid , the pseudomagnetic field @xmath11 is created by twisting the crystal of a weyl material.,scaledwidth=50.0% ] the paper is organized as follows . in sec .
[ sec : eikonal ] we describe the eikonal approximation in application to weyl quasiparticles .
the properties of magnetic and pseudomagnetic lenses for weyl quasiparticles are considered in sec .
[ sec : lens ] .
the discussion and summary of the main results are given in sec .
[ sec : summary - discussions ] .
in this section , we formulate the eikonal approximation for the motion of weyl quasiparticles in the presence of both magnetic and pseudomagnetic fields . according to ref .
@xcite , the berry curvature not only modifies the semiclassical equations of quasiparticle motion in background fields @xcite , but also changes their dispersion relations .
thus , we will begin our analysis by presenting the corresponding relations for chiral quasiparticles . in the framework of the chiral kinetic theory @xcite , it is straightforward to include corrections linear in the background field @xmath23 , where @xmath9 and @xmath13 denote the ordinary magnetic and pseudomagentic fields , respectively , and @xmath24 is chirality of the left- ( @xmath25 ) and right - handed ( @xmath26 ) quasiparticles .
however , by noting that the eikonal approximation for the conventional magnetic lens requires the inclusion of quadratic terms in the magnetic field @xcite , we need the dispersion relations for the weyl quasiparticles that are also valid to the second order in @xmath27 .
the energy dispersion for a general band structure with nonzero berry curvature was obtained in ref .
@xcite . the corresponding explicit expression for weyl quasiparticles
was derived by the present authors in ref .
@xcite . for the quasiparticles of positive energy ( electrons ) , the corresponding relation reads @xmath28 -\frac{(\mathbf{b}_{\lambda}\cdot[\mathbf{e}_{\lambda}\times\mathbf{p}])}{p}\right\ } , \label{second - energy-2}\end{aligned}\ ] ] where @xmath29 is the fermi velocity , @xmath30 is the speed of light , @xmath31 is the momentum of quasiparticles , and @xmath32 is the electron charge .
note that the second and third terms in eq .
( [ second - energy-2 ] ) describe corrections due to the berry curvature . using the standard eikonal approximation @xcite for charged particles ,
we derive the equation for the abbreviated action @xmath33 of weyl quasiparticles by making the following replacement in the dispersion relation ( [ second - energy-2 ] ) : @xmath34 where @xmath35 is an effective vector potential that describes the background field @xmath27 .
note that the abbreviated action @xmath36 is related to the full action via @xmath37 , where @xmath38 denotes the quasiparticle energy and @xmath39 is time . for a constant @xmath27 in the @xmath19 direction , the effective vector potential in the symmetric gauge
is given by @xmath40 then , by substituting eq .
( [ lens - eikonal - p - replace ] ) into eq .
( [ second - energy-2 ] ) , we obtain the following eikonal equation for the abbreviated action @xmath36 of weyl quasiparticles with energy @xmath38 : @xmath41^{3/2 } } -\frac{e^2 \hbar^2v_fb_{\lambda}^2}{16c^2 } \frac{(\nabla_zs_0)^2}{\left[(\bm{\nabla}s_0)^2+\frac{e^2}{4c^2}r_{\perp}^2b_{\lambda}^2\right]^{5/2 } } , \label{lens - eikonal - epsilon}\end{aligned}\ ] ] where , in view of the symmetry in the problem , we assumed that @xmath42 depends on @xmath43 and @xmath44 , and used @xmath45 in order to obtain an analytical solution to eq .
( [ lens - eikonal - epsilon ] ) , we will use the paraxial approximation , which is well justified when the quasiparticles propagate sufficiently close to the optical axis .
in other words , we will assume that @xmath46 is small and expand the solution in powers of @xmath46 . in the case
@xmath47 , the action should describe a free quasiparticle moving with momentum @xmath48 , i.e. , @xmath49 in the case of a nonzero @xmath27 , we will seek @xmath36 in a similar form , i.e. , @xmath50 where constant @xmath51 and function @xmath52 are to be determined . indeed , by matching the actions at the boundaries of the ( pseudo-)magnetic lens , one can easily show that the eikonal of a quasiparticle moving in ( pseudo-)magnetic fields should also contain only even powers of @xmath46 . it is worth mentioning that the form of the abbreviated action in eq .
( [ lens - eikonal - s0 ] ) is standard in the paraxial optics @xcite . by moving the second term on the right - hand side of eq .
( [ lens - eikonal - epsilon ] ) to the left - hand side and squaring both sides of the equation , we find @xmath53 + \frac{e^2v_f^2}{4c^2}r_{\perp}^2b_{\lambda}^2 , \label{lens - eikonal - epsilon-1}\end{aligned}\ ] ] where terms of order @xmath54 were dropped .
then , by keeping the terms up to quadratic order in @xmath55 and making use of the following relations : @xmath56 , \label{lens - eikonal - eqs - ee}\end{aligned}\ ] ] we rewrite eq .
( [ lens - eikonal - epsilon-1 ] ) in the form @xmath57 + \frac{2\lambda b_{\lambda}}{b^ { * } } \left\{c^3 + r_{\perp}^2c\left[\frac{3}{2}c a^{\prime}(z)+(a(z))^2\right]\right\ } \nonumber\\ & & + \frac{1}{2}\left(\frac{b_{\lambda}}{b^{*}}\right)^2 \left\{c^2 + r_{\perp}^2\left [ ca^{\prime}(z ) -2(a(z))^2\right]\right\ } -\frac{r_{\perp}^2 eb_{\lambda}^2}{2c \hbar b^ { * } } c^4=0 .
\label{lens - eikonal - epsilon-2}\end{aligned}\ ] ] note that in the last equation , we introduced the reference value of the magnetic field @xmath58 associated with the quasiparticle energy @xmath38 . as is easy to check , this field
is defined so that the corresponding magnetic length @xmath59 is comparable to the de broglie wavelength of the weyl quasiparticle @xmath60 . in fact , they are related as follows : @xmath61 . it should be noted that the validity of the quasiparticles energy ( [ second - energy-2 ] ) is restricted only to the case of sufficiently weak magnetic and pseudomagnetic fields , i.e. , @xmath62 . at the zeroth order in @xmath55 , eq .
( [ lens - eikonal - epsilon-2 ] ) reduces to the following equation : @xmath63 this equation has four nontrivial solutions , i.e. , @xmath64 where we used an expansion in powers of the small parameter @xmath65 . by taking into account
that @xmath51 should be equal to one in the limit of vanishing fields , @xmath47 , we conclude that the physical solution is given by @xmath66 .
now , by equating the terms quadratic in @xmath46 in eq .
( [ lens - eikonal - epsilon-2 ] ) , we obtain the following first - order differential equation for the function @xmath52 : @xmath67 where @xmath68 \left [ c^2-\frac{3}{2 } c^4 + \frac{\lambda b_{\lambda}}{b^ { * } } c-\frac{1}{2 } \left(\frac{b_{\lambda}}{b^{*}}\right)^2 \right]^{-1 } \simeq 1+\frac{5}{4 } \left(\frac{b_{\lambda}}{b^{*}}\right)^2 , \\
a_2 ^ 2&=&\frac { eb_{\lambda}^2}{4c \hbar b^ { * } } c^4 \left[\frac{3}{2 } c^4 -c^2 -\frac{\lambda b_{\lambda}}{b^ { * } } c+\frac{1}{2 } \left(\frac{b_{\lambda}}{b^{*}}\right)^2 \right]^{-1 } \simeq \frac { eb_{\lambda}^2}{2c \hbar b^ { * } } \left[1 - 2\lambda \frac{b_{\lambda}}{b^ { * } } \right ] .
\label{lens - eikonal - epsilon - r-1-a2}\end{aligned}\ ] ] in the next section we will use eqs .
( [ lens - eikonal - epsilon - r-1-comp ] ) , ( [ lens - eikonal - epsilon - r-1-a1 ] ) , and ( [ lens - eikonal - epsilon - r-1-a2 ] ) to analyze the motion of weyl quasiparticles in magnetic and pseudomagnetic fields .
let us begin the analysis from the simplest case of uniform magnetic and pseudomagnetic fields , @xmath69 . in the regions
@xmath70 and @xmath71 , the fields are absent .
therefore , @xmath72 and @xmath73 there and the solutions to eq .
( [ lens - eikonal - epsilon - r-1-comp ] ) are given by @xmath74 where @xmath75 and @xmath76 are the integration constants that will be fixed by the boundary conditions .
on the other hand , by solving eq .
( [ lens - eikonal - epsilon - r-1-comp ] ) in the region with nonzero background fields , i.e. , @xmath20 , and using eqs .
( [ lens - eikonal - epsilon - r-1-a1 ] ) and ( [ lens - eikonal - epsilon - r-1-a2 ] ) , we obtain the following solution : @xmath77 where @xmath78 is another integration constant that should be also determined by matching the solutions for @xmath52 at @xmath79 and @xmath80 , i.e. , @xmath81 finally , after excluding @xmath78 from these equations , we obtain the following lens equation relating @xmath75 and @xmath76 for the quasiparticles of a given chirality @xmath82 : @xmath83 where @xmath75 and @xmath76 can be interpreted as the coordinates of the quasiparticle source and its image , respectively .
indeed , we see that when the source is placed at the left focal point , i.e. , @xmath84 , the position of the image @xmath76 goes to infinity . similarly , when @xmath85 , the location of the image is near the right focal point , i.e. , @xmath86 .
therefore , @xmath87 and @xmath88 are the locations of the principal foci and @xmath89 is the principal focal length . in the case under consideration , we find that @xmath90 and @xmath91 by making use of the small - field asymptotes for the functions @xmath92 and @xmath93 given by eqs .
( [ lens - eikonal - epsilon - r-1-a1 ] ) and ( [ lens - eikonal - epsilon - r-1-a2 ] ) , we then obtain the following explicit expressions for functions @xmath94 and @xmath89 : @xmath95 these analytical expressions are the key characteristics of the ( pseudo-)magnetic lens and are the main results of this paper .
when the paraxial approximation is justified , these results should be valid for arbitrary weyl and dirac materials . as is easy to see from eq .
( [ lens - eikonal - lens - f - app-2 ] ) , the focal lengths for the quasiparticles of opposite chiralities are different in general .
this remains true even in the limit of the vanishing pseudomagnetic field ( i.e. , @xmath96 but @xmath97 ) .
in such a case a small quantitative difference between @xmath98 and @xmath99 is connected with the berry curvature effects .
this is in contrast to the case of the vanishing magnetic field ( @xmath100 but @xmath101 ) , when the focal lengths for the quasiparticles of opposite chiralities are exactly the same in the formalism at hand . in order to get a better insight into the parametric dependencies of the principal focal length on the strength of the ( pseudo-)magnetic field
, we will use the numerical value of the fermi velocity for cd@xmath0as@xmath1 @xcite , i.e. , @xmath102 . in this case , the characteristic magnetic field can be estimated as @xmath103 the dependence of the focal length @xmath89 on the pseudomagnetic field @xmath14 at two fixed values of the magnetic field @xmath100 and @xmath104 are shown in the left and right panels of fig .
[ fig : focal - f - b5-b ] for @xmath105 , respectively . as expected , in the presence of both magnetic
_ and _ pseudomagnetic fields , the focal lengths are different for the quasiparticles of opposite chirality . in practice , this property may provide an efficient means for a space separation of chiral charges inside a weyl material . as is easy to see from the analytical expression ( [ lens - eikonal - lens - f - app-2 ] ) , as well as from the numerical results in fig .
[ fig : focal - f - b5-b ] , the dependence of the focal length on the ( pseudo-)magnetic field is quasi - periodic . also , focal length is formally divergent at the following discrete value of the background field : @xmath106 , where @xmath107 .
the mathematical reason for the divergencies is clear from the analytical result in eq .
( [ lens - eikonal - lens - f - app-2 ] ) , which has the sine function in the denominator .
it should be noted , however , that the paraxial approximation for the eikonal equation breaks down in the vicinity of such divergencies .
this is clear from the fact that the expansion in powers of @xmath108 in eq .
( [ lens - eikonal - s0 ] ) becomes unreliable when @xmath52 is too large .
in such a situation , a geometrical optics approach fails and one needs to use exact , rather than approximate solutions to the wave equation . given by eq .
( [ lens - eikonal - lens - f - app-2 ] ) for quasiparticles of chirality @xmath26 ( red solid line ) and @xmath25 ( blue dashed line ) as a function of pseudomagnetic field .
while the left panel corresponds to zero magnetic field @xmath100 , the right one is plotted for @xmath109 .
we set @xmath105 and @xmath110.,title="fig:",scaledwidth=45.0% ] given by eq .
( [ lens - eikonal - lens - f - app-2 ] ) for quasiparticles of chirality @xmath26 ( red solid line ) and @xmath25 ( blue dashed line ) as a function of pseudomagnetic field .
while the left panel corresponds to zero magnetic field @xmath100 , the right one is plotted for @xmath109 .
we set @xmath105 and @xmath110.,title="fig:",scaledwidth=45.0% ] in order to illuminate the effects of the berry curvature , it is instructive to compare the focal length in eq .
( [ lens - eikonal - lens - f - app-2 ] ) with its counterpart when such effects are neglected , i.e. , @xmath111 the corresponding relative difference @xmath112 is plotted in fig .
[ fig : focal - f - compare ] . as we see , the effects of the berry curvature quantified by @xmath113 are about @xmath114 orders of magnitude smaller than the focal lengths .
we checked , however , that the corresponding effects may become noticeable at sufficiently large magnetic and pseudomagnetic fields or near the divergencies . between the focal length @xmath115 given by eq .
( [ lens - eikonal - lens - f - app - landau ] ) and the focal length @xmath89 with the effects of the berry curvature ( [ lens - eikonal - lens - f - app-2 ] ) as a function of pseudomagnetic field .
the red solid line denotes the right - handed quasiparticles and the blue dashed one corresponds to the left - handed ones .
the left panel corresponds to zero magnetic field @xmath100 , while @xmath109 in the right one .
we set @xmath105 and @xmath110.,title="fig:",scaledwidth=45.0% ] between the focal length @xmath115 given by eq .
( [ lens - eikonal - lens - f - app - landau ] ) and the focal length @xmath89 with the effects of the berry curvature ( [ lens - eikonal - lens - f - app-2 ] ) as a function of pseudomagnetic field .
the red solid line denotes the right - handed quasiparticles and the blue dashed one corresponds to the left - handed ones .
the left panel corresponds to zero magnetic field @xmath100 , while @xmath109 in the right one .
we set @xmath105 and @xmath110.,title="fig:",scaledwidth=45.0% ] the model situation with a constant ( pseudo-)magnetic field @xmath116 inside a quasiparticle lens considered in the previous subsection is not very realistic .
indeed , in the case of real solenoids and torsion - induced strains there are always fringing fields at the ends . in order to get a better insight into the underlying physics , in this subsection we study the case of a nonuniform ( pseudo-)magnetic field .
in particular , we consider the effective field with the following spatial profile : @xmath117 such a configuration mimics well the fringing fields at the ends of the ( pseudo-)magnetic lens and , at the same time , allows one to obtain an analytical solution in the paraxial approximation . because of the the unit step functions @xmath118 in eq .
( [ inhom - b - profile ] ) , the background field @xmath119 is still effectively confined in the region where the background ( pseudo-)magnetic field is nonzero @xmath20 . by limiting ourselves to the case of sufficiently weak ( pseudo-)magnetic fields , we can neglect the effects of the berry curvature in the eikonal approximation .
this is supported by our findings in the previous subsection , showing that the main contribution to the spatial separation of weyl quasiparticles with different chirality comes from the terms of the leading - order in the ( pseudo-)magnetic field .
thus , by dropping the terms due to the berry curvature in the quasiparticle energy , we have @xmath120 instead of eq .
( [ second - energy-2 ] ) . in this case
, the analog of the differential equation ( [ lens - eikonal - epsilon - r-1-comp ] ) reads @xmath121 it has the following analytical solution inside the solenoid : @xmath122 } , \label{inhom - b - eikonal}\end{aligned}\ ] ] where @xmath123 is an integration constant , and @xmath124 by matching the solutions outside the solenoid ( [ lens - eikonal - a2-sol-1 ] ) with that in eq .
( [ inhom - b - eikonal ] ) at @xmath79 and @xmath80 , we obtain the following lens equation : @xmath125 where @xmath126 , @xmath127 } + l \sin{\left[2f_{\lambda}\ , \mbox{arccot}\left(\frac{2\xi}{l}\right)\right]}}{2\left(4\xi^2f_{\lambda}^2-l^2\right ) \sin{\left[2f_{\lambda}\ , \mbox{arccot}\left(\frac{2\xi}{l}\right)\right]-8l\xi f_{\lambda } \cos{\left[2f_{\lambda}\ , \mbox{arccot}\left(\frac{2\xi}{l}\right)\right ] } } } , \label{inhom - b - g}\end{aligned}\ ] ] and @xmath128 } -4l\xi f_{\lambda } \cos{\left[2f_{\lambda}\ , \mbox{arccot}\left(\frac{2\xi}{l}\right)\right]}}. \label{inhom - b - f}\end{aligned}\ ] ] as is easy to check , @xmath129 , where @xmath115 is the solution in the case of the homogeneous field given by eq .
( [ lens - eikonal - lens - f - app - landau ] ) . the dependence of the focal length @xmath130 on the pseudomagnetic field strength @xmath14 is presented in fig .
[ fig : focal - inhom - b5-b ] , where the left and right panels correspond to @xmath131 and @xmath132 , respectively .
as we see from this figure , the effect of the inhomogeneity is weak at small values of the pseudomagnetic field , but it increases with @xmath14 .
the case of moderately large @xmath133 , i.e. , @xmath134 , depicted in the right panel is almost indistinguishable from the case of uniform fields , as it should be for the weakly varying ( pseudo-)magnetic field . on the other hand , at small values of @xmath133 ( i.e. , @xmath135 ) , the dependence of @xmath130 on the ( pseudo-)magnetic field is very different .
qualitatively , with decreasing @xmath133 , the period of the focal length oscillations increases .
furthermore , we note that for @xmath136 , we can still use eq .
( [ lens - eikonal - lens - f - app - landau ] ) , but with the following replacement of the lens size : @xmath137 where we used the explicit form of @xmath138 in eq .
( [ inhom - b - profile ] ) in order to perform the integration . given by eq .
( [ inhom - b - f ] ) for quasiparticles of both chiralities at @xmath100 ( red solid line ) as well as for @xmath26 ( blue dashed lines ) and @xmath25 ( green dotted lines ) at @xmath139 . while the left panel represents the results at @xmath131 , the right one corresponds to @xmath132 .
we set @xmath105 and @xmath110.,title="fig:",scaledwidth=45.0% ] given by eq .
( [ inhom - b - f ] ) for quasiparticles of both chiralities at @xmath100 ( red solid line ) as well as for @xmath26 ( blue dashed lines ) and @xmath25 ( green dotted lines ) at @xmath139 . while the left panel represents the results at @xmath131 , the right one corresponds to @xmath132 .
we set @xmath105 and @xmath110.,title="fig:",scaledwidth=45.0% ]
in this study we investigated the conceptual possibility of pseudomagnetic lenses that can be used to focus the beams of chiral quasiparticles in weyl materials .
we found that the maximum flexibility in controlling the beams of weyl quasiparticles is achieved when one combines the magnetic and pseudomagnetic lenses , as shown schematically in fig .
[ fig : illustration ] .
indeed , the presence of both magnetic _ and _ pseudomagnetic fields allows one to achieve different magnitudes of the effective magnetic fields exerted on the left- and right - handed quasiparticles .
this provides enough control to manipulate the focal lengths @xmath89 independently for each chirality . in order to obtain the geometric optics description for the beams of weyl quasiparticles in background magnetic and pseudomagentic fields
, we used the conventional eikonal approximation for solving the semiclassical wave equation .
such an approximation is well justified for weyl materials when the de broglie wavelengths of quasiparticles in the beam are small compared to the spatial scale of the background - field inhomogeneities , as well as the sample size . in this study
we argued that , because of the nontrivial topology of weyl fermions , the corresponding eikonal equation is affected by the berry curvature in a nontrivial way .
the corrections due to berry curvature effects are rather small when the magnetic and pseudomagnetic fields are much weaker than the reference value @xmath140 , defined in eq .
( [ lens - eikonal - bcrit ] ) , which is the scale set by the quasiparticle energy .
this regime seems to be relevant for applications in dirac semimetals such as cd@xmath0as@xmath1 when the effective background field is weaker than about @xmath141 . for stronger fields , or for other types of dirac / weyl materials
, it is possible that the berry curvature could lead to substantial corrections in the quantitative description of the ( pseudo-)magnetic lenses .
in fact , these corrections may lead to a spatial separation of the beams of chiral quasiparticles even by the ordinary magnetic field alone .
in such a case , a more refined analysis using quantum , rather than semiclassical equations would be required . in the framework of a simplified model with a spatially uniform ( pseudo-)magnetic field inside the lens , we derived the explicit general expressions for the principal foci and the focal length .
the corresponding results include the effects due to the berry curvature and are reliable only in the case of not too strong background fields .
in addition , in order to address the role of inhomogeneous fields , we also studied a simple model with a nonuniform pseudomagnetic field , see eq .
( [ inhom - b - profile ] ) . in this case
, we found that the effect of a highly nonuniform field is similar to that of a uniform one , but with the effective length of the lens @xmath142 given by eq .
( [ inhom - b - leff ] ) .
it is worth noting that in this study we solved the eikonal equation for weyl quasiparticles by using the paraxial approximation .
clearly , this is adequate only when the beam of chiral quasiparticles remains close to the axis of the ( pseudo-)magnetic lens . in practice , of course , this condition may break down and , then , one would have to reanalyze the problem by using numerical methods .
in such a regime , optical aberrations will appear and further complicate the situation . while all these issues may be of real importance for making pseudomagnetic lenses in practice , they are beyond the scope of the conceptual study presented here
. it is intriguing to suggest that the pseudomagnetic lenses could allow for new and powerful ways of controlling and manipulating the chiral beams of quasiparticles inside dirac and weyl materials .
for example , they may open an experimental possibility to control the spatial distributions of the electric current densities due to weyl quasiparticles , depending on their chirality . by focusing the beams of the left- and right - handed quasiparticles in different spatial regions
, one could achieve a chiral distillation " and/or steady states of matter with a nonzero chiral asymmetry .
this could allow various new applications that utilize the possibility of a _ chiral spectrometer _ , in which spatial separation of chiral charges may be detected via the local probes .
the work of e.v.g . was partially supported by the program of fundamental research of the physics and astronomy division of the nas of ukraine .
the work of v.a.m . and p.o.s .
was supported by the natural sciences and engineering research council of canada .
the work of i.a.s . was supported by the u.s .
national science foundation under grant no .
phy-1404232 .
o. vafek and a. vishwanath , ann .
condensed matter phys . * 5 * , 83 ( 2014 ) .
a. a. burkov , j. phys . : condens .
matter * 27 * , 113201 ( 2015 ) .
h. b. nielsen and m. ninomiya , nucl .
b * 185 * , 20 ( 1981 ) [ erratum : nucl .
b * 195 * , 541 ( 1982 ) ] ; nucl .
b * 193 * , 173 ( 1981 ) .
s. borisenko , q. gibson , d. evtushinsky , v. zabolotnyy , b. buchner , and r. j. cava , phys .
rev . lett . * 113 * , 027603 ( 2014 ) .
m. neupane , s .- y .
xu , r. sankar , n. alidoust , g. bian , c. liu , i. belopolski , t .-
chang , h .-
jeng , h. lin , a. bansil , f. chou , and m. z. hasan , nature commun . * 5 * , 3786 ( 2014 )
. z. k. liu , b. zhou , y. zhang , z. j. wang , h. m. weng , d. prabhakaran , s .- k .
mo , z. x. shen , z. fang , x. dai , z. hussain , and y. l. chen , science * 343 * , 864 ( 2014 ) . c. zhang , z. yuan , s. xu , z. lin , b. tong , m. z. hasan , j. wang , c. zhang , and s. jia , phys .
b * 95 * , 085202 ( 2017 ) . s .- y .
xu , i. belopolski , n. alidoust , m. neupane , g. bian , c. zhang , r. sankar , g. chang , z. yuan , c .- c .
huang , h. zheng , j. ma , d. s. sanchez , b. wang , a. bansil , f. chou , p. p. shibayev , h. lin , s. jia , and m. z. hasan , science * 349 * , 613 ( 2015 ) .
b. q. lv , h. m. weng , b. b. fu , x. p. wang , h. miao , j. ma , p. richard , x. c. huang , l. x. zhao , g. f. chen , z. fang , x. dai , t. qian , and h. ding , phys .
x * 5 * , 031013 ( 2015 ) .
x. huang , l. zhao , y. long , p. wang , d. chen , z. yang , h. liang , m. xue , h. weng , z. fang , x. dai , and g. chen , phys .
x * 5 * , 031023 ( 2015 ) .
i. belopolski , s .- y .
xu , y. ishida , x. pan , p.
yu , d. s. sanchez , m. neupane , n. alidoust , g. chang , t .-
chang , y. wu , g. bian , h. zheng , s .-
huang , c .- c .
lee , d. mou , l. huang , y. song , b. wang , g. wang , y .- w .
yeh , n. yao , j. rault , p. lefevre , f. bertran , h .-
jeng , t. kondo , a. kaminski , h. lin , z. liu , f. song , s. shin , and m. z. hasan , arxiv:1512.09099 .
s. borisenko , d. evtushinsky , q. gibson , a. yaresko , t. kim , m. n. ali , b. buechner , m. hoesch , and r. j. cava , arxiv:1507.04847 . g. sundaram and q. niu , phys .
b * 59 * , 14915 ( 1999 ) .
d. xiao , j. shi , and q. niu , phys .
lett . * 95 * , 137204 ( 2005 ) [ erratum : phys .
lett . * 95 * , 169903 ( 2005 ) ] . c. duval , z. horvath , p. a. horvathy , l. martina , and p. stichel , mod .
b * 20 * , 373 ( 2006 ) .
y. gao , s. a. yang , and q. niu , phys .
b * 91 * , 214405 ( 2015 )
. m. v. berry , proc .
a * 392 * , 45 ( 1984 ) .
d. xiao , m. c. chang , and q. niu , rev .
phys . * 82 * , 1959 ( 2010 ) .
d. t. son and n. yamamoto , phys .
. lett . * 109 * , 181602 ( 2012 ) .
m. a. stephanov and y. yin , phys .
* 109 * , 162001 ( 2012 ) .
d. t. son and n. yamamoto , phys .
d * 87 * , 085016 ( 2013 ) .
s. l. adler , phys . rev . * 177 * , 2426 ( 1969 ) ; j. s. bell and r. jackiw , nuovo cim .
a * 60 * , 47 ( 1969 ) .
m. a. zubkov , annals phys .
* 360 * , 655 ( 2015 ) . a.
cortijo , y. ferreiros , k. landsteiner , and m. a. h. vozmediano , phys .
115 * , 177202 ( 2015 ) .
a. cortijo , d. kharzeev , k. landsteiner , and m. a. h. vozmediano , phys .
b * 94 * , 241405 ( 2016 ) .
a. g. grushin , j. w. f. venderbos , a. vishwanath , and r. ilan , phys . rev .
x * 6 * , 041046 ( 2016 ) .
d. i. pikulin , a. chen , and m. franz , phys .
x * 6 * , 041021 ( 2016 ) .
t. liu , d. i. pikulin , and m. franz , phys .
b * 95 * , 041201 ( 2017 ) .
e. v. gorbar , v. a. miransky , i. a. shovkovy and p. o. sukhachov , arxiv:1702.02950 . | it is proposed that strain - induced pseudomagnetic fields in weyl materials could be used as chirality sensitive lenses for beams of weyl quasiparticles .
the study of the ( pseudo-)magnetic lenses is performed by using the eikonal approximation for describing the weyl quasiparticles propagation in magnetic and strain - induced pseudomagnetic fields .
analytical expressions for the locations of the principal foci and the focal length are obtained in the paraxial approximation in the models with uniform as well as nonuniform effective magnetic fields inside the lens .
the results show that the left- and right - handed quasiparticles can be focused at different spatial locations when both magnetic and pseudomagnetic fields are applied .
it is suggested that the use of the magnetic and pseudomagnetic lenses could open new ways of producing and manipulating beams of chiral weyl quasiparticles . |
substantial populations of intergalactic qso absorption - line systems have been detected in highly - ionized oxygen ( ) with the hubble space telescope ( hst ) and the _ far ultraviolet spectroscopic explorer _ ( _ fuse _ ) . because they are probably hot ( @xmath8 k )
, these absorbers may represent a portion of the `` missing baryons '' that are predicted ( cen & ostriker 1999 ; dav et al .
1999 ) to arise at low redshift from continuous structure formation in the warm - hot intergalactic medium ( whim ) .
early surveys of these absorbers with hst ( tripp , savage , & jenkins 2000 ; savage et al .
2002 , tripp 2005 ) found unexpectedly high number densities per unit redshift , @xmath0/@xmath1 @xmath9 , down to 30 - 50 m sensitivity .
the recent _ fuse _ study by danforth & shull ( 2005 , hereafter ds05 ) has refined these estimates of @xmath0/@xmath1 and measured their column density distribution , @xmath0/@xmath10 .
ds05 assumed uniform 0.1 solar metallicity ( [ o / h ] @xmath11 ) and @xmath12 k to estimate the total contribution of the absorbers to the baryon density , @xmath13 , where @xmath14 ( spergel et al . 2003 ) km s@xmath15 mpc@xmath15 ) @xmath16 . ] .
the _ fuse _ data can not measure the gas metallicity precisely but they allow for estimates of typical values .
ds05 detect 40 absorbers which span 2.5 dex in @xmath17 ( ) but only 1.5 dex in @xmath17 ( ) . by combining photoionization and collisional ionization equilibrium ( cie ) models with a theoretical relation between igm density and @xmath18 , ds05 derive a typical metallicity of @xmath19 .
we derive a conservative lower limit [ o / h ] @xmath20 from a pure cie model that matches the broad locus of points in figure 1 of ds05 , assuming the maximum fractional ionization , @xmath21 , of at @xmath22 k. however , according to equation 6 of ds05 this metallicity limit implies @xmath23 , close to the total cosmic baryon density . because a major portion of the whim is thought to lie in hotter gas accessible only to x - ray telescopes ( fang et al .
2002 ; mathur , weinberg , & chen 2003 ; mckernan et al .
2003 ; nicastro et al .
2005 ) , it is likely that the true metallicity of these absorbers lies closer to the fiducial value [ o / h ] @xmath11 assumed by ds05 .
metallicity [ o / h ] @xmath24 in @xmath25% of the detected ly@xmath26 clouds ( ds05 ) is too high and too common to have been in place by @xmath27 , where schaye et al .
( 2003 ) find filling factors of 2
10% for gas with [ c / h ] @xmath28 . the
absorbers therefore imply significant later enrichment by galaxies .
conversely , if the metallicities of these absorbers lie near our lower limits , a portion of the sample would need to be removed to limit their cosmic density .
thus the mean metallicity , the dispersion in the metallicity , and the possibility of other mechanisms for producing are essential to understanding the contribution of absorbers to @xmath29 . to produce detectable absorption , an intervening cloud must ( 1 ) have the proper ionization conditions to ionize oxygen to , and ( 2 ) have sufficient metallicity to make the lines strong enough to detect .
the first condition can be met in cie gas produced in shocks ( shull & mckee 1979 ) or in hot - cold interfaces ( fox et al .
2004 ; indebetouw & shull 2004 , and references therein ) , though for some igm conditions photoionization by a metagalactic ionizing background ( haardt & madau 1996 ) can also produce .
the ionization mechanism need not be caused by , nor occur near , a galaxy , but in the standard picture of structure formation metals must be created in galaxies . when absorber / galaxy relationships are considered , there is no necessary connection between galaxies and the _ ionization _ of the gas , while there is a necessary connection between galaxies and the _ metallicity _ of the gas .
thus even if the absorbers fill only a portion of the igm hot - gas budget , they can still constrain the distribution of metals . in this paper
we study the dispersal of igm metals by galaxies using only _ observables _ and simple geometric arguments .
our model does not attempt to explain the shock - heating or photoionization mechanisms that cause or maintain the ionization of , but rather the metals that must have some link to nearby galaxies .
our model has significant advantages over simulations ( cen et al .
2002 ; fang & bryan 2001 ; chen et al .
2003 ; yoshida et al .
2005 ) or semi - analytic calculations ( furlanetto et al .
2005 ) that predict @xmath0/@xmath1 from _ a priori _ models .
these models either do not treat metal distribution in a self - consistent way , by adopting a uniform or density - dependent metallicity , or they use complicated models to calculate ionization structures and obtain reasonable ionization fractions .
by contrast , our model avoids all these complicated issues . by decoupling these entangled physical processes
, we can then focus on key issues , namely the volume that can be enriched by galaxies , and the minimum luminosity of the galaxies that can enrich the nearby igm .
our method also allows us to evaluate the hypothesis that the intergalactic absorbers are related to the hvcs seen in the galaxy . combining some simple geometrical arguments we derive the expected column density distribution , @xmath0/@xmath10 .
when compared with observations , our simple model provides significant constraints on the distribution of metals and their connection with galaxies .
our model compares the distribution of igm metals expressed by @xmath0/@xmath1 to the incidence of qso - galaxy `` pairs '' expected per redshift interval .
we calculate the total number of absorbers @xmath30 that can be encountered by one random sightline along a distance @xmath31 to a background qso .
the spatial density of halos with absolute magnitude between @xmath32 and @xmath33 is @xmath34 , where @xmath35 will ultimately be determined by sdss . assuming each halo can enrich the surrounding volume to a radius of @xmath36 , the total number of such intercepted absorbers within a distance of @xmath31 along the line of sight ( los ) toward background quasars then is @xmath37 here @xmath38 is the los covering fraction , representing the chances of intercepting absorbers within each halo . in the next section we will see that @xmath39 when we relate the local high velocity clouds ( hvcs ) and the intergalactic absorbers . in principle the integral in the last term of equation 1
could be stated in terms of mass and derived from , e.g. a press - schechter function , but in practice we use observations .
we begin with the sdss data release 3 catalog of qsos ( all spectroscopically confirmed ) and galaxies . to maximize the size of the sample and
therefore provide good statistics , we use all sdss photometric sources determined to be galaxies by photometric selection , then calculate photometric redshifts , @xmath40 , from the sdss @xmath41 photometry using the software of blanton et al .
these redshifts have typical errors @xmath42 , which make individual pairs quite uncertain but which wash out in the average .
for each qso sightline , we assign to all galaxies within 42@xmath43 an impact parameter @xmath44 to the qso sightline at the galaxy @xmath40 .
we require @xmath45 to minimize spurious detections on the low end and incompleteness at the high end .
we correct for incompleteness in faint galaxies at the higher redshifts by assigning each galaxy ( or magnitude bin of galaxies ) a unique pathlength @xmath46 based on its absolute magnitude and the sdss flux limits .
then , for a given spherical `` halo '' or `` enrichment '' radius @xmath36 and a los covering fraction @xmath38 , we can derive the total number and @xmath0/@xmath1of pairs , with cuts for any galaxy property ( i.e. , color , morphology , luminosity ) measurable from the @xmath41 photometry and @xmath47 .
figure [ dndzfig ] shows differential and cumulative distributions of @xmath0/@xmath1 achieved by galaxies of varying @xmath48 , given a uniform impact parameter @xmath44 and @xmath49 .
the key measurements are @xmath0/@xmath1 = 3 , 9 , 17 , 19 , and 21 for @xmath50 100 , 50 , 30 , 15 , and 10 m , respectively ( ds05 ) .
this mapping between galaxy populations and absorber incidence is the major result of this study .
spherical enrichment zones must be quite large , @xmath51 mpc , and uniform ( @xmath49 ) for enrichment by massive galaxies ( @xmath52 ) to explain the observed numbers of absorbers . for @xmath53 kpc , low - mass galaxies ( @xmath54 ) can reproduce the observed @xmath0/@xmath1 . to obtain @xmath0/@xmath1= 21 with dwarf galaxies only , we require uniform @xmath55 kpc for @xmath56 to @xmath57 and @xmath58 kpc for @xmath59 to @xmath57 .
in passing we note our model can also constrain @xmath36 for halos corresponding to high-@xmath17 ( ) ly@xmath26 clouds , i.e. @xmath60 kpc at @xmath52 is needed to reproduce @xmath0/@xmath1 = 1 for lyman limit systems ( keeney et al .
2005 ) .
= 1.1 ds05 find a power - law dependence of @xmath0/@xmath1 on @xmath61 , @xmath0/@xmath62 for absorbers with @xmath63 ( ) @xmath64 .
the behavior of @xmath0/@xmath10 at low @xmath17 is of critical importance to estimating @xmath65 .
however , below @xmath6 ( ) @xmath66 , the observed numbers of absorbers drop below the power - law fit , and well below the numbers expected from simulations that assume uniform or overdensity - dependent metallicity in the whim ( figure [ dndnfig ] ) .
ds05 can not say whether the apparent turnover below @xmath6 ( ) @xmath67 13.5 is real or due to survey incompleteness below @xmath68 m . integrating this power law down to @xmath6 ( ) = 13.0 ( 12.5 m ) yields @xmath0/@xmath1 = 106 .
this high number density requires [ o / h ] @xmath69 over @xmath70 kpc spheres for @xmath71 ( figure 1 ) , an unlikely outcome .
we believe it is more likely that the observed turnover below @xmath6 ( ) @xmath67 13.5 is real and not a consequence of incompleteness in the ds05 survey .
this prediction could be verified by a survey with more pathlength coverage down to @xmath6 ( ) = 13.0 ( 12.5 m ) .
if this deficit relative to theoretical expectations is real , our model suggests that metals are not as widely or uniformly distributed throughout the low - density igm as simulations assume or find them to be .
qso absorption - line studies of the dispersal of igm metals generally offer conflicting guidance because they suffer from the practical difficulty of identifying the faintest galaxy with which a given absorber might be associated .
these studies can not generally exclude the presence of a faint galaxy nearer the sightline than a known bright one , and so they tend to overestimate the dispersal of metals .
this effect is seen in the data points in figure 1 , which show @xmath32 and @xmath44 of nearest - neighbor galaxies to absorbers with galaxy surveys of varying depth and sky coverage .
there is a general trend of metals spread to @xmath72 kpc , with some notable exceptions .
the @xmath73 absorber toward pg0953 + 415 ( savage et al .
2002 ) suggests @xmath74 solar enrichment to @xmath70 kpc , but this field has not been surveyed much below @xmath75 .
the pks2155 - 304 absorbers ( shull , tumlinson , & giroux 2003 ) provide the best case for metal dispersal to @xmath76 kpc , because this field has been surveyed to @xmath77 , but the absorbers appear to be associated with a galaxy group that may complicate constraints on metal dispersal .
chen et al . ( 2001 ) systematically studied near 50 galaxies over a range of luminosity and found @xmath78 at @xmath79 kpc and @xmath80 at @xmath81 kpc , with typical limiting @xmath82 m .
this result does not rely on a post - facto galaxy survey , and so provides more robust evidence that luminous galaxies have associated metals out to @xmath83 kpc .
stocke et al .
( 2004 ) found a post - starburst dwarf with @xmath84 associated with a 6% solar metallicity absorber in the sightline to 3c 273 .
this galaxy , while starbursting , apparently drove a metal - enriched superwind @xmath85 kpc into the igm .
thus there is evidence that metal enrichment spreads to @xmath86 kpc from all galaxies , but evidence for further propagation is less clear .
we stress that the known @xmath0/@xmath1 of absorbers represents the minimum budget of metals that must be explained in terms of enrichment by galaxies .
there are surely other reservoirs of metals that meet condition 2 above ( detectable metallicity ) but not 1 ( conditions that produce ) .
any difficulty associated with distributing the oxygen seen in can only be compounded by the addition of these other reservoirs ( e.g. , photoionized gas in systems ) to the budget of igm metals .
finally , figure 1 shows results for @xmath87 and so presents the most optimistic case of enrichment .
our model can also interpret the intergalactic o vi absorbers as analogs of the local highly - ionized high - velocity clouds ( sembach et al .
2003 , hereafter s03 ) . to do so
, we must first derive a mapping between the covering fraction , @xmath88 , of hvcs with column density @xmath17 as seen along radial sightlines out from the galaxy , to @xmath38 , the covering fraction along a los that follows a chord through an intervening galaxy halo .
the key assumption here is that the hvcs are not _ unique _ , i.e. , an observer located in another galaxy similar to our galaxy should be able to detect similar hvcs with a similar pattern of column density distribution .
let us first consider hvcs with column densities between @xmath89 and @xmath90 , and with a covering fraction @xmath91 the total covering fraction @xmath88 is the sum of @xmath91 .
assuming there are a total of @xmath92 such absorbers uniformly distributed within the halo ( of a galaxy or a group ) , and each absorber has a spherical geometry with radius @xmath93 , we find @xmath92 and the spatial number density of the absorbers within the halo @xmath94 are @xmath95 respectively , where @xmath36 is the radius of the halo . these absorbers project onto the sky a total area @xmath96 , such that their covering fraction along the los is @xmath97 we derive all quantities from observation - @xmath91 from table 4 of s03 and the spatial density of galaxies from sdss .
the total los covering fraction @xmath38 then is @xmath98 . adopting the value of @xmath88 from s03
, we find @xmath38 ranges from @xmath99 to @xmath100 down to 30 m , which justifies the value of 1 we adopt in the previous section .
our results appear in figure [ dndnfig ] , where we show that hvc - like absorbers can make a significant contribution to the observed numbers if massive galaxies have large halos , or if all galaxies down to @xmath101 ( @xmath102 ) have hvcs out to @xmath86 kpc .
highly ionized gas can arise in galaxy superwinds ( heckman et al .
2001 ) , in gas tidally stripped from galaxies ( cox et al .
2004 ) , and in accreting gas ( maller & bullock 2004 ; birnboim & dekel 2002 ) . the @xmath84 dwarf 71@xmath103 kpc away from a metal - enriched absorber toward 3c 273 ( stocke et al .
2004 ) is one example at the faint end of the luminosity function .
the _ fuse _ studies of galactic hvcs ( sembach et al .
2003 ; fox et al .
2004 ; collins , shull , giroux 2004 ) show that can arise in the extended halo regions of larger galaxies ( @xmath86 kpc ) and not necessarily correspond to the whim . from figure [ dndnfig ] , it appears that if these processes have high @xmath38 out to @xmath86 kpc around galaxies down to @xmath104 , they may contribute @xmath0/@xmath1 @xmath105 to the overall population of absorbers . given the potentially important contributions of absorbers to @xmath106 , their utility as pointers to the hotter whim seen in x - ray absorption , and considering the many mechanisms that can produce near galaxies , it is important to obtain an unbiased observation of the incidence of and metals within @xmath107 kpc of galaxies over a range of mass .
until such an observation is made , we can not know how to correct the inferred baryon densities for these contaminating effects .
our observations can also guide the design of such studies by constraining the fields over which galaxy surveys must be executed .
based on our model of the distribution of igm metals and the observed numbers of igm absorbers , we draw the following conclusions : 1 . unless they eject metals to @xmath108 mpc , luminous galaxies can not explain @xmath0/@xmath1 for metal - line absorbers in the low-@xmath109 igm . if low - mass galaxies ( @xmath110 ) disperse metals to @xmath111 kpc , @xmath0/@xmath1 for the absorbers can be explained .
uniform enrichment to @xmath112 1 mpc is implied by integrating the power - law dependence of @xmath0/@xmath10 ( ds05 ) to @xmath6 ( ) = 13.0 .
thus the turnover seen by ds05 in @xmath0/@xmath10 below @xmath6 ( ) @xmath113 ( 50 m ) is more likely a real effect than a consequence of survey incompleteness .
a survey sensitive to @xmath114 m could confirm this result and further constrain the dispersal of igm metals .
conversely , additional absorbers below @xmath6 ( ) @xmath113 would imply that metals are quite widespread .
3 . a high covering fraction of from `` hvcs '' within @xmath107 kpc from galaxies can contribute @xmath0/@xmath1 = @xmath115 , a non - negligible portion of the observed total that would not contribute to @xmath116 .
thanks to charles danforth and mike shull for sharing their results before publication and to charles danforth and john stocke for constructive comments .
we began this work in the hospitality of the kavli institute for theoretical physics at the university of california , santa barbara , supported in part by nsf under grant no .
phy99 - 07949
. j. t. is supported by the astronomy and astrophysics department at the university of chicago and hst go-9874.01-a .
t. f. was supported by nasa through _ chandra _ postdoctoral fellowship pf3 - 40030 issued by the _ chandra _ x - ray observatory center , which is operated by the smithsonian astrophysical observatory for and on behalf of nasa under contract nas8 - 39073 .
birnboim , y. , & dekel , a. 2002 , mnras , 345 , 349 blanton , m. r. , et al .
2003 , , 592 , 819 cen , r. , & ostriker , j. p. 1999
, , 519 , l109 chen , x. , weinberg , d. h. , katz , n. , dav , r. 2003 , , 594 , 42 collins , j. a. , shull , j. m. , & giroux , m. l. 2004 , , 605 , 216 cox , t. j. , et al .
2004 , , 607 , l87 danforth , c. w. , & shull , j. m. 2005 , , in press fang , t. , & bryan , g. l. 2001 , , 561 , l31 fang , t. , et al .
2002 , , 572 , 127 fox , a. j. , et al .
2004 , apj , 602 , 738 furlanetto , s. , phillips , l. a. , & kamionkowski , m. 2005 , , in press haardt , f. , & madau , p. 1996
, , 461 , 20 heckman , t. , et al .
2001 , apj , 554 , 1021 indebetouw , r. , & shull , j. m. 2004 , , 605 , 205 keeney , b. a. , et al .
2005 , , submitted ( astro - ph/0409448 ) maller , a. h. , & bullock , j. s. 2004 , mnras , 355 , 694 mathur , s. , weinberg , d. h. , & chen , x. 2003 , , 582 , 82 mckernan , b. , et al .
2003 , , 583 , l83 nicastro , f. , et al . 2005 , nature , in press prochaska , j. x. , chen , h .- w . , howk , j. c. , weiner , b. j. , & mulchaey , j. 2004 , , 617 , 718 savage , b. d. , et al .
2002 , , 564 , 631 schaye , j. , aguirre , a. , kim , t .- s . , theuns , t. , rauch , m. , & sargent , w. l. w. 2003 , , 596 , 768 shull , j. m. , & mckee , c. f. 1979 , , 227 , 131 shull , j. m. , tumlinson , j. , & giroux , m. l. 2003 , , 594 , l107 spergel , d. n. , et al .
2003 , , 148 , 175 stocke , j. t. , et al .
2004 , , 609 , 94 tripp , t. m. , savage , b. d. , & jenkins , e. b. 2000 , , 534 , l1 tripp , t. m. , giroux , m. l. , stocke , j. t. , tumlinson , j. , & oegerle , w. r. 2001 , , 563 , 724 tripp , t. m. 2005 , in astrophysics in the far ultraviolet , in press tumlinson , j. , shull , j. m. , giroux , m. l. , & stocke , j. t. 2005 , , 620 , in press yoshida , n. , furlanetto , s. r. , & hernquist , l. 2005 , , 618 , 91 | we use the observed number and column - density distributions of intergalactic absorbers to constrain the distribution of metals in the low - redshift intergalactic medium ( igm ) . in this simple model the metals in the absorbers are assumed to be produced in and propagated from low - redshift galaxies drawn from a real sample , in this case the sloan digital sky survey ( sdss ) .
this model can explain the observed @xmath0/@xmath1 of metals seen in absorbers if these metals are dispersed out to @xmath2 kpc by galaxies down to @xmath3 .
massive galaxies ( @xmath4 ) by themselves can not provide the necessary enrichment unless they can enrich volumes out to @xmath5 mpc .
alternatively , our model allows an estimate of the fraction of absorbers directly caused by galaxies rather than hot igm . with this assumption
we explore the possible connections between the intergalactic absorbers and the known populations of highly - ionized high - velocity clouds ( hvcs ) surrounding the milky way .
our model predicts that more sensitive , complete surveys optimized to uncover weaker absorbers will find the tentative turnover below @xmath6 ( ) @xmath7 to be a real effect resulting from the apparently limited volumes over which galaxies can enrich the igm .
if so , it would indicate that metals are not as widespread throughout the low - density igm as they are assumed to be in cosmological simulations of the whim . |
the ability to control the rotational motion and to angularly confine molecules has various applications in molecular sciences .
this includes studies of steric effects in chemical reactions , both bimolecular and photoinduced , and the possibility to investigate molecules from their own point of view , the molecular frame .
the latter mitigates the usual blurring of experimental observables caused by the random orientation of molecules in uncontrolled samples .
access to molecular frame measurements is crucial in several applications , notably in various modern schemes aiming at observing the ( coupled ) motion of nuclei and electrons during chemical reactions @xcite .
methods based on the use of moderately intense , nonresonant , near - infrared laser pulses have proven particularly useful for controlling the alignment and , in conjunction with weak dc electric fields , orientation of a broad range of molecules .
alignment refers to the confinement of molecule - fixed axes along laboratory - fixed axes , and orientation usually refers to the molecular dipole moment pointing in a particular direction @xcite . for a linear molecule
, only a single axis needs to be confined in space to ensure complete rotational control .
this can be achieved by a linearly polarized laser pulse , which will align the most polarizable axis ( mpa ) , i.e. the internuclear axis of the molecule .
this is termed one - dimensional ( 1d ) alignment @xcite .
combined with a ( weak ) static electric field it can also control the head - versus - tail order of a polar molecule , i.e. , induce 1d orientation @xcite .
complete rotational control of asymmetric top molecules requires the confinement of three molecular axes to laboratory frame fixed axes , resulting in 3d alignment . in the adiabatic limit , where the laser pulse is turned on slower than the rotational periods of the molecule
, it has been shown that an elliptically polarized laser pulse can induce 3d alignment @xcite . for polar molecules , where the permanent dipole moment ( dm ) is parallel to the mpa
it has also been shown that 3d orientation , defined as 3d alignment and a unique direction of the dm , can be achieved by combining the elliptically polarized laser pulse with a weak static electric field parallel to the major polarization axis @xcite . for most asymmetric top molecules ,
the dm does , however , not coincide with any principal axis of polarizability .
while 3d alignment is expected to work well for these less symmetric molecules , it remains to be explored if the combined action of a linearly or elliptically polarized laser pulse and a weak or strong static electric field can efficiently induce 3d orientation .
also , as discussed here , the precise meaning of 3d orientation must be specified for molecules with low symmetry . in the current work
we investigate 3d alignment and orientation of asymmetric top molecules where the dm is not parallel to a principal axis of polarizability .
our studies are motivated by the fact that many important biomolecules , e.g. , amino acids , nucleic acids , peptides , and dna strands , belong to this class of molecules .
controlling how they are turned in space would be of significant value in novel and emerging schemes for time - resolved molecular imaging @xcite .
following the conclusions from the current work , this three - dimensional control is indeed possible .
our studies focus on 6-chloropyridazine-3-carbonitrile ( , cpc ) .
the molecule is chosen because the dm is off - set by 57.1@xmath0 from the mpa and because the atomic composition makes it possible to determine its 3-dimensional spatial orientation through coulomb explosion imaging .
a sketch of the molecular structure of cpc and the position of the mpa and dm vector are shown in fig : mol - structure .
the planar molecule consists of an aromatic pyridazine ring with a nitrile and a chlorine substituent . the n - n bond is shorter than the c - n and c - c bonds in the aromatic ring causing the bond angle of the substituents on the ring to bend slightly towards the pyridazine nitrogens , as can be seen from the energy optimized geometry of the molecule shown in fig : mol - structure .
quantum chemical calculations determine the electric dipole moment of cpc to be with @xmath1 , @xmath2 , @xmath3 , i.e. , in the molecular plane and at an angle of 57.1@xmath4with respect to the @xmath5 axis .
the static polarizability components of cpc are determined to be @xmath6 , @xmath7 and @xmath8 .
most aspects of the experimental setup have been described previously @xcite and only a few pertinent details will be given here .
a few mbar of cpc ( chemfuture pharmatech , @xmath9 chemical purity ) was seeded in a helium carrier gas at a backing pressure of and expanded into vacuum through a pulsed even - lavie valve @xcite heated to c. the expansion was skimmed twice before entering an electrostatic deflector , where the molecules were deflected according to the effective dipole moment @xmath10 of their specific rotational quantum state @xcite .
the quantum - state - dispersed molecular beam entered a velocity map imaging ( vmi ) spectrometer where it was crossed at by two collinear laser beams .
the laser beams were focused by a spherical lens ( @xmath11 ) mounted on a motorized translation stage .
this allowed for the height of the foci to be scanned with high precision .
hereby it is possible to measure the vertical intensity profile of the molecular beam such that the effect of the electrostatic deflector can be characterized .
furthermore , the focused laser beams can be directed to the most deflected molecules which are the ones residing in the lowest lying rotational states and , thereby , those that undergo the strongest alignment and orientation @xcite .
the molecules were aligned and oriented by the combined effect of pulses from one of the laser beams ( yag pulse , @xmath12 , @xmath13 , @xmath14 , @xmath15 , injection seeded ) and the weak static electric field from the vmi spectrometer ( @xmath16was varied between values of and ) .
the polarization of the yag pulse can be rotated to rotate the molecular alignment with respect to the static field direction which is fixed by the vmi spectrometer axis
see sec .
[ sec : orientation ] .
the yag beam was overlapped in space and time with pulses from a second laser ( probe pulse : , , , @xmath17 ) .
these short pulses multiply ionized the molecules , which then fragmented into charged ions .
the ions were projected onto a 2-dimensional particle detector in order to detect their recoil directions . for cpc molecules ,
ion momenta were recorded to determine the spatial orientation of the bond axis with respect to the laboratory frame .
the or fragment ion distributions were recorded to provide information about the orientation of the molecular plane in the laboratory frame .
all experiments were conducted on deflected , state - selected molecular samples at a repetition rate of , limited by the yag laser .
theoretically we investigated the rotational dynamics of the cpc molecule in combined static electric and nonresonant laser fields polarized either linearly or elliptically . due to the complexity of this system
, we retreated to a quasi - static description .
we assumed that the interaction with the laser pulse can be described within the adiabatic limit .
we applied a two - photon rotating - wave approach averaging over the rapid oscillations of the nonresonant field . in the framework of the rigid - rotor approximation
, we solved the time - independent schrdinger equation of the cpc molecule in a field configuration equivalent to the experimental one .
it is convenient to define a laser - polarization - fixed frame ( lpff ) @xmath18 . for the elliptically polarized field ,
the major polarization axis defines the @xmath19-axis and the minor polarization the @xmath20-axis . for the linearly polarized field ,
the @xmath19-axis is defined by the polarization axis .
the homogeneous electrostatic field of strength @xmath21 is contained in the @xmath22-plane at an angle @xmath23 with respect to the @xmath19 axis .
the relation between the lpff and the molecular fixed frame ( mff ) @xmath24 ( defined in fig : mol - structure ) is given by the euler angles @xmath25 @xcite .
the hamiltonian of this system is @xmath26 with the rotational constant @xmath27 , @xmath28 and @xmath29 and the interaction operators @xmath30 and
@xmath31 with the dc and ac electric fields , respectively . the stark interaction reads @xmath32 with the absolute value of the electric dipole moment @xmath33 , and its two components @xmath34 and @xmath35 .
the angles between the electric field and @xmath36 , and the mff @xmath5 and @xmath37-axes , @xmath38 , @xmath39 and @xmath40 , respectively , are given by the relations @xmath41 the interaction of the molecule with a nonresonant elliptically polarized laser field can be written as @xmath42 where @xmath43 and @xmath44 are the intensities of the polarization components along the lpff @xmath20 and @xmath19 axes , respectively .
the total intensity is @xmath45 , and @xmath46 is used here .
@xmath47 , and @xmath48 are the @xmath49-th diagonal element of the polarizability tensor , with @xmath50 .
@xmath51 the dielectric constant and @xmath52 is the speed of light .
@xmath53 are the angles between the lpff @xmath54-axis and the mff @xmath55-axis , and they are related to the euler angles as follows @xmath56 if the laser field is linearly polarized , the interaction with this field is obtained by setting @xmath57 and @xmath58 in eqn : hamil_laser_ellip .
the time - independent schrdinger equation of the hamiltonian eqn : hamiltonian was solved by expanding the wave function in a basis set formed by linear combinations of field - free symmetric top wave functions , i.e. wigner functions @xcite . for each field configuration
, we constructed a basis that respects the symmetries of the corresponding irreducible representation @xcite .
let us shortly summarize the symmetries of this system in the mixed - field configurations . in the field - free case ,
they are given by the spatial group so(3 ) and the molecular point group d@xmath59 @xcite .
as a consequence , the total angular momentum @xmath60 and its projection @xmath61 onto the @xmath19-axis of the lpff are good quantum numbers , but the projection of @xmath60 onto the @xmath5-axis of the mff ( @xmath62 ) is not well defined .
the symmetries of the hamiltonian eqn : hamiltonian with a linearly polarized laser and a dc electric field tilted by an angle @xmath23 have been analyzed in detail in ref . .
for an elliptically polarized laser field in the @xmath22 plane and with the dc field parallel to the @xmath19-axis , i.e. , @xmath63 , a @xmath64-rotation around the lpff @xmath19-axis and the reflection on the @xmath22-plane ( the laser polarization plane ) are the symmetry operations and @xmath61 is not a good quantum number , but its parity is . for the other two cases , @xmath65 and @xmath66 , the system has the same symmetries as in the corresponding field configuration with a linearly polarized laser field , see ref . .
we start by showing that a linearly polarized yag pulse induces 1d alignment of the cpc molecules . for this purpose ,
the emission directions of ions are detected .
the expected action of the yag pulse is that it aligns the mpa along its polarization axis and as such an experimental observable that provides direct and precise information about the spatial orientation of this axis would be ideal . unlike in higher - symmetry molecules ,
e.g. , iodobenzene @xcite , no such observable exists .
the emission direction of ions comes close , assuming axial recoil along the c - cl bond axis , since the c - cl axis is only off - set by 3 degrees from the mpa .
when only the linearly - polarized probe pulse is applied , polarized perpendicular to the detector , the image shown in fig:1d - alignment(a ) is circularly symmetric , as expected for randomly oriented molecules . .
( c ) the degree of alignment @xmath67as a function of @xmath68 . ]
when the yag pulse is included the ions tightly localize along its polarization axis parallel to the detector plane , see fig:1d - alignment(b ) .
these observations show that the c - cl bond axes of the cpc molecules are aligned along the yag pulse polarization axis , i.e. , that 1d alignment is induced .
the degree of alignment is quantified by determining the average value of @xmath69 , @xmath67 , where @xmath70 is the angle between the yag pulse polarization and the projection of a ion velocity vector on the detector screen . only a confined radial range is used to determine @xmath67 .
this range at the outermost part of the images is marked by circles in fig:1d - alignment(b ) .
it corresponds to ions originating from a highly directional coulomb explosion process .
the derived values are plotted as a function of the yag pulse intensity , @xmath68 in fig:1d - alignment(c ) .
@xmath67rises from 0.5 , the value characterizing a sample of randomly oriented molecules , at @xmath71 to @xmath72 at the highest value of @xmath68 .
this behaviour is fully consistent with many previous studies of 1d adiabatic alignment @xcite .
the @xmath67values determined underestimate the true degree of alignment due to the offset between the c - cl axis and the mpa and imperfect axial recoil . in order to investigate the effect on the molecular alignment when the yag pulse polarization is changed from linear to elliptical ,
an ellipticity ratio of 3:1 was applied , i.e. , the intensity along the major polarization axis of the yag pulse is three times the intensity along the minor axis . for these measurements , and images , displayed in fig : cpc - alignment , are used to infer information about the molecular alignment . .
the images represent either a `` side - view '' when the major polarization axis is parallel to the detector ( vertical in fig : cpc - alignment ) , i.e. , the molecules are watched from the side , or an `` end - view '' when the major polarization axis is perpendicular to the detector , i.e. , the molecules are watched from the end @xcite . for ions ,
the molecule is imaged in side - view .
fig : cpc - alignment(a1 ) shows the image obtained with the probe pulse by itself , polarized vertically , and serves as a reference . in panel ( a2 ) measurements including the linearly polarized yag pulse
are shown .
the ions appear as two distinct areas at large radii along the vertical axis and as two pairs of wings protruding nearly horizontally from the vertical centerline .
the 1d alignment means that the mpa of the molecules is confined along the vertical @xmath19 axis and it implies that the recoiling ions from the group will be ejected vertically , either up or down depending on the orientation of the molecule .
these ions form the two distinct centerline regions of signal similar to the ion structure ( fig:1d - alignment(b ) ) used for the determination of the 1d alignment discussed above .
the wing structure is interpreted as ions originating from coulomb explosion of the n atoms in the aromatic ring .
since the linearly polarized yag pulse does not impose any constraint on the rotation of the ring the ions will be emitted in a double - torus - like pattern . upon projection on the 2d detector plane
this gives the wing - structure .
covariance analysis @xcite confirms that in the wing - structure the two ions from a single molecule are predominately produced on the same sides of the tori supporting this interpretation .
when the yag pulse polarization is changed to elliptical the image in fig : cpc - alignment ( a3 ) is obtained .
the ions from the ring are confined close to the vertical axis , whereas the ion emission structure from the cn group is practically unchanged .
this shows that the alignment of the mpa is not much changed while the molecular plane is no longer free to rotate , but instead it is confined to the polarization plane .
this demonstrates that the molecule is 3d aligned .
the corresponding end - view images of in row ( b ) corroborate this interpretation : with a linearly polarized yag pulse the ions emerge in the circularly symmetric pattern shown in panel ( b2 ) , corresponding to free rotation of the molecular plane around the yag pulse polarization axis .
( note that in this measurement the probe pulse polarization is perpendicular to the detector plane ) .
for an elliptically polarized yag pulse , the ions are angularly localized around the vertical minor polarization axis , i.e. , the molecular plane is confined to the polarization plane .
the radial structures of panel ( b2 ) and ( b3 ) are the same , confirming that the long axes of the molecules remain aligned along the major polarization axis .
thus , the images confirm that the cpc molecules are 3d aligned by the elliptically polarized yag pulse .
first we discuss the orientation that results from the combined action of a linearly polarized , moderately intense laser and a weak static electric field . all previous studies comprised molecules where the permanent dipole moment was parallel to the mpa including linear rotors @xcite , symmetric tops @xcite and asymmetric tops @xcite .
in these cases 1d orientation , defined as 1d alignment and a preferred direction of the permanent dipole moment , was induced . for studies employing ion imaging , as in the current work
, orientation was observed when the molecule was rotated away from the side - view geometry used in pure alignment measurements ; see fig : cpc_exp_setup_1d_3d_orient(a ) for a sketch of this experimental approach .
( the angle between the static field direction and the ( major ) polarization axis of the yag pulse ) is changed . ] in practice this was done by rotating the yag pulse polarization to angles where @xmath73 .
this provides a component of the static field along the dipole moment which mixes the pendular states of the tunneling doublet to form the corresponding oriented states .
the experimental findings showed that the degree of orientation increased monotonically as @xmath23 was rotated from @xmath74 towards @xmath75 or @xmath76 @xcite .
later experiments and analysis have identified this behaviour as resulting from nonadiabatic dynamics in the mixed - field orientation @xcite . in the following
we investigate if the angular offset of the dipole moment from the mpa in cpc influences the efficiency of mixed - field orientation and if the degree of orientation peaks when the mpa or the permanent dipole moment is directed along the static field from the vmi spectrometer .
the experimental observables used are the ion images which provide information about the orientation of the mpa .
examples of ion images recorded for @xmath77 @xmath78 are shown as insets in fig : beta . between the static field and the ( major ) polarization axis of the yag pulse for different field strengths .
the values are determined from the distribution of ions .
the insets show raw ion images at ( left ) @xmath77 and ( right ) @xmath79 . ] for @xmath77 @xmath78 , more ( less ) ions are detected on the upper half of the detector than on the lower half . in analogy with previous studies we interpret these observations as 1d orientation due to the combined effect of the yag laser field and the static electric extraction field @xcite , where the dm @xmath5-component orients along the projection of the dc electric field onto the mpa .
this implies that the partially negative nitrile end will be directed towards the repeller electrode where the potential is highest and the c - cl bond towards the extractor electrode where the potential is lowest see fig : cpc_exp_setup_1d_3d_orient(a ) . as a consequence
, ions are expected to be ejected upwards ( downwards ) for @xmath77 @xmath78 .
this is in agreement with the up - down asymmetry in the images .
the degree of 1d orientation is quantified by dividing the number of ions detected on the upper half of the detector by the total amount of ions detected ( @xmath80 ) , corrected for the slight up - down detector asymmetry .
this asymmetry parameter , is plotted in fig : beta as a function of @xmath23 recorded for two values of the static electric field .
the figure shows that the degree of orientation increases gradually as the mpa is rotated towards the direction of the static electric field .
this behaviour is similar to that observed for molecules such as ocs @xcite and iodobenzene @xcite where the mpa and the permanent dipole moment are parallel .
the experimental findings are rationalized by our computational treatment discussed in sec : theory - results .
next we investigated 3d orientation .
in previous studies 3d orientation was defined as 3d alignment occurring together with a preferred direction of the permanent dipole moment axis , which coincides with the mpa for molecules with rotational point group symmetry like iodobenzene and benzonitrile ( molecular point group @xmath81 ) . in the case of molecules like cpc ( molecular point group @xmath82 ) full 3d orientation [ fig : cpc_exp_setup_1d_3d_orient(c ) ] requires preferred directions in space for at least two components of the dipole moment . for cpc
this would be the @xmath5-axis and the @xmath37-axis see fig : mol - structure .
a slightly less complete 3d confinement of the molecular rotations is 3d alignment and 1d orientation of one of the principal axes of polarizability .
this can be either the @xmath5-axis ( the mpa ) as illustrated in fig : cpc_exp_setup_1d_3d_orient(b ) or the @xmath37-axis which will be discussed in sec : theory - results . in the experiment we measured images at different values of @xmath23 for an elliptically polarized yag pulse with the same parameters as the one used to induce 3d alignment ( where @xmath83 or @xmath74 ) .
the results , displayed in fig : beta , are very similar to the results obtained with the linearly polarized yag pulse and establish that 1d orientation of the @xmath5-axis occurs .
no information about the @xmath37-axis can be extracted from this ion species .
images of and ions were also recorded .
both ion species are confined along the vertical centerline of the images which shows that 3d alignment occurs - similar to the case presented in fig : cpc - alignment(a3 ) and fig : cpc - alignment(b3 ) as expected .
neither of these ion species are , however , well suited to extract information about the orientation of the @xmath37-axis . at this point
we , therefore , conclude that the cpc molecules are , at least , 3d aligned and simultaneously have their @xmath5-axis 1d oriented .
in this subsection we consider only the influence of a static electric field on the cpc molecule , i.e. , @xmath71 .
the @xmath34 ( @xmath35 ) term of the stark effect interaction couples states with different parity under inversion along the molecular @xmath5 ( @xmath37 ) axis . as the dc field strength
is increased , the electric dipole moment @xmath36 gets oriented along the electric field axis .
the expectation values @xmath84 , @xmath85 and @xmath86 , see eqn : cos_smu , [ eqn : cos_sz ] and [ eqn : cos_sy ] , measure the orientation of @xmath36 and of the molecular @xmath5 and @xmath37 axes , respectively .
they are presented as a function of the dc field strength @xmath87 in fig : orientation_cosines . for the rotational ground state and @xmath88 , we compute @xmath89 , @xmath90 , and @xmath91 . in such a weak field
the orientation of the @xmath37-axis , corresponding to the largest dipole moment component , is smaller than the orientation of the @xmath5 axis , see fig : orientation_cosines , because the energy gap from the ground state to the first level with odd parity under inversion along the @xmath37-axis ( @xmath92 ) is larger than to the first level with odd parity under the inversion along the @xmath5-axis ( @xmath93 ) .
when @xmath87 is increased , the hybridization of the pendular levels increases , and this trend in the orientation is inverted ; we encounter that @xmath94 for @xmath95 , see fig : orientation_cosines . in the strong - dc - field regime @xmath96 , @xmath97 and @xmath98 .
( thick solid line ) , @xmath85 ( dashed line ) and @xmath86 ( dot - dashed line ) of cpc , and @xmath85 ( thin solid line ) of 0-cpc as a function of the electric field strength @xmath87 .
the field configuration is @xmath99 and @xmath71 . ] to investigate the influence of @xmath35 on the dc - field orientation , we have considered a molecule with the same rotational constants and polarizability as cpc , but with @xmath100 d and @xmath101 d. when this @xmath102-cpc molecule is exposed to an electric field , only its @xmath5-axis gets oriented along the @xmath19-axis . for weak dc fields , the ground states of the cpc and @xmath102-cpc molecules show similar values of @xmath85 ; we find relative differences between @xmath103 and @xmath104 for @xmath105 . by increasing @xmath87 , these relative differences increase and are larger than @xmath106 for @xmath107 .
the 0-cpc always orients better , i.e. , its orientation cosine @xmath85 is larger than the corresponding ones @xmath84 and @xmath85 of the cpc .
both molecules share the same field - free energy level structure , but the @xmath34 and @xmath35 stark interactions couple different states , which provoke a larger orientation for 0-cpc in despite of its smaller dipole moment . only in the strong dc - field regime , when the pendular levels are strongly hybridized these two systems show a similar orientation .
we obtain @xmath108 for the cpc ground state and @xmath109 for the 0-cpc ground state at @xmath110 .
we now consider the molecule in a linearly polarized strong laser field , when tunneling doublets of aligned states are formed @xcite . in an additional tilted weak electric field , the terms in @xmath34 and @xmath35 in eqn : hamil_electric couple states in the same doublet and between neighbouring doublets , respectively .
for the experimentally employed field - strengths , the interaction due to the nonresonant laser field dominates . for @xmath111 , the energy splittings of the sublevels in the lowest two pendular doublets
are smaller than @xmath112 @xmath113 , the energies of the two doublets differ by @xmath114 @xmath113 , and the mpa is strongly aligned along the @xmath19-axis with @xmath115 for these four levels . for a weak dc field ,
the @xmath114 @xmath113 energy gap between two consecutive doublets is larger than the interaction due to the dc field : for @xmath88 , @xmath116 @xmath113 and @xmath117 @xmath113 .
note that these quantities provide upper bounds to the dc field interactions because the angular dependence in eqn : hamil_electric is set to @xmath118 , which holds only for fully oriented states . as a consequence , for weak dc fields ,
the coupling is only significant between states in the same doublet and the states become oriented or antioriented along the lpff @xmath19-axis , but no orientation of the molecular @xmath37-axis is achieved .
this can be illustrated by a comparison between the cpc and @xmath102-cpc results in this field configuration .
for @xmath119 , these molecules present the same mixed - field orientation of the @xmath5-axis @xmath120 with relative differences smaller than @xmath121 . at the experimental field regime ,
the mixed - field orientation of both systems is dominated by the stark interaction due to @xmath34 , and the contribution of @xmath35 can be neglected . in fig : orientation_cosines_linearly_polarized_laser we observe that in this regime the orientation cosines along the lpff @xmath19 axis of the cpc ground state , i.e. , @xmath120 and @xmath85 , are larger than those along the lpff @xmath20 axis , i.e. , @xmath122 and @xmath86 . for the cpc ground state ,
the degrees of orientation and alignment are presented in tab : lineal for @xmath15 and @xmath123 .
as @xmath87 is increased ( to values @xmath124 or @xmath125 in tab : lineal and up to @xmath126 in fig : orientation_cosines_linearly_polarized_laser ) , the coupling due to @xmath35 is enhanced and the molecular @xmath37-axis gets oriented along the lpff @xmath20-axis , whereas @xmath120 and @xmath85 keep constant values . in this strong electric field regime , the cpc molecule is completely 3d oriented .
thus , the difference between the cpc and @xmath102-cpc systems appears only for strong dc fields where they are 3d and 1d oriented , respectively .
however , even in this regime , they still have the same value of @xmath120 .
( thick solid line ) , @xmath85 ( dashed line ) , @xmath122 ( dotted line ) , and @xmath86 ( dot - dashed line ) calculated as a function of the electric field strength @xmath87 for a linearly ( black ) and an elliptically ( green ) polarized laser pulse . the field configurations are @xmath127 and @xmath15 .
the ellipticity ratio is 3:1 . ]
.orientation and alignment of the ground state of the cpc molecule in a dc electric field and an linearly polarized yag pulse of @xmath15 forming an angle of @xmath123 .
[ cols="^,^,^,^",options="header " , ] if the dc electric field forms an angle of @xmath128 with the major polarization axis of the laser pulse the following cosine expectation values were obtained for @xmath15 and @xmath129 : @xmath130 , @xmath131 , @xmath132 , @xmath133 .
this means that the cpc molecule is 3d aligned and the @xmath5-axis 1d oriented similar to the case for higher symmetry molecules studied previously @xcite .
the @xmath37-axis is equally likely to point either upward or downward. if @xmath23 is changed to @xmath134 the results are : @xmath130 , @xmath131 , @xmath135 , @xmath136 , this shows that the molecules are still 3d aligned , now with the @xmath37-axis 1d oriented along the static electric field whereas the @xmath5-axis points either forward or backward. analogous features are found for the excited rotational / pendular levels .
the complexity of their field - dressed dynamics is significantly enhanced due to the large number of avoided crossings .
these avoided crossings provoke abrupt changes on their directional properties , which play an important role for the mixed - field orientation of the molecular beam @xcite . for the experimentally accessed regime of dc field strengths ,
@xmath120 is practically independent of @xmath137 , whereas @xmath122 increases until a large orientation is reached for high intensities of elliptically polarized pulse , e.g. , @xmath138 and @xmath139 for and @xmath140 , @xmath141 and @xmath123 .
regarding the behaviour of the orientation cosines @xmath120 and @xmath122 versus @xmath23 , three different regimes are observed : i ) for weak alignment lasers , when the pendular doublets are not yet formed , or the energy splitting between two neighbouring doublets is larger than the dc - field interaction , @xmath120 or @xmath122 monotonically increase with @xmath23 , respectively ; ii ) for stronger laser fields , the energy separations between the doublets are significantly reduced , and the orientation is independent of @xmath23 ; iii ) if the dc - field interaction is much larger than the laser - field interaction , the orientation in both directions reaches a maximum at @xmath142 , because the effect of the static field becomes optimal at this field configuration . in particular , this time - independent description predicts an orientation of the mpa along the @xmath19 axis independent of @xmath23 and @xmath87 .
thus , the smooth behaviour of @xmath80 versus @xmath23 in fig : beta can not be reproduced with this theoretical treatment .
indeed , the authors have recently demonstrated that only a time - dependent study can reproduce the intriguing physical phenomena taking place in the mixed - field orientation experiments @xcite of @xmath81 symmetric molecules .
the detailed quasi - static description of the cpc molecule in mixed dc and linearly or elliptically polarized non - resonant ac fields presented here provides a solid basis for a future time - dependent study of this system , which is beyond the scope of this work .
let us remark that the knowledge of the adiabatic energy structure in mixed - fields is required for an adequate interpretation of the non - adiabatic phenomena taking place in the field - dressed dynamics , such as the formation of pendular doublets @xcite . for completeness
, we theoretically investigated the mixed - field orientation of thermal samples of cpc , in order to mimic the state - selection , assuming that the alignment and orientation processes are adiabatic @xcite .
for an elliptically polarized laser with @xmath15 , @xmath88 , and @xmath123 , the molecular sample at @xmath118 k is strongly aligned but practically not oriented , consistent with experimental findings @xcite : @xmath143 ( @xmath144 , @xmath145 ) , @xmath146 , @xmath147 and @xmath148 . by reducing the temperature to @xmath149 k , the alignment is slightly improved to @xmath150 ( @xmath151 , @xmath152 ) and the orientation is strongly increased @xmath153 , @xmath154 and @xmath155 .
this demonstrates even that for our very cold molecular beams ( @xmath156 k ) the state - selection of low - energy rotational states is crucial for the creation of orientation @xcite .
we have performed a combined experimental and theoretical investigation of mixed - field orientation of the 6-chloropyridazine-3-carbonitrile ( cpc ) molecule .
our studies are motivated by the fact that this molecule represents the large class of important species where the permanent dipole moment does not coincide with any of the three principal axes of polarizability .
experimentally we showed that the combination of an elliptically polarized laser pulse and a weak static electric field , not coinciding with either the major or the minor polarization axis of the light field , leads to 3d alignment of the molecule and 1d orientation of the most polarizable axis along the major polarization axis .
the experiment is not capable of determining whether the second most polarizable axis is also oriented along the minor polarization axis but our calculation shows that it should be the case .
this situation represents the most comprehensive degree of rotational control and is termed complete 3d orientation .
if the elliptically polarized pulse is polarized such that the major ( minor ) polarization axis is parallel to the static field the molecule is 3d aligned and only the most ( second most ) polarizable axis is 1d oriented .
furthermore , our calculations showed that complete 3d orientation can also be achieved using a linearly polarized laser pulse and a strong dc electric field arranged under an angle similar to the angle between the most polarizable axis and the dipole moment .
overall , it is clear that mixed - field orientation with appropriately polarized laser fields and weak dc fields is an effective tool for confining how complex molecules are turned in space .
even stronger control will be achievable in upcoming experiments combining strong dc electric fields and linearly - polarized laser fields .
the degree of angular control demonstrated provides excellent prospects for the recording of molecular movies of complex molecules using ion- , electron- , or photon - imaging experiments .
we are grateful to frank jensen for calculating the structure , dipole moment , and polarizability components of cpc . this work has been supported by the excellence cluster `` the hamburg center for ultrafast imaging structure , dynamics and control of matter at the atomic scale '' of the deutsche forschungsgemeinschaft , including the mildred dresselhaus award for r.g.f ..
the work was supported by the danish council for independent research ( natural sciences ) , the lundbeck foundation , and the carlsberg foundation .
financial support by the spanish project fis2011 - 24540 ( micinn ) , the grants p11-fqm-7276 and fqm-4643 ( junta de andaluca ) , and the andalusian research group fqm-207 is gratefully appreciated .
j.j.o . acknowledges the support of me under the program fpu .
part of this work was done while r.g.f .
was visitor at the kavli institute for theoretical physics , university of california at santa barbara within the program `` fundamental science and applications of ultracold polar molecules '' and she gratefully acknowledges partial financial support from the national science foundation grant no .
nsf phy11 - 25915 .
36ifxundefined [ 1 ] ifx#1 ifnum [ 1 ] # 1firstoftwo secondoftwo ifx [ 1 ] # 1firstoftwo secondoftwo `` `` # 1''''@noop [ 0]secondoftwosanitize@url [ 0 ]
+ 12$12 & 12#1212_12%12@startlink[1]@endlink[0]@bib@innerbibempty link:\doibase 10.1126/science.1169183 [ * * , ( ) ] link:\doibase 10.1039/c0cp01585 g [ * * , ( ) ] link:\doibase 10.1088/0034 - 4885/74/9/096101 [ * * , ( ) ] link:\doibase 10.1063/1.4719816 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.109.133202 [ * * , ( ) ] link:\doibase 10.1146/annurev - physchem-032511 - 143720 [ * * , ( ) ] link:\doibase 10.1146/annurev - physchem-032511 - 143708 [ * * , ( ) ] link:\doibase 10.1103/revmodphys.75.543 [ * * , ( ) ] @noop link:\doibase 10.1063/1.479917 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.90.083001 [ * * , ( ) ] link:\doibase 10.1080/01442350600847746 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.102.023001 [ * * , ( ) ] link:\doibase 10.1038/nphys1225 [ * * , ( ) ] link:\doibase 10.1088/1367 - 2630/11/10/105040 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.85.2470 [ * * , ( ) ] link:\doibase 10.1103/physreva.72.063401 [ * * , ( ) ] link:\doibase 10.1039/b910423b [ * * , ( ) ] link:\doibase 10.1103/physrevlett.92.198102 [ * * , ( ) ] link:\doibase 10.1038/nphys1666 [ * * , ( ) ] link:\doibase 10.1063/1.3194287 [ * * , ( ) ] link:\doibase 10.1039/c1cp21143a [ * * , ( ) ] link:\doibase 10.1063/1.481405 [ * * , ( ) ] @noop _ _ ( , ) link:\doibase 10.1063/1.3624774 [ * * , ( ) ] link:\doibase 10.1103/physreva.70.013403 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.99.143602 [ * * , ( ) ] _ _ , @noop ph.d .
thesis , , ( ) link:\doibase 10.1103/physrevlett.108.193001 [ * * , ( ) ] link:\doibase 10.1103/physreva.86.043437 [ * * , ( ) ] link:\doibase 10.1103/physreva.88.033416 [ * * , ( ) ] link:\doibase 10.1039/c1cp21195a [ * * , ( ) ] _ _ , @noop ph.d .
thesis , , ( ) _ _ , @noop ph.d .
thesis , , ( ) @noop `` , '' ( ) | the mixed - field orientation of an asymmetric - rotor molecule with its permanent dipole moment non - parallel to the principal axes of polarizability is investigated experimentally and theoretically .
we find that for the typical case of a strong , nonresonant laser field and a weak static electric field complete 3d orientation is induced if the laser field is elliptically polarized and if its major and minor polarization axes are not parallel to the static field .
for a linearly polarized laser field solely the dipole moment component along the most polarizable axis of the molecule is relevant resulting in 1d orientation even when the laser polarization and the static field are non parallel .
simulations show that the dipole moment component perpendicular to the most - polarizable axis becomes relevant in a strong dc electric field combined with the laser field .
this offers an alternative approach to 3d orientation by combining a linearly - polarized laser field and a strong dc electric field arranged at an angle equal to the angle between the most polarizable axis of the molecule and its permanent dipole moment .
= 1 |
we focus on rare - event simulation for addressing reliability problems corresponding to dynamic systems . to compute the rare event ( failure ) probability for a dynamic system , both input ( excitation ) and modeling uncertainties
should be quantified and propagated .
therefore , a probability model must be chosen to describe the uncertainty in the future input for the system and then a chosen deterministic or stochastic system model is used , preferably in conjunction with a probability model describing the associated modeling uncertainties , to propagate these uncertainties .
these input and system models define a probabilistic description of the system output ( response ) . for example
, the problem of interest might be to compute the small failure probability for a highly reliable dynamic system such as a bridge or building under uncertain future earthquake excitation , or for an aircraft under uncertain excitation by turbulence , using a finite - element structural model to approximate the dynamics of the system .
this model will usually be subject to both parametric uncertainty ( what values of the model parameters best represent the behavior of the system ? ) and non - parametric modeling uncertainty ( what are the effects of the aspects of the system behavior not captured by the dynamic model ? ) .
the treatment of input uncertainty has a long history in dynamic reliability theory and random vibrations , now more commonly called stochastic dynamics , but the treatment of modeling uncertainty is more recent .
usually the dynamic model of the system is represented by a time - dependent bvp ( boundary - value problem ) involving pdes ( partial differential equations ) or by a set of coupled odes ( ordinary differential equations ) .
typically the failure event is defined as any one of a set of performance quantities of interest exceeding its specified threshold over some time interval .
this is the so - called _ first - passage problem_. this challenging problem is characterized by a lack of analytical solutions , even for the simplest case of a single - degree - of - freedom linear oscillator subject to excitation that is modeled as a gaussian process .
approximate analytical methods exist that are usually limited in scope and their accuracy is difficult to assess in a given application @xcite .
semi - analytical methods from structural reliability theory such as form and sorm ( first- and second - order reliability methods ) @xcite can not be applied directly to the first - passage problem and are inapplicable , anyway , because of the high - dimensional nature of the discrete - time input history @xcite .
standard monte carlo simulation has general applicability but it is computationally very inefficient because of the low failure probabilities . as a consequence , advanced stochastic simulation schemes are needed .
we assume that initially there is a continuous - time deterministic model of the real dynamic system that consists of a state - space model with a finite - dimensional state @xmath1 at time @xmath2 and this is converted to a discrete - time state - space model using a numerical time - stepping method to give : @xmath3 where @xmath4 is the input at discrete time @xmath2 .
if the original model consists of a bvp with pdes describing a response @xmath5 where @xmath6 , then we assume that a finite set of basis functions @xmath7 is chosen ( e.g. global bases such as fourier and hermite polynomials or localized ones such as finite - element interpolation functions ) so that the solution is well approximated by : @xmath8 then a numerical method is applied to the bvp pdes to establish time - dependent equations for the vector of coefficients @xmath9 $ ] so that the standard state - space equation in ( [ model ] ) still applies .
for example , for a finite - element model of a structural system , @xmath7 would be local interpolation functions over the elements .
then , expressing the bvp in weak form , a weighted residual or galerkin method could be applied to give a state - space equation for the vector of coefficients @xmath10 @xcite .
suppose that a positive scalar performance function @xmath11 is a quantity of interest and that the rare event @xmath12 of concern is that @xmath11 exceeds a threshold @xmath13 over some discrete - time interval @xmath14 : @xmath15 where @xmath10 satisfies ( [ model ] ) .
the performance function @xmath11 may involve exceedance of multiple performance quantities of interest @xmath16 above their corresponding thresholds @xmath17 .
this can be accomplished by aggregating them using the _ max _ and _ min _ operators in an appropriate combination on the set of @xmath18 s ; for example , for a _
pure series _ failure criterion , where the threshold exceedance of _ any _ @xmath19 represents failure , one takes the aggregate performance failure criterion as @xmath20 , while for a _ pure parallel _ failure criterion , where _ all _ of the @xmath18 must exceed their thresholds before failure is considered to have occurred , one takes the aggregate performance failure criterion as @xmath21 .
if the uncertainty in the input time history vector @xmath22\in\mathbb{r}^d$ ] @xmath23 is quantified by a probability distribution for @xmath24 that has a pdf ( probability density function ) @xmath25 with respect to lebesgue integration over @xmath26 , then the rare - event probability is given by : @xmath27 the pdf @xmath25 is assumed to be readily sampled . although _ direct sampling _ from a high - dimensional pdf is not possible in most cases , multi - dimensional gaussians are an exception because the gaussian vector can be readily transformed so that the components are independent and the pdf is a product of one - dimensional gaussian pdfs . in many applications ,
the discrete - time stochastic input history is modeled by running discrete - time gaussian white noise through a digital filter to shape its spectrum in the frequency domain , and then multiplying the filtered sequence by an envelope function to shape it in the time domain , if it is non - stationary . the model in ( [ model ] ) may also depend on uncertain parameters @xmath28 which includes the initial values @xmath29 if they are uncertain .
then a prior pdf @xmath30 may be chosen to quantify the uncertainty in the value of vector @xmath31 .
some of the parameters may characterize the pdf for input @xmath24 which can then be denoted @xmath32 .
it is convenient to re - define vector @xmath24 to also include @xmath31 , then the new pdf @xmath25 is @xmath33 in terms of the previous pdfs .
we assume that model parameter uncertainty is incorporated in this way , so the basic equations remain the same as ( [ model ] ) , ( [ f ] ) , and ( [ pf ] ) . when model uncertainty is incorporated , the calculated @xmath34 has been referred to as the robust rare - event probability @xcite , meaning robust to model uncertainty , as in robust control theory .
the standard _ monte carlo simulation
_ method ( mcs ) is one of the most robust and straightforward ways to simulate rare events and estimate their probabilities .
the method was originally developed in @xcite for solving problems in mathematical physics . since
then mcs has been used in many applications in physics , statistics , computer science , and engineering , and currently it lays at the heart of all random sampling - based techniques @xcite . the basic idea behind mcs is to observe that the probability in ( [ pf ] ) can be written as an expectation : @xmath35 } p_\mathcal{e}=\int_{\mathbb{r}^{d}}{i_\mathcal{e}(u)p(u)du}=\mathbb{e}_p[i_\mathcal{e}],\ ] ] where @xmath36 is the indicator function of @xmath12 , that is @xmath37 if @xmath38 and @xmath39 otherwise , and @xmath40 is the dimension of the integral . recall that the strong law of large numbers @xcite states that if @xmath41 are independent and identically distributed ( i.i.d . )
samples of vector @xmath24 drawn from the distribution @xmath25 , then for any function @xmath42 with finite mean @xmath43 $ ] , the sample average @xmath44 converges to the true value @xmath43 $ ] as @xmath45 almost surely ( i.e. with probability 1 ) .
therefore , setting @xmath46 , the probability in ( [ pf = e[i ] ] ) can be estimated as follows : @xmath47 it is straightforward to show that @xmath48 is an _ unbiased _ estimator of @xmath34 with mean and variance : @xmath49&= \mathbb{e}_p\left[\frac{1}{n}\sum_{i=1}^ni_\mathcal{e}(u_i)\right]\\ & = \frac{1}{n}\sum_{i=1}^n\mathbb{e}_p[i_\mathcal{e}]=p_\mathcal{e},\\ \mathrm{var}_p[{p}_\mathcal{e}^{mcs}]&=\mathrm{var}_p\left[\frac{1}{n}\sum_{i=1}^ni_\mathcal{e}(u_i)\right]\\ & = \frac{1}{n^2}\sum_{i=1}^n\mathrm{var}_p[i_\mathcal{e}]=\frac{p_\mathcal{e}(1-p_\mathcal{e})}{n}. \end{split}\ ] ] furthermore , by the central limit theorem @xcite , as @xmath45 , @xmath48 is distributed asymptotically as gaussian with this mean and variance . _ frequentist interpretation of mcs : _ the frequentist interpretation of mcs focuses on the forward problem , arguing that if @xmath50 is large so that the variance of @xmath48 is relatively small , then the value @xmath51 based on ( [ mcsestimate ] ) for a specific set of @xmath50 samples @xmath52 drawn from @xmath25 should be close to the mean @xmath34 of @xmath48 .
the sample mean estimate @xmath51 is very intuitive and , in fact , simply reflects the frequentist definition of probability : @xmath51 is the ratio between the number of trials where the event @xmath12 occurred , @xmath53 , and the total number of trials @xmath50 . _
bayesian interpretation of mcs : _ the same mcs estimate @xmath51 has a simple bayesian interpretation ( e.g. @xcite ) , which focuses on the inverse problem for the specific set of @xmath50 samples @xmath52 drawn from @xmath25 . following the bayesian approach @xcite , the unknown probability @xmath34 is considered as a stochastic variable whose value in @xmath54 $ ] is uncertain .
the _ principle of maximum entropy _
@xcite leads to the uniform prior distribution for @xmath34 , @xmath55 , @xmath56 , which implies that all values are taken as equally plausible a priori .
since samples @xmath41 are i.i.d , the binary sequence @xmath57 is a sequence of bernoulli trials , and so for the forward problem , @xmath58 is distributed according to the binomial distribution with parameters @xmath50 and @xmath34 , @xmath59 .
therefore , for the set of @xmath50 samples , the likelihood function is @xmath60 . using _
bayes theorem _ , the posterior distribution for @xmath34 , @xmath61 , is therefore the beta distribution @xmath62 , i.e. @xmath63 where the beta function @xmath64 is the normalizing constant that equals @xmath65 here . the mcs estimate is the _
maximum a posteriori ( map ) estimate _ , which is the mode of the posterior distribution ( [ posterior ] ) and therefore the most probable value of @xmath34 a posteriori : @xmath66 notice that the posterior pdf in ( [ posterior ] ) gives a complete description of the uncertainty in the value of @xmath34 based on the specific set of @xmath50 samples of @xmath24 drawn from @xmath25 .
the posterior distribution in ( [ posterior ] ) is in fact the original bayes result @xcite , although bayes theorem was developed in full generality by laplace @xcite .
the standard mcs method for estimating the probability in ( [ pf ] ) is summarized in the following pseudo - code .
+ ' '' '' height 0.6pt ' '' '' * monte carlo simulation * ' '' '' ' '' '' ` input : ` @xmath67 @xmath50 , total number of samples . `
algorithm : ` set @xmath68 , number of trials where the event @xmath12 occurred . * for * @xmath69 * do * sample the input excitation @xmath70 . compute the system trajectory @xmath71 using the system model ( [ model ] ) with @xmath72 .
* if * @xmath73 @xmath74 * end if * * end for * ` output : ` @xmath75 @xmath76 , mcs estimate of @xmath34 @xmath75 @xmath77 , posterior pdf of @xmath34 ' '' '' _ assessment of accuracy of mcs estimate : _ for the frequentist interpretation , the _ coefficient of variation _ ( c.o.v . ) for the estimator @xmath48 given by ( [ mcsestimate ] ) , conditional on @xmath34 and @xmath50 , is given by ( [ mean_var ] ) : @xmath78}}{\mathbb{e}_p[{p}_\mathcal{e}^{mcs}]}=\sqrt{\frac{1-p_\mathcal{e}}{np_\mathcal{e}}}.\ ] ] this can be approximated by replacing @xmath34 by the estimate @xmath79 for a given set of @xmath50 samples
@xmath52 : @xmath80 for the bayesian interpretation , the posterior c.o.v . for the stochastic variable @xmath34 ,
conditional on the set of @xmath50 samples , follows from ( [ posterior ] ) : @xmath81}}{\mathbb{e}[{p}_\mathcal{e}|\widehat{n}_\mathcal{e},n]}\\ & = \frac{\sqrt{1-\frac{\widehat{n}_\mathcal{e}+1}{n+2}}}{\sqrt{(n+3)\left(\frac{\widehat{n}_\mathcal{e}+1}{n+2}\right)}}\longrightarrow\sqrt{\frac{1-\widehat{p}_\mathcal{e}^{mcs}}{n\widehat{p}_\mathcal{e}^{mcs}}}=\widehat{\delta}_n^{mcs } , \end{split}\ ] ] as @xmath45 .
therefore , the same expression @xmath82 can be used to assess the accuracy of the mcs estimate , even though the two c.o.v.s have distinct interpretations .
the approximation @xmath82 for the two c.o.v.s reveals both the main advantage of the standard mcs method and its main drawback .
the main strength of mcs , which makes it very robust , is that its accuracy does not depend on the geometry of the domain @xmath83 and its dimension @xmath84 .
as long as an algorithm for generating i.i.d .
samples from @xmath25 is available , mcs , unlike many other methods ( e.g. numerical integration ) , does not suffer from the `` curse of dimensionality . ''
moreover , an irregular , or even fractal - like , shape of @xmath12 will not affect the accuracy of mcs .
on the other hand , the serious drawback of mcs is that this method is not computationally efficient in estimating the _ small probabilities _
@xmath34 corresponding to _ rare events _ , where from ( [ cov ] ) , @xmath85 therefore , to achieve a prescribed level of accuracy @xmath86 , the required total number of samples is @xmath87 . for each sampled excitation @xmath88 , a system analysis usually computationally very intensive is required to compute the corresponding system trajectory @xmath89 and to check whether @xmath88 belongs to @xmath12 .
this makes mcs excessively costly and inapplicable for generating rare events and estimating their small probabilities .
nevertheless , essentially all sampling - based methods for estimation of rare event probability are either based on mcs ( e.g. importance sampling ) or have it as a part of the algorithm ( e.g. subset simulation ) .
the _ importance sampling _ ( is ) method belongs to the class of _ variance reduction techniques _ that aim to increase the accuracy of the estimates by constructing ( sometimes biased ) estimators with a smaller variance @xcite .
it seems it was first proposed in @xcite , soon after the standard mcs method appeared .
the inefficiency of mcs for rare event estimation stems from the fact that most of the generated samples @xmath90 do not belong to @xmath12 so that the vast majority of the terms in the sum ( [ mcsestimate ] ) are zero and only very few ( if any ) are equal to one . the basic idea of is is to make use of the information available about the rare event @xmath12 to generate samples that lie more frequently in @xmath12 or in the _ important region _ @xmath91 that accounts for most of the probability content in ( [ pf ] ) . rather than estimating @xmath34 as an average of many 0 s and very few 1 s like in @xmath51 , is seeks to reduce the variance by constructing an estimator of the form @xmath92 , where @xmath93 is an appreciable fraction of @xmath50 and the @xmath94 are small but not zero , ideally of the same order as the target probability , @xmath95 .
specifically , for an appropriate pdf @xmath96 on the excitation space @xmath26 , the integral in ( [ pf = e[i ] ] ) can be re - written as follows : @xmath97 .
\end{split}\ ] ] the is estimator is now constructed similarly to ( [ mcsestimate ] ) by utilizing the law of large numbers : @xmath98 where @xmath41 are i.i.d .
samples from @xmath96 , called the _ importance sampling density _ ( isd ) , and @xmath99 is the _ importance weight _ of sample @xmath88 .
the is estimator @xmath100 converges almost surely as @xmath45 to @xmath34 by the strong law of large numbers , provided that the support of @xmath96 , i.e. the domain in @xmath26 where @xmath101 , contains the support of @xmath102 .
intuitively , the latter condition guarantees that all points of @xmath12 that can be generated by sampling from the original pdf @xmath25 , can also be generated by sampling from the isd @xmath96 .
note that if @xmath103 , then @xmath104 and is simply reduces to mcs , @xmath105 . by choosing the isd @xmath96 appropriately , is aims to obtain an estimator with a smaller variance . the is estimator @xmath100 is also _ unbiased _ with mean and variance : @xmath106&=\mathbb{e}_q\left[\frac{1}{n}\sum_{i=1}^ni_\mathcal{e}(u_i)w(u_i)\right]\\ & = \frac{1}{n}\sum_{i=1}^n\mathbb{e}_q\left[\frac{i_\mathcal{e}p}{q}\right]=p_\mathcal{e},\\ \mathrm{var}_q[{p}_\mathcal{e}^{is}]&=\frac{1}{n^2}\sum_{i=1}^n\mathrm{var}_q\left[\frac{i_\mathcal{e}p}{q}\right]\\ & = \frac{1}{n}\left(\mathbb{e}_q\left[\frac{i_\mathcal{e}p^2}{q^2}\right]-p_\mathcal{e}^2\right ) . \end{split}\ ] ] the is method is summarized in the following pseudo - code
. + ' '' '' height 0.6pt ' '' '' * importance sampling * ' '' '' ' '' '' ` input : ` @xmath67 @xmath50 , total number of samples .
@xmath67 @xmath96 , importance sampling density . `
algorithm : ` set @xmath107 , counter for the number of samples in @xmath12 . * for * @xmath69 * do * sample the input excitation @xmath108 . compute the system trajectory @xmath71 using the system model ( [ model ] ) with @xmath72 . *
if * @xmath73 @xmath109 compute the importance weight of the @xmath110 sample in @xmath12 , @xmath111 .
* end if * * end for * @xmath112 , the total number of trials where the event @xmath12 occurred . `
output : ` @xmath75 @xmath113 , is estimate of @xmath34 . ' '' '' the most important task in applying is for estimating small probabilities of rare events is the construction of the isd , since the accuracy of @xmath114 depends critically on @xmath96 .
if the isd is `` good '' , then one can get great improvement in efficiency over standard mcs . if , however , the isd is chosen inappropriately so that for instance @xmath68 or the importance weights have a large variation
, then is will yield a very poor estimate .
both scenarios are demonstrated below in the section `` illustrative example '' .
it is straightforward to show that the _ optimal _ isd , which minimizes the variance in ( [ mean var is ] ) , is simply the original pdf @xmath25 conditional on the domain @xmath12 : @xmath115 indeed , in this case , all generated sample excitations satisfy @xmath116 , so their importance weights @xmath117 , and the is estimate @xmath118 .
moreover , just one sample ( @xmath119 ) generated from @xmath120 is enough to find the probability @xmath34 exactly .
note , however , that this is a purely theoretical result since in practice sampling from the conditional distribution @xmath121 is challenging , and , most importantly , it is impossible to compute @xmath120 : this would require the knowledge of @xmath34 , which is unknown .
nevertheless , this result indicates that the isd @xmath96 should be chosen as close to @xmath120 as possible .
in particular , most of the probability mass of @xmath96 should be concentrated on @xmath12 .
based on these considerations , several ad hoc techniques for constructing isds have been developed , e.g. variance scaling and mean shifting @xcite . in the special case of linear dynamics and gaussian excitation ,
an extremely efficient algorithm for estimating the rare - event probability @xmath34 in ( [ pf ] ) , referred to as isee ( _ importance sampling using elementary events _ ) , has been presented @xcite .
the choice of the isd exploits known information about each elementary event , defined as an outcrossing of the performance threshold @xmath13 in ( [ f ] ) at a specific time @xmath122 . the c.o.v .
of the isee estimator for @xmath50 samples of @xmath24 from @xmath25 is given by @xmath123 where the proportionality constant @xmath124 is close to @xmath125 , regardless of how small the value of @xmath34 .
in fact , @xmath124 decreases slightly as @xmath34 decreases , exhibiting the opposite behavior to mcs .
in general , it is known that in many practical cases of rare event estimation it is difficult to construct a good isd that leads to a low - variance is estimator , especially if the dimension of the uncertain excitation space @xmath26 is large , as it is in dynamic reliability problems @xcite . a geometric explanation as to why is in often inefficient in high dimensions is given in @xcite .
au @xcite has presented an efficient is method for estimating @xmath34 in ( [ pf ] ) for elasto - plastic systems subject to gaussian excitation . in recent years
, substantial progress has been made by tailoring the _ sequential importance sampling _ ( sis ) methods @xcite , where the isd is iteratively refined , to rare event problems .
sis and its modifications have been successfully used for estimating rare events in dynamic portfolio credit risk @xcite , structural reliability @xcite , and other areas .
the _ subset simulation _ ( ss ) method @xcite is an advanced stochastic simulation method for estimating rare events which is based on _ markov chain monte carlo _ ( mcmc ) @xcite .
the basic idea behind ss is to represent a very small probability @xmath34 of the rare event @xmath12 as a product of larger probabilities of `` more - frequent '' events and then estimate these larger probabilities separately . to implement this idea ,
let @xmath126 be a sequence of nested subsets of the uncertain excitation space starting from the entire space @xmath127 and shrinking to the target rare event @xmath128 .
by analogy with ( [ f ] ) , subsets @xmath129 can be defined by relaxing the value of the critical threshold @xmath13 : @xmath130 where @xmath131 . in the actual implementation of ss ,
the number of subsets @xmath132 and the values of intermediate thresholds @xmath133 are chosen adaptively . using the notion of conditional probability and exploiting the nesting of the subsets
, the target probability @xmath34 can be factorized as follows : @xmath134 an important observation is that by choosing the intermediate thresholds @xmath133 appropriately , the conditional events @xmath135 can be made more frequent , and their probabilities can be made large enough to be amenable to efficient estimation by mcs - like methods .
the first probability @xmath136 can be readily estimated by standard mcs : @xmath137 where @xmath138 are i.i.d .
samples from @xmath25 . estimating the remaining probabilities
@xmath139 , @xmath140 , is more challenging since one needs to generate samples from the conditional distribution @xmath141 , which , in general , is not a trivial task . notice that a sample @xmath24 from @xmath142 is one drawn from @xmath25 that lies in @xmath143 .
however , it is not efficient to use mcs for generating samples from @xmath142 : sampling from @xmath25 and accepting only those samples that belong to @xmath143 is computationally very expensive , especially at higher levels @xmath144 . in standard ss ,
samples from the conditional distribution @xmath142 are generated by the _ modified metropolis algorithm _
( mma ) @xcite which belongs to the family of mcmc methods for sampling from complex probability distributions that are difficult to sample from directly @xcite .
an alternative strategy _ splitting _ is described in the next section .
the mma algorithm is a component - wise version of the original metropolis algorithm @xcite .
it is specifically tailored for sampling from high - dimensional conditional distributions and works as follows .
first , without loss of generality , assume that @xmath145 , i.e. components of @xmath24 are independent .
this assumption is indeed not a limitation , since in simulation one always starts from independent variables to generate correlated excitation histories @xmath24 .
suppose further , that some vector @xmath146 is already distributed according to the target conditional distribution , @xmath147 .
mma prescribes how to generate another vector @xmath148 and it consists of two steps : 1 .
generate a `` candidate '' state @xmath149 as follows : first , for each component @xmath150 of @xmath149 , sample @xmath151 from the symmetric univariate _
proposal distribution _
@xmath152 centered on the @xmath153 component of @xmath154 , where symmetry means that @xmath155 ; then , compute the _ acceptance ratio _
@xmath156 ; finally , set @xmath157 2 .
accept or reject the candidate state @xmath149 : @xmath158 it can be shown that @xmath159 generated by mma is indeed distributed according to the target conditional distribution @xmath142 when @xmath154 is @xcite . for a detailed discussion of mma ,
the reader is referred to @xcite .
the procedure for generating conditional samples at level @xmath144 is as follows .
starting from a `` seed '' @xmath160 , one can now use mma to generate a sequence of random vectors @xmath138 , called a _
markov chain _
, distributed according to @xmath142 . at each step ,
@xmath161 is used to generate the next state @xmath162 .
note that although these mcmc samples are identically distributed , they are clearly not independent : the correlation between successive samples is due to the proposal pdfs @xmath163 at level @xmath144 that govern the generation of @xmath162 from @xmath164 .
nevertheless , @xmath138 can still be used for statistical averaging as if they were i.i.d , although with certain reduction in efficiency @xcite . in particular , similarly to ( [ p1 ] ) , the conditional probability @xmath139 , can be estimated as follows : @xmath165 to obtain an estimator for the target probability @xmath34 , it remains to multiply the mcs ( [ p1 ] ) and mcmc ( [ pi ] ) estimators of all factors in ( [ product ] ) .
in real applications , however , it is often difficult to rationally define the subsets @xmath166 in advance , since it is not clear how to specify the values of the intermediate thresholds @xmath133 . in ss , this is done adaptively .
specifically , let @xmath167 be the mcs samples from @xmath25 , @xmath168 be the corresponding trajectories from ( [ model ] ) , and @xmath169 be the resulting performance values .
assume that the sequence @xmath170 is ordered in non - increasing order , i.e. @xmath171 , renumbering the samples where necessary .
define the first intermediate threshold @xmath172 as follows : @xmath173 where @xmath174 is a chosen probability satisfying @xmath175 .
this choice of @xmath172 has two immediate consequences : first , the mcs estimate of @xmath176 in ( [ p1 ] ) is exactly @xmath174 , and , second , @xmath177 not only belong to @xmath178 , but also are distributed according to the conditional distribution @xmath179 .
each of these @xmath180 samples can now be used as mother seeds in mma to generate @xmath181 offspring , giving a total of @xmath182 samples @xmath183 .
since these seeds start in the stationary state @xmath179 of the markov chain , this mcmc method gives _ perfect sampling _
, i.e. no wasteful burn - in period is needed .
similarly , @xmath184 is defined as @xmath185 where @xmath186 are the ( ordered ) performance values corresponding to excitations @xmath187 . again by construction , the estimate ( [ pi ] ) gives @xmath188 , and @xmath189 .
the ss method proceeds in this manner until the target rare event @xmath12 is reached and is sufficiently sampled .
all but the last factor in ( [ product ] ) are approximated by @xmath174 , and the last factor @xmath190 , where @xmath191 is the number of samples in @xmath12 among @xmath192 .
the method is more formally summarized in the following pseudo - code . ' '' '' height 0.6pt ' '' '' * subset simulation * ' '' '' ' '' '' ` input : ` @xmath67 @xmath182 , number of samples per conditional level .
@xmath67 @xmath174 , level probability ; e.g. @xmath193 .
@xmath67 @xmath163 , proposal distributions ; e.g. @xmath194 .
` algorithm : ` set @xmath195 , number of conditional level .
set @xmath196 , number of the mcs samples in @xmath12 . sample the input excitations @xmath197 .
compute the corresponding trajectories @xmath198 .
* for * @xmath199 * do * * if * @xmath200 * do * @xmath201 * end if * * end for * * while * @xmath202 * do * @xmath203 , a new subset @xmath129 is needed .
sort @xmath204 so that @xmath205 .
define the @xmath206 intermediate threshold : @xmath207 .
* for * @xmath208 * do * using @xmath209 as a seed , use mma to generate @xmath181 additional states of a markov chain @xmath210 . * end for * renumber : @xmath211 .
compute the corresponding trajectories @xmath212 . * for * @xmath199
* do * * if * @xmath213 * do * @xmath214 * end if * * end for * * end while * @xmath215 , number of levels , i.e. subsets @xmath129 in ( [ filtration ] ) and ( [ subsets ] ) .
@xmath216 , total number of samples .
` output : ` @xmath75 @xmath217 , ss estimate of @xmath218 . ' '' '' implementation details of ss , in particular the choice of level probability @xmath174 and proposal distributions @xmath219 , are thoroughly discussed in @xcite .
it has been confirmed that @xmath193 proposed in the original paper @xcite is a nearly optimal value .
the choice of @xmath163 is more delicate , since the efficiency of mma strongly depends on the proposal pdf variances in a non - trivial way : proposal pdfs with both small and large variance tend to increase the correlation between successive samples , making statistical averaging in ( [ pi ] ) less efficient . in general , finding the optimal variance of proposal distributions is a challenging task not only for mma , but also for almost all mcmc algorithms .
nevertheless , it has been found in many applications that using @xmath194 , the gaussian distribution with mean @xmath220 and variance @xmath221 , yields good efficiency if @xmath222 and @xmath25 is a multi - dimensional gaussian with all variances equal to @xmath223 .
for an adaptive strategy for choosing @xmath163 , the reader is referred to @xcite ; for example , @xmath224 can be chosen so that the observed average acceptance rate in mma , based on a subset of samples at level @xmath144 , lies in the interval @xmath225 $ ] .
it can be shown @xcite that , given @xmath34 , @xmath174 , and the total number of samples @xmath50 , the c.o.v .
of the ss estimator @xmath226 is given by @xmath227 where @xmath228 and @xmath229 is approximately a constant that depends on the state correlation of the markov chain at each level .
numerical experiments show that @xmath230 gives a good approximation to the c.o.v . and that @xmath231 if the proposal variance @xmath232 for each level is appropriately chosen @xcite .
it follows from ( [ cov2 ] ) that @xmath233 for mcs , while for ss , @xmath234 .
this drastically different scaling behavior of the c.o.v.s with small @xmath34 directly exhibits the improvement in efficiency . to compare an advanced stochastic simulation algorithm directly with mcs , which is always applicable ( but not efficient ) for rare event estimation
, @xcite introduced the relative computation efficiency of an algorithm , @xmath235 , which is defined as the ratio of the number of samples @xmath236 required by mcs to the number of samples @xmath237 required by the algorithm for the same c.o.v .
the _ relative efficiency _ of ss is then @xmath239 for @xmath230 , @xmath240 , and @xmath193 . for rare events
, @xmath241 is very large , and , as expected , ss outperforms mcs ; for example , if @xmath242 , then @xmath243 .
in recent years , a number of modifications of ss have been proposed , including ss with splitting @xcite ( described in the next section ) , hybrid ss @xcite , two - stage ss @xcite , spherical ss @xcite , and ss with delayed rejection @xcite .
a bayesian post - processor for ss , which generalizes the bayesian interpretation of mcs described above , was developed in @xcite . in the original paper @xcite ,
ss was developed for estimating reliability of complex civil engineering structures such as tall buildings and bridges at risk from earthquakes .
it was applied for this purpose in @xcite and @xcite .
ss and its modifications have also been successfully applied to rare event simulation in fire risk analysis @xcite , aerospace @xcite , nuclear @xcite , wind @xcite and geotechnical engineering @xcite , and other fields .
a detailed exposition of ss on an introductory level and a matlab code implementing the above pseudo - code is given in @xcite .
for more advanced and complete reading , the fundamental monograph on ss @xcite is strongly recommended .
in the previously presented stochastic simulation methods , samples of the input and output discrete - time histories , @xmath244 and @xmath245 , are viewed geometrically as vectors @xmath24 and @xmath246 that define points in the vector spaces @xmath247 and @xmath248 , respectively . in the splitting method , however , samples of the input and output histories are viewed as trajectories defining paths of length @xmath249 in @xmath250 and @xmath251 , respectively .
samples that reach a certain designated subset in the input or output spaces at some time are treated as `` mothers '' and are then split into multiple offspring trajectories by separate sampling of the input histories subsequent to the splitting time .
these multiple trajectories can themselves subsequently be treated as mothers if they reach another designated subset nested inside the first subset at some later time , and so be split into multiple offspring trajectories .
this is continued until a certain number of the trajectories reach the smallest nested subset corresponding to the rare event of interest .
splitting methods were originally introduced by kahn and harris @xcite and they have been extensively studied ( for example , @xcite ) .
we describe splitting here by using the framework of subset simulation where the only change is that the conditional sampling in the nested subsets is done by splitting the trajectories that reach each subset , rather than using them as seeds to generate more samples from markov chains in their stationary state . as a result ,
only standard monte carlo simulation is needed , instead of mcmc simulation .
the procedure in @xcite is followed here to generate offspring trajectories at the @xmath206 level @xmath252 of subset simulation from each of the mother trajectories in @xmath129 constructed from samples from the previous level , except that we present it from the viewpoint of trajectories in the input space , rather than the output space .
therefore , at the @xmath206 level , each of the @xmath180 sampled input histories @xmath164 , @xmath208 , from the previous level that satisfy @xmath253 , as defined in ( [ subsets ] ) ( so the corresponding output history @xmath254 satisfies @xmath255 ) , are split at their first - passage time @xmath256 this means that the mother trajectory @xmath164 is partitioned as @xmath257 $ ] where @xmath258 $ ] and @xmath259 $ ] ; then a subtrajectory sample @xmath260 $ ] is drawn from @xmath261 where the last equation follows if one assumes independence of the @xmath262 ( although it is not necessary ) . also , @xmath263 . note that the new input sample @xmath264 $ ] also lies in @xmath129 since it has the subtrajectory @xmath265 in common with @xmath164 , which implies that the corresponding outputs at the first - passage time @xmath266 are equal : @xmath267 .
the offspring trajectory @xmath268 is a sample from @xmath25 lying in @xmath129 and so , like its mother @xmath164 , it is a sample from @xmath269 .
this process is repeated to generate @xmath181 such offspring trajectories from each mother trajectory , giving a total of @xmath270 input histories that are samples from @xmath269 at the @xmath206 level .
the pseudo - code for the splitting version of subset simulation is the same as the previously presented pseudo - code for the mcmc version except that the part describing the generation of conditional samples at level @xmath144 using the mma algorithm is replaced by : ' '' '' height 0.6pt ' '' '' * generation of conditional samples * + * at level @xmath144 with splitting * ' '' '' ' '' '' * for * @xmath208 * do * using @xmath271 as a mother trajectory , generate @xmath181 offspring trajectories by splitting of this input trajectory .
* end for * ' '' '' to generate the same number of samples @xmath182 at a level , the splitting version of subset simulation is slightly more efficient than the mcmc version using mma because when generating the conditional samples , the input offspring trajectories @xmath272 $ ] already have available the first part @xmath273 of the corresponding output trajectory @xmath274 $ ] .
thus , ( [ model ] ) need only be solved for @xmath275 starting from the final value of @xmath273 ( which corresponds to the first - passage time of the trajectory ) .
a disadvantage of the splitting version is that it can not handle parameter uncertainty in the model in ( [ model ] ) since the offspring trajectories must use ( [ model ] ) with the same parameter values as their mothers .
furthermore , the splitting version applies only to dynamic problems , as considered here .
the mcmc version of subset simulation can handle parameter uncertainty and is applicable to both static and dynamic uncertainty quantification problems .
ching , au and beck @xcite discuss the statistical properties of the estimators corresponding to ( [ p1 ] ) and ( [ pi ] ) when the sampling at each level is done by the trajectory splitting method .
they show that as long as the conditional probability in subset simulation satisfies @xmath276 , the coefficient of variation for @xmath34 when estimating it by ( [ product ] ) and ( [ pi ] ) is insensitive to @xmath174 .
ching , beck and au @xcite also introduce a hybrid version of subset simulation that combines some advantages of the splitting and mcmc versions when generating the conditional samples @xmath277 at each level .
it is limited to dynamic problems because of the splitting but it can handle parameter uncertainty through using mcmc .
all three variants of subset simulation are applied to a series of benchmark reliability problems in @xcite ; their results imply that for the same computational effort in the dynamic benchmark problems , the hybrid version gives slightly better accuracy for the rare - event probability than the mcmc version . for a comparison between these results and
those of other stochastic simulation methods that are applied to some of the same benchmark problems ( e.g. spherical subset simulation , auxiliary domain method and line sampling ) , the reader may wish to check @xcite .
to illustrate mcs , is , and ss with mcmc and splitting for rare event estimation , consider the following forced lorenz system of ordinary differential equations : @xmath278 where @xmath279 defines the system state at time @xmath2 and @xmath280 is the external excitation to the system .
if @xmath281 , these are the original equations due to e. n. lorenz that he derived from a model of fluid convection @xcite . in this example , the three parameters @xmath282 , and @xmath13 are set to @xmath283 , @xmath284 , and @xmath285 .
it is well - know ( e.g. @xcite ) that in this case , the lorenz system has three unstable equilibrium points , one of which is @xmath286 that lies on one `` wing '' of the `` butterfly '' attractor .
let @xmath287 be the initial condition , and @xmath10 be the corresponding solution .
lorenz showed @xcite that the solution of ( [ lorenz1],[lorenz2],[lorenz3 ] ) with @xmath281 always ( for any @xmath2 ) stays inside the bounding ellipsoid @xmath288 : @xmath289 ) enclosed in the bounding ellipsoid @xmath288 ( top ) and the corresponding response function @xmath290 ( bottom ) , where @xmath291 $ ] , @xmath292 .
the right top panel shows the solution of the forced lorenz system ( @xmath293 ) that corresponds to an excitation @xmath294 .
as it is clearly seen , this solution leaves the ellipsoid @xmath288 .
according to the response function @xmath290 shown in the right bottom panel , this first - passage event happens around @xmath295 . ]
suppose that the system is now excited by @xmath296 , where @xmath297 is the standard brownian process ( gaussian white noise ) and @xmath124 is some scaling constant .
the uncertain stochastic excitation @xmath280 makes the corresponding system trajectory @xmath10 also stochastic .
let us say that the event @xmath12 occurs if @xmath10 leaves the bounding ellipsoid @xmath288 during the time interval of interest @xmath298 $ ] .
the discretization of the excitation @xmath24 is obtained by the standard discretization of the brownian process : @xmath299 where @xmath300s is the sampling interval , @xmath301 , and @xmath302 are i.i.d .
standard gaussian random variables .
the target domain @xmath83 is then @xmath303 where the system response @xmath304 at time @xmath305 is @xmath306 figure [ fig1 ] shows the solution of the unforced lorenz system ( with @xmath307 so @xmath308 ) , and an example of the solution of the forced system ( with @xmath293 ) that corresponds to excitation @xmath294 ( slightly abusing notation , @xmath309 means that the corresponding gaussian vector @xmath310 ) .
_ monte carlo simulation : _ for @xmath293 , figure [ fig2 ] shows the probability @xmath34 of event @xmath12 as a function of @xmath311 estimated using standard mcs : @xmath312 where @xmath313 are i.i.d .
samples from the standard @xmath84-dimensional gaussian pdf @xmath314 . for each value of @xmath311 , @xmath315 samples were used .
when @xmath316 the accuracy of the mcs estimate ( [ eq : mcs_ex ] ) begins to degenerate since the total number of samples @xmath50 becomes too small for the corresponding target probability .
moreover , for @xmath317 , none of the @xmath50 generated mcs samples belong to the target domain @xmath12 , making the mcs estimate zero .
figure [ fig2 ] shows , as expected , that @xmath34 is an increasing function of @xmath311 , since the more time the system has , the more likely its trajectory eventually penetrates the boundary of ellipsoid @xmath288 . of event @xmath12 where @xmath293 as a function of duration time @xmath311 . for each value of @xmath318 $ ] ,
@xmath315 samples were used in mcs and @xmath319 samples per conditional level were used in the two versions of ss . the mcs and ss / splitting estimates for @xmath34 are zero for @xmath317 and @xmath320 , respectively .
the bottom panel shows the total computational effort automatically chosen by both ss algorithms . ] _ importance sampling : _ is is a variance reduction technique and , as it was discussed in previous sections , its efficiency critically depends on the choice of the isd @xmath321 . usually some geometric information about the target domain @xmath12 is needed for constructing a good isd .
to get some intuition , figure [ fig3 ] shows the domain @xmath12 for two lower dimensional cases : @xmath322 , @xmath323 ( @xmath324 ) and @xmath325 , @xmath323 ( @xmath326 ) .
notice that in both cases , @xmath12 consists of two well separated subsets , @xmath327 , which are approximately symmetric about the origin .
this suggests that a good isd must be a mixture of two distributions @xmath328 and @xmath329 , that effectively sample @xmath330 and @xmath331 , @xmath332 in two dimensional case @xmath324 , where @xmath322 , @xmath323 , and @xmath333 .
@xmath315 samples were generated and marked by red circles ( respectively , green dots ) if they do ( respectively , do not ) belong to @xmath12 .
right panel : the same as on the left panel but with @xmath326 and @xmath325 . ] in this example , three different isds , denoted @xmath334 and @xmath335 , are considered : case 1 : : : @xmath336 , where @xmath337 .
that is , we first generate a sample @xmath338 and then take isd @xmath339 as the mixture of gaussian pdfs centered at @xmath340 and @xmath341 .
case 2 : : : @xmath342 , where @xmath343 is obtained as follows .
first we generate @xmath344 samples from @xmath314 , and define @xmath343 to be the sample in @xmath12 with the smallest norm .
sample @xmath343 can be interpreted as the `` best representative '' of @xmath330 ( or @xmath331 ) , since @xmath345 has the largest ( among generated samples ) value .
we then take isd @xmath346 as the mixture of gaussian pdfs centered at @xmath343 and @xmath347 .
case 3 : : : to illustrate what happens if one ignores the geometric information about two components of @xmath12 , we choose @xmath348 , as given in case 2 .
let @xmath322 and @xmath333 .
the dimension of the uncertain excitation space is then @xmath349 .
table [ tab1 ] shows the simulation results for the above three cases as well as for standard mcs .
the is method with @xmath339 , on average , correctly estimates @xmath34
. however the c.o.v . of the estimate is very large , which results in large fluctuations of the estimate in independent runs .
is with @xmath346 works very well and outperforms mcs : the c.o.v .
is reduced by half .
finally , is with @xmath335 completely misses one component part of the target domain @xmath12 , and the resulting estimate is about half of the correct value .
note that the c.o.v . in this case
is very small , which is very misleading .
lll & @xmath350 & @xmath351 + mcs & @xmath352 & 17@xmath353 + is @xmath339 & @xmath354 & 132.4@xmath353 + is @xmath346 & @xmath352 & 8.3@xmath353 + is @xmath335 & @xmath355 & 5.5@xmath353 + as a function of duration time @xmath311 .
solid red and dashed blue curves correspond to mcs and is with @xmath346 , respectively . in this example ,
@xmath333 and @xmath315 samples for each value of @xmath311 is used .
it is clearly visible how the is estimate degenerates as the dimension @xmath84 goes from 10 ( @xmath322 ) to 100 ( @xmath356 ) . ]
it was mentioned in previous sections that is is often not efficient in high dimensions because it becomes more difficult to construct a good isd @xcite . to illustrate this effect , is with @xmath346 was used to estimate @xmath34 for a sequence of problems where the total duration time gradually grows from @xmath322 to @xmath356 .
this results in an increase of the dimension @xmath84 of the underlying uncertain excitation space from 10 to 100 .
figure [ fig4 ] shows how the is estimate degenerates as the dimension @xmath84 of the problem increases .
while is is accurate when @xmath349 ( @xmath322 ) , it strongly underestimates the true value of @xmath34 as @xmath84 approaches @xmath357 ( @xmath356 ) . _
subset simulation : _ ss is a more advanced simulation method and , unlike is , it does not suffer from the curse of dimensionality . for @xmath293 ,
figure [ fig2 ] shows the estimate of the target probability @xmath34 as a function of @xmath311 using ss with mcmc and splitting . for each value of @xmath311
, @xmath319 samples were used in each conditional level in ss .
unlike mcs , ss is capable of efficiently simulating very rare events and estimating their small probabilities .
the total computational effort , i.e. the total number @xmath50 of samples automatically chosen by ss , is shown in the bottom panel of figure [ fig2 ] .
note that the larger the value of @xmath34 , the smaller the number of conditional levels in ss , and , therefore , the smaller the total number of samples @xmath50 .
the total computational effort in ss is thus a decreasing function of @xmath311 . in this example
, the original mcmc strategy @xcite for generating conditional samples outperforms the splitting strategy @xcite that exploits the causality of the system : while the ss / mcmc method works even in the most extreme case ( @xmath358 ) , the ss / splitting estimate for @xmath34 becomes zero for @xmath320 .
this chapter examines computational methods for rare - event simulation in the context of uncertainty quantification for dynamic systems that are subject to future uncertain excitation modeled as a stochastic process .
the rare events are assumed to correspond to some time - varying performance quantity exceeding a specified threshold over a specified time duration , which usually means that the system performance fails to meet some design or operation specifications . to analyze the reliability of the system against this performance failure , a computational model for the input - output behavior of the system
is used to predict the performance of interest as a function of the input stochastic process discretized in time .
this dynamic model may involve explicit treatment of parametric and non - parametric uncertainties that arise because the model only approximately describes the real system s behavior , implying that there are usually no true values of the model parameters and the accuracy of its predictions are uncertain . in the engineering literature , the mathematical problem to be solved numerically for the probability of performance failure , commonly called the failure probability , is referred to as the first - passage reliability problem .
it does not have an analytical solution and numerical solutions must face two challenging aspects : 1 . the vector representing the time - discretized stochastic process that models the future system excitation lies in an input space of high dimension ; 2 . the dynamic systems of interest are assume to be highly reliable so that their performance failure is a rare event , that is , the probability of its occurrence , @xmath34 , is very small . as a result , standard monte carlo simulation and importance sampling methods are not computationally efficient for first - passage reliability problems . on the other hand ,
subset simulation has proved to be a general and powerful method for numerical solution of these problems .
like mcs , it is not affected by the dimension of the input space and for a single run , it produces a plot of @xmath34 vs threshold @xmath13 covering @xmath359 $ ] , where @xmath132 is the number of levels used . for a critical appraisal of methods for first - passage reliability problems in high dimensions ,
the reader may wish to check schuller et al @xcite .
several variants of subset simulation have been developed motivated by the goal of further improving the computational efficiency of the original version , although the efficiency gains , if any , are modest .
all of them have an accuracy described by a coefficient of variation for the estimate of the rare - event probability that depends on @xmath360 rather than @xmath361 as in standard monte carlo simulation .
for all methods covered in this section , the dependence of this coefficient of variation on the number of samples @xmath50 is proportional to @xmath362 .
therefore , in the case of very low probabilities , @xmath34 , it still requires thousands of simulations ( large @xmath50 ) of the response time history based on a dynamic model as in ( [ model ] ) in order to get acceptable accuracy . for complex models ,
this computational effort may be prohibitive .
one approach to reduce the computational effort when estimating very low rare - event probabilities is to utilize additional information about the nature of the problem for specific classes of reliability problems ( e.g. @xcite ) .
another more general approach is to construct surrogate models ( meta - models ) based on using a relatively small number of complex - model simulations as training data .
the idea is to use a trained surrogate model to rapidly calculate an approximation of the response of the complex computational model as a substitute when drawing new samples .
various methods for constructing surrogate models have been applied in reliability engineering , including response surfaces @xcite , support vector machines @xcite , neural networks @xcite , and gaussian process modeling ( kriging ) @xcite .
the latter method is a particularly powerful one because it also provides a probabilistic assessment of the approximation error .
it deserves further exploration , especially with regard to the optimal balance between the accuracy of the surrogate model as a function of the number of training samples from the complex model , and the accuracy of the estimate of the rare - event probability as a function of the total number of samples from both the complex model and the surrogate model .
ching , j. , au , s.k .
, beck , j.l . : reliability estimation for dynamical systems subject to stochastic excitation using subset simulation with splitting . comput .
methods appl .
194 , 1557 - 1579 ( 2005 ) .
jalayer , f. , beck , j.l . :
effects of two alternative representations of ground - motion uncertainty on probabilistic seismic demand assessment of structures .
earthquake eng . and structural dynamics 37 , 61 - 79 ( 2008 ) .
katafygiotis , l.s . , cheung , s.h .
: a two - stage subset simulation - based approach for calculating the reliability of inelastic structural systems subjected to gaussian random excitations .
methods appl .
194 , 1581 - 1595 ( 2005 ) .
pellissetti , m.f . ,
schuller , g.i . ,
pradlwarter , h.j .
, calvi , a. , fransen , s. , klein , m. : reliability analysis of spacecraft structures under static and dynamic loading .
computers @xmath363 structures 84 , 1313 - 1325 ( 2006 ) .
taflanidis , a.a . ,
beck , j.l . : analytical approximation for stationary reliability of certain and uncertain linear dynamic systems with higher dimensional output .
earthquake engineering and structural dynamics 35 , 1247 - 1267 ( 2006 ) .
valdebenito , m.a .
, pradlwarter , h.j . , schuller , g.i . : the role of the design point for calculating failure probabilities in view of dimensionality and structural nonlinearities .
structural safety 32 , 101 - 111 ( 2010 ) .
zuev , k. : subset simulation method for rare event estimation : an introduction . in : m. beer
et al ( eds . ) encyclopedia of earthquake engineering .
springer - verlag berlin heidelberg , ( 2015 ) .
available on - line at http://arxiv.org/abs/1505.03506 .
zuev , k.m . ,
beck , j.l .
katafygiotis , l.s . : bayesian post - processor and other enhancements of subset simulation for estimating failure probabilities in high dimensions .
computers @xmath363 structures 92 - 93 , 283 - 296 ( 2012 ) . | rare events are events that are expected to occur infrequently , or more technically , those that have low probabilities ( say , order of @xmath0 or less ) of occurring according to a probability model . in the context of uncertainty quantification , the rare events often correspond to failure of systems designed for high reliability , meaning that the system performance fails to meet some design or operation specifications .
as reviewed in this section , computation of such rare - event probabilities is challenging .
analytical solutions are usually not available for non - trivial problems and standard monte carlo simulation is computationally inefficient .
therefore , much research effort has focused on developing advanced stochastic simulation methods that are more efficient . in this section ,
we address the problem of estimating rare - event probabilities by monte carlo simulation , importance sampling and subset simulation for highly reliable dynamic systems . |
a group action is one of the most important tool in the study of mathematical objects .
finite group actions in symplectic geometry appear in various contexts , such as mirror symmetry , singularity theory , etc . in this paper
, we will construct finite group actions on lagrangian floer theory .
lagrangian floer homology ( developed by floer@xcite , oh@xcite , fukaya - oh - ohta - ono @xcite ) has been a powerful tool , which revolutionized the study of symplectic topology .
furthermore , its categorical versions , fukaya category ( fukaya @xcite ) and fukaya - seidel category ( seidel @xcite , @xcite ) , are the main players of homological mirror symmetry as envisioned by kontsevich @xcite .
when a symplectic manifold as well as a lagrangian submanifold admits a finite group action , it is indeed quite natural to expect that such an action exists on lagrangian floer theory .
but there are several technical difficulties to define such an action as we will explain below . in the case of @xmath3-actions on exact symplectic manifolds
, seidel has defined @xmath3-equivariant fukaya categories @xcite .
another known case is when a group acts freely on the set of connected components of a given exact lagrangian submanifold in an exact symplectic manifold .
in such a case , it is relatively easy to define a group action , since one can use pull - back data ( such as spin structure , perturbation data , etc ) to make floer theory compatible with the given group action .
such construction was proposed by seidel @xcite in his study of mirror symmetry of quartic surfaces in an exact setting .
let us now describe our main results .
we will first explain the exact case , and later the case of general closed symplectic manifolds .
for an exact symplectic manifold @xmath4 , consider a @xmath2-invariant exact lagrangian submanifold @xmath5 , which means that a @xmath2-action preserves the lagrangian submanifold .
suppose that @xmath5 is spin and that its spin structure is @xmath2-invariant .
we show that the @xmath2-action on @xmath5 gives rise to a group cohomology class in @xmath6 , the spin profile of @xmath5 " , which is the obstruction to extending the group action to the spin bundle of @xmath5 .
the main observation is that if two lagrangian submanifolds have the same spin profile , then it is possible to define a group action on the lagrangian floer cohomology of the pair in the exact case .
thus for each cohomology class @xmath7 , an @xmath1-equivariant brane is defined as a @xmath2-invariant exact lagrangian submanifold whose associated spin profile is @xmath1 , with a @xmath2-invariant grading .
let @xmath4 be an exact symplectic manifold . for @xmath8 , an _
@xmath1-equivariant fukaya category _ can be defined as follows .
an object of @xmath1-equivariant fukaya category is given by an @xmath1-equivariant brane .
morphism between two @xmath1-equivariant branes @xmath9 are given by @xmath10 , which has a @xmath2-action .
there are @xmath11-operations @xmath12 which are compatible with the @xmath2-action .
i.e. for @xmath13 @xmath14 to be more precise , we should equip each @xmath1-equivariant brane with a @xmath2-equivariant flat complex vector bundle as in section [ sec : orbibundle ] .
the @xmath1-orbifolded fukaya category is defined to have the same set of objects as the @xmath1-equivariant fukaya category , but we take @xmath2-invariant parts of morphisms of the @xmath1-equivariant category . then , the @xmath1-orbifolded fukaya category naturally inherits an @xmath11-structure from the @xmath1-equivariant category from . as an application
, we define both equivariant and orbifolded fukaya - seidel categories of a @xmath2-invariant lefschetz fibration .
namely , for each @xmath7 , we define _
@xmath1-equivariant fukaya - seidel category of @xmath15 _ ( see definition [ def : fs ] , [ def : fs2 ] ) . by taking @xmath2-invariant parts on morphisms , and the induced @xmath11-structure
, we obtain the _
@xmath1-orbifolded fukaya - seidel _ category of @xmath16 .
let us turn our discussion to the case of a general closed symplectic manifold ( which is not exact ) with an effective action of a finite group @xmath2 .
consider a lagrangian submanifold @xmath17 ( @xmath18 ) , which has an induced @xmath2-action .
suppose that @xmath17 has a @xmath2-invariant spin structure , with a spin profile @xmath19 for @xmath18 .
fix a base path @xmath20 \to m$ ] such that @xmath21 for @xmath18 .
let @xmath22 be the energy zero subgroup @xmath23 of @xmath24 as in definition [ def : galpha ] . if @xmath25 , then the lagrangian floer filtered cochain complex of @xmath26 admits a @xmath22-action
. more generally , filtered @xmath11-bimodule of the pair @xmath27 admits a natural @xmath22-action .
suppose that @xmath28 is connected and let @xmath29 .
then @xmath30 , and we have a @xmath2-equivariant filtered @xmath11-bimodule quasi - isomorphism @xmath31 the method developed in this paper , together with the work of fukaya in @xcite should define equivariant fukaya categories in the general case . more detailed introduction and description are now in order . in this paper , we introduce the notion of a _ spin profile _ of a lagrangian submanifold with a @xmath2-action .
such a notion is deeply related to the orientation of the moduli space of @xmath32-holomorphic discs and the @xmath2-action on it .
recall that if a lagrangian submanifold is spin , a choice of spin structure determines canonical orientations of the moduli spaces of @xmath32-holomorphic discs ( see @xcite ) .
hence , we need to analyze the relation between the group action on @xmath5 and that of its spin structure .
we assume that the related spin structure is @xmath2-invariant , which means that for a principal spin bundle @xmath33 , the pull - back bundle @xmath34 via the action of @xmath35 is isomorphic to the original bundle @xmath36 .
in fact , the precise isomorphism between @xmath34 and @xmath36 ( as in @xmath2-equivariant sheaves ) , which can be considered as a lift of the @xmath37-action to the one on spin bundle , will be necessary to determine the sign of the @xmath37-action on the floer complex .
in general , these isomorphisms are not necessarily compatible with each other , and the @xmath2-action on @xmath5 does _ not _ lift to the one on the spin bundle .
hence , one can not expect that any orientation preserving action of a symplectic manifold gives rise to an action on its fukaya category .
a _ spin profile _ is defined as the group cohomology class in @xmath6 which measures the obstruction to extending the action of @xmath2 on a manifold to the one on its spin structure .
we show that if spin profiles of two lagrangian submanifolds coincide , then we are able to define a group action on lagrangian floer theory of a pair ( we need to develop @xmath2-novikov theory additionally for non - exact symplectic manifolds ) .
the main point is that the exact signs from failures of group actions on spin bundles cancel out for lagrangians with the same spin profiles .
this means that the equivariant fukaya category is defined for each choice of spin profiles in @xmath6 .
construction of the @xmath3-equivariant fukaya category by seidel @xcite , in fact , corresponds to the non - trivial class in @xmath38 .
we explain several properties of spin profiles , and also the change of spin profiles under the change of spin structures .
let @xmath4 be an exact symplectic manifold with an effective action of a finite group @xmath2 .
we fix a class @xmath8 , and consider @xmath1-equivariant branes as explained above .
it may happen that some of @xmath37-action send @xmath5 to @xmath39 which is disjoint from @xmath5 , in which case we can take @xmath40 to be the new @xmath2-invariant lagrangian submanifold .
we do not assume that @xmath5 is connected in the exact case .
in general , the equivariant fukaya category of @xmath41 should also have @xmath2-equivariant lagrangian immersions to @xmath41 . in the sequel of this paper @xcite
, we will investigate extensively such @xmath2-equivariant lagrangian immersions , and explore the relationship of @xmath42-equivariant immersions into @xmath43 with chen - ruan orbifold cohomology of @xmath44 $ ] .
in fact , one should consider a more general notion of orbifold embeddings @xcite , but they were shown to be equivalent to @xmath2-equivariant immersions for global quotients in @xcite , and hence it is enough to study the latter .
there is another technical issue , which are equivariant transversalities of the related moduli spaces of holomorphic discs .
namely , we would like to choose perturbation data equivariantly , so that we have strict @xmath2-equivariant @xmath11-operations . in the exact case ,
we follow seidel @xcite to achieve it with aid of an algebraic machinery . in the general case later , kuranishi structure ( as in @xcite , @xcite , @xcite ) will be used , which is in fact a device to achieve this kind of equivariant transversality we construct a version of donaldson - fukaya categories ( @xmath45 in definition [ donald ] ) , and produce a @xmath2-action on it from the weak @xmath2-action on the related floer complex ( see @xcite for the weak @xmath2-actions on morse complexes ) .
then , the homological perturbation type lemma of seidel ( lemma 4.3 of @xcite ) enables us to define the @xmath1-equivariant fukaya category for an exact symplectic manifold . by taking @xmath2-invariant parts on morphism spaces ,
the @xmath1-orbifolded fukaya category of @xmath44 $ ] is obtained .
this should be regarded as a first approximation of orbifold theory , as we should also add bulk deformation by twisted sectors as in @xcite . as an application
, we define _ the directed equivariant ( and the orbifolded ) fukaya - seidel category _ of a @xmath2-invariant lefschetz fibration @xmath46 , where @xmath41 is equipped with a @xmath2-action , and @xmath47 is @xmath2-invariant . in mirror symmetry , orbifolds and orbifolded landau - ginzburg models
appear naturally , and have been actively investigated .
our work provides the necessary framework to state homological mirror symmetry in these cases by defining appropriate orbifold fukaya categories .
we only deal with @xmath2-invariant lefschetz fibrations in this paper and leave the case of an orbifolded @xmath48-model which is not lefschetz fibration and of a @xmath2-equivariant @xmath48-model to the future investigation . for a closed symplectic manifold which is not exact ,
the lagrangian floer cohomology is the morse - novikov homology on path spaces between two lagrangian submanifolds .
hence , we need to develop @xmath2-novikov theory .
we first recall @xmath2-morse theory from our previous work @xcite . given a morse - smale function @xmath49 on a compact manifold @xmath41
, one can define the morse - smale - witten complex ( morse complex for short ) by counting gradient flow lines between critical points of @xmath50 .
if @xmath50 is @xmath2-invariant , we studied how to define a finite group action on the morse complex , whose homology of the @xmath2-invariant part is isomorphic to the singular homology of the quotient orbifold .
one of the main observation of @xcite is that one should consider not only the action on critical points , but also on the orientation spaces ( defined from their unstable manifolds ) .
now , morse - novikov theory ( novikov theory for short ) generalizes morse theory to the case of a general closed morse one form @xmath51 , instead of @xmath52 .
novikov theory is a kind of morse theory associated to a morse function on a certain novikov covering ( so that the pull - back of the given closed one form becomes exact ) , together with the action of the group ring of its deck transformation group . in this paper , we define finite group actions on novikov theory .
the main difficulty is that the base point chosen for novikov theory is _ not _ preserved by the group action .
we resolve this issue by introducing the subgroup @xmath53 of energy zero elements of @xmath2 , where @xmath35 is said to be energy zero if there is an energy zero path ( with respect to the action functional ) from the base point to its @xmath37-action image .
and we will construct an action of energy zero subgroup on the novikov complex using this energy zero path .
together with an appropriate orbifold novikov ring coefficient , taking @xmath53-invariant part defines @xmath2-novikov theory by using spin profiles and @xmath2-novikov theory , we can define group actions on lagrangian floer cohomology groups for non - exact cases as follows .
let @xmath4 be a closed symplectic manifold with an effective @xmath2-action .
fix @xmath8 and consider a lagrangian submanifold @xmath17 ( for @xmath18 ) whose spin profile is @xmath1 . for a base path @xmath24 between @xmath28 and @xmath54 , take an energy zero surface @xmath55 connecting @xmath24 and @xmath56 for @xmath37 in the energy zero subgroup @xmath23 ( definition [ def : galpha ] ) as an analogue of @xmath2-novikov theory .
we assume that @xmath57 for simplicity . for each generator of the floer complex of a pair @xmath27
, one can consider its orientation space , which is the determinant bundle of the associated cauchy - riemann operator .
this construction already appeared in @xcite , and seidel also used orientation spaces in exact cases .
then , the actual group action is defined to be an action on orientation spaces , where the action image is obtained by gluing the determinant bundles of @xmath55 ( more precisely , @xmath58 ) to the naive @xmath37-action image of a generator of the floer complex ( see definition [ def : aact ] ) . here
the lagrangian bundle data along the boundary @xmath59 will be given a specific spin structure , and @xmath60 is defined to be 0 ( resp .
1 ) if such spin structure is trivial ( resp .
non - trivial ) .
then , the following identity from proposition [ prop : compadm ] @xmath61 guarantees the cocycle condition of the group action , provided that the spin profiles of @xmath28 and @xmath54 agree with each other . in this way , we get a well - defined action of @xmath2 on the floer cochain complex @xmath62 between @xmath28 and @xmath54 . with equivariant kuranishi perturbations , we show that the floer differential as well as the @xmath11-bimodule structure on @xmath62 is compatible with the @xmath2-action .
therefore , the bimodule structure descends to the @xmath2-invariant part of @xmath62 , which should be interpreted as a morphism space between @xmath63 $ ] and @xmath64 $ ] in the fukaya category of the orbifold @xmath44 $ ] .
invariance of the floer cohomologies under @xmath2-invariant hamiltonians can be proven as well .
that is , if @xmath54 is the hamiltonian deformation of @xmath28 by a @xmath2-invariant hamiltonian function on @xmath41 , then the assumption @xmath65 always holds and moreover , @xmath66 and @xmath67 are @xmath2-equivariantly isomorphic as @xmath66-modules .
for the purpose of mirror symmetry , we need to consider complex flat line bundles on lagrangian submanifolds . with @xmath2-action , @xmath2-equivariant structures of such bundles should be also taken into account .
the constructions in this paper can be adapted to the cases with @xmath2-equivariant line bundles by following the standard constructions .
such a @xmath2-equivariant bundle may be considered as an orbi - bundle on the orbi - lagrangian @xmath68 $ ] .
hence an object of an orbifolded category @xmath69 is given by a @xmath2-invariant lagrangian ( or a @xmath2-equivariant immersion ) equipped with a @xmath2-equivariant line bundle on it . here
morphisms of @xmath70 are obtained by taking the @xmath2-invariant part of morphisms in the @xmath2-equivariant category .
recall that @xmath71-branes of mirror symmetry are given by derived categories of coherent sheaves .
when there is a group action , it is natural to work with @xmath2-equivariant sheaves .
the standard naming convention for equivariant sheaves are unfortunately different from ours .
namely , what we call an orbifolded category ( taking @xmath2-invariant parts on morphisms of an equivariant category ) , is called an equivariant category for the case of sheaves .
we will see later that if @xmath72 is given by a lagrangian torus fiber bundle equipped with an action of @xmath2 compatible with bundle structure , then its dual torus bundle @xmath73 naturally admits a @xmath2-action .
in general , if @xmath73 is given as a syz ( strominger - yau - zaslow ) mirror of @xmath72 , we conjecture that the ( derived ) orbifold category @xmath74 is equivalent the derived category of @xmath2-equivariant sheaves on @xmath73 .
when the mirror of @xmath72 is a landau - ginzburg model @xmath75 , it turns out that @xmath76 is invariant under the @xmath2-action on @xmath73 .
we conjecture that homological mirror symmetry is the equivalence between the @xmath2-equivariant matrix factorization category of @xmath76 and orbifolded fukaya category @xmath69 .
( see section [ mirrorsymm ] for the precise statements including spin profile data ) character group of @xmath2 acts on both sides of mirror by twisting the equivariant structures on bundles .
namely , given an orbifold bundle ( or @xmath2-equivariant bundle ) for @xmath77 , one can change @xmath2-equivariant structure by multiplying a character .
we expect that these actions are compatible with homological mirror symmetry .
we will look into toric cases , and present more conjectures and examples .
let us also mention that in a forthcoming work with siu - cheong lau , we will analyze mirror symmetry phenomena for @xmath2-equivariant immersions into finite group quotients , and construct a _ localized homological mirror functor _ for immersed lagrangian submanifolds .
+ we will assume throughout the paper that @xmath2 is a finite group . and
@xmath2-action are always assumed to be effective , proper , and orientation preserving , unless stated otherwise . a submanifold @xmath78 preserved by the @xmath2-action of @xmath41
is called @xmath2-invariant .
here , the action of @xmath35 sends @xmath5 to @xmath5 , but it may not fix @xmath5 pointwisely .
in fact , if @xmath5 is a lagrangian submanifold of @xmath41 and @xmath37 fixes every point of @xmath5 , then it is easy to see that @xmath37 fixes a neighborhood of @xmath5 in @xmath41 .
since the action on @xmath41 is assumed to be effective , such @xmath37 should be the identity element of @xmath2 .
let @xmath79 be a field containing @xmath80 , and we will use @xmath81 or @xmath82 for floer theory .
we use the following @xmath79-normalization convention due to seidel ( * ? ? ?
* section ( 12e ) ) .
this is because , roughly speaking , we are only interested in the signs , rather than magnitudes in orientations spaces . for any @xmath83-dimensional real vector space @xmath84 ,
we denote by @xmath85 the @xmath83-dimensional @xmath79-vector space generated by the two orientations of @xmath84 , with the relation that the sum of two generators is zero .
an @xmath79-linear isomorphism @xmath86 induces a @xmath79-linear isomorphism @xmath87 .
normalization will be also used in families , turning real line bundles into local coefficient systems with fiber @xmath79 .
it can be thought of as passing to the associated bundle via the homomorphism given by @xmath88 to simplify notations , we denote also by @xmath89 the diffeomorphism of @xmath41 induced by the action of @xmath37 ( sometimes written as @xmath90 ) . for @xmath91 ,
we denote by @xmath92 the induced map on the fundamental group @xmath93 .
we denote by @xmath94 , the concatenation of two paths @xmath95 and @xmath96 .
+ we have benefited very much from paul seidel s work @xcite , @xcite .
we also thank him for helpful comments to our work .
we are indebted to kenji fukaya for lemma [ lem : fukaya ] and valuable advices .
we are grateful to kaoru ono for the kind explanations of orientations in @xcite , and to siu - cheong lau for several helpful discussions especially on character group actions .
we would like to express our gratitude to yong - geun oh , kwokwai chan , hsian - hua tseng , mainak poddar , nick sheridan , hiroshige kajiura , gyoung - seog lee , hyung - seok shin , sangwook lee for their help .
this work was supported by the national research foundation of korea ( nrf ) grant funded by the korea government ( mest ) ( no .
in this section , we review novikov theory by considering @xmath2-novikov theory under the assumption [ assum : x_0fixed ] . in this case , @xmath2-novikov theory is given by an equivariant version of the standard one .
also , we will review @xmath2-morse theory from our previous work @xcite , which is needed for the proper definition of the group action on a novikov complex .
[ assum : x_0fixed ] there exists a point @xmath97 such that @xmath98 for all @xmath35 .
this assumption enables us to have a natural @xmath2-action on the fundamental group @xmath99 .
we will fix a base point to be @xmath100 , and simply write @xmath101 for @xmath102 in this section .
note that the @xmath2-action on @xmath103 is compatible with the group operations of @xmath103 .
i.e. the @xmath2-action gives a group homomorphism @xmath104 .
for a closed @xmath83-form @xmath51 on @xmath41 , we call @xmath51 a _ morse @xmath83-form _ if the graph of @xmath51 in @xmath105 is transversal to the zero section .
assume further that @xmath51 is @xmath2-invariant .
@xmath51 gives rise to a homomorphism @xmath106 , @xmath107 for @xmath108 .
note that @xmath109 is @xmath2-invariant as well , i.e. @xmath110 . let @xmath111 be the covering space associated to the kernel of @xmath109 , denoted by @xmath112 .
( i.e. for the universal covering @xmath113 of @xmath41 , we have @xmath114 ) .
we fix a base point @xmath115 in @xmath116 such that @xmath117 .
[ lem : coveraction ] there is an induced @xmath2-action on @xmath116 and @xmath118 is @xmath2-equivariant . since @xmath109 is @xmath2-invariant , @xmath112 is preserved by @xmath119 .
hence for any @xmath35 , @xmath120 thus , the composition map @xmath121 has a unique lift @xmath122 such that @xmath123 .
the lemma , then , follows from the standard covering theory .
the deck transformation group @xmath124 of the covering @xmath125 is given by @xmath126 , and the novikov ring @xmath127}$ ] is a completion of the group algebra of @xmath124 : @xmath128 } = \left\ { \left .
\sum a_i h_i \ , \right| \ , a_i \in r , \ , h_i \in \gamma,\,\ , |\{i : a_i \neq 0 , i_\eta ( h_i ) < c\ } | < \infty \,\ , \forall c \in { \mathbb{r}}\right\},\ ] ] where @xmath79 is a commutative ring with unity
the induced @xmath2-action on @xmath129 is trivial . since @xmath109 is @xmath2-invariant , @xmath2 acts on @xmath129
however , for @xmath108 , @xmath130 where @xmath131 .
we have two group actions on @xmath116 , which are the action of deck transformation group @xmath124 and the induced action of @xmath2 .
the following lemma will play an important role later .
[ ggamma ] the @xmath124-action and the @xmath2-action on @xmath116 commute with each other .
namely for each @xmath35 , and @xmath132 , we have @xmath133 we first work over the universal cover @xmath134 .
note that we can lift the @xmath2-action on @xmath41 to the @xmath2-action on the universal cover @xmath134 similar as in lemma [ lem : coveraction ] . for a deck transformation @xmath135 of @xmath134
, we denote the corresponding loop in @xmath101 by @xmath136 .
the action of @xmath37 takes the loop @xmath136 to the loop @xmath137 , and we claim that @xmath138 for the expression @xmath139 to make sense , we first check that the composition @xmath140 is a deck transformation of @xmath141 : @xmath142 to prove the claim , it is enough to compare actions of both sides in on the fiber @xmath143 since the covering is regular .
take any point @xmath144 in the fiber @xmath143 .
we consider the path @xmath145 which is the lift of the loop @xmath136 starting from @xmath144 . then , its end point gives @xmath146 hence , @xmath147 is a loop based at @xmath100 , and we get a corresponding deck transformation , say @xmath148 .
let @xmath149 be the lift of @xmath137 starting from @xmath144 .
then , @xmath150 sends @xmath144 to @xmath151 . to get
, we have to show that @xmath152 let @xmath153 be the lifting of @xmath136 which begins at @xmath154 instead of @xmath144 .
@xmath155 is the lifting of @xmath147 which starts from @xmath144 because @xmath156 and @xmath157 now @xmath158 which proves the claim . by the previous lemma ,
@xmath137 and @xmath136 are the same elements modulo @xmath159 . projecting down from @xmath134 to @xmath116 , we obtain for @xmath132 that @xmath160 on @xmath116 .
we will give more geometric proof of lemma [ ggamma ] in subsection [ decktrans ] .
the @xmath2-action on @xmath124 induces the trivial @xmath2-action on @xmath127}$ ] as well .
note that @xmath161 is exact by construction .
so , we can find a smooth function @xmath162 such that @xmath163 from the transversality assumption on @xmath51 , @xmath164 is a morse function on @xmath116 .
[ invmorse ] @xmath164 is @xmath2-invariant .
since @xmath51 is @xmath2-invariant and @xmath47 is @xmath2-equivariant , we have for each @xmath35 , @xmath165 as we assumed that @xmath41 is connected , there is a constant @xmath166 such that for all @xmath167 , @xmath168 putting @xmath169 in the place of @xmath170 , we get @xmath171 since @xmath172 where @xmath173 is the order of @xmath35 . by summing up
, we have @xmath174 and , hence @xmath175 .
we conclude that @xmath176 .
in fact , with the assumption [ assum : x_0fixed ] , we have @xmath177 .
but the above argument works without the assumption [ assum : x_0fixed ] .
[ lem : isolocalgroup ] the isotropy group @xmath178 is the same as @xmath179 for all @xmath167 .
it is easy to see that @xmath180 . for @xmath181 , @xmath182 and @xmath183
may not be homotopic ( regarding @xmath183 as a path from @xmath100 to @xmath170 ) , but they differ by the @xmath184-action .
more precisely @xmath185 sends @xmath182 to @xmath183 . before constructing the @xmath2-novikov complex
, we briefly recall the construction of morse complexes for global quotients from @xcite ( using the language of orientation spaces in @xcite ) . for the global quotient
orbifold @xmath44 $ ] , consider a @xmath2-invariant morse - smale function @xmath49 , and denote by @xmath186 , the induced map on the quotient .
@xmath2-action sends a critical point of @xmath50 to another critical point , but this naive group action on @xmath187 is not the right action to consider as observed in @xcite . for the critical point @xmath188 , consider the unstable manifold @xmath189 of the negative gradient flow of @xmath50 , and its tangent space @xmath190 . in @xcite
, we observed that @xmath2 should act on the determinant bundle of `` unstable directions '' , @xmath191 of each critical point .
( we remark that such determinant line bundles also appear for morse complexes of morse - bott functions . )
orientation space @xmath192 at @xmath188 is defined as @xmath193 without a group action , such additional data are always trivial .
however when there is a group action , @xmath2 may act on orientation spaces in a non - trivial way .
this is a crucial difference , in the sense that the critical point @xmath188 is fixed by the @xmath194-action , but @xmath188 together with its orientation space may not be fixed by the @xmath194-action , as @xmath194 may reverse the orientation of unstable directions ( such a critical point was called non - orientable in @xcite ) .
hence , when we take the @xmath2-invariant part , non - orientable critical points disappear , and rightfully so since the topology of the sublevel sets of the orbifold does not change at such a critical point ( see @xcite ) .
we refer readers to @xcite for more details .
orientation spaces naturally arise also when we consider the assignment of signs for each isolated gradient flow in the definition of the boundary operator of morse complexes .
consider two critical points @xmath188 and @xmath195 with morse indices @xmath196 .
then , the space @xmath197 of gradient flows between @xmath188 and @xmath195 can be identified with @xmath198 where @xmath199 is the stable manifold of @xmath195 for @xmath200 . in terms of orientations , can be written as follows : for @xmath201 , @xmath202 where @xmath203 is the dual space with respect to the canonical pairing @xmath204 . in the isomorphism
, two copies of @xmath205 should appear but cancelled .
these copies come from the relation of @xmath199 and @xmath206 , and from taking the intersection of them in @xmath41 .
( see ( 12.2 ) of @xcite for the floer theoretic analogue ) .
therefore , this also explains why orientation spaces @xmath192 s are needed to define a @xmath2-action . here
, @xmath192 is a one dimensional real vector space , and we will be only concerned about the sign of the elements of these vector spaces , not the actual magnitudes when identified with @xmath207 .
hence , we use the normalization convention , following @xcite as explained in the introduction .
we set up the morse complex of @xmath50 as follows .
define @xmath208 we recall that the differential of the morse complex can be understood as follows .
the time translation of @xmath209 gives an element in @xmath210 , and from the equality , this determines an isomorphism @xmath211 .
we use @xmath212 to define a boundary map @xmath213 this is a coordinate - free way of writing the usual boundary operator of the morse complex .
namely , a trajectory has an induced orientation from the transversal intersection @xmath214 and we assign @xmath215 if it agrees with the direction of the flow itself ( the time translation ) and @xmath216 otherwise .
now , we can consider a natural @xmath2-action on the morse complex @xmath217 .
it is not hard to see that the @xmath2-action commutes with the morse differential in this new setting .
hence , we can take the @xmath2-invariant part @xmath218 , which is the morse complex for @xmath44 $ ] .
its homology is isomorphic to the singular homology of the quotient space of @xmath44 $ ] with @xmath79-coefficients for @xmath219 . here
@xmath80 is needed since over @xmath220 , local morse data are much more complicated .
one difficulty of building up @xmath2-morse theory is the equivariant transversality problem .
namely , it is hard to make a generic perturbation of either the morse function or the metric so that it becomes both morse - smale and @xmath2-invariant .
there are two ways to overcome this issue . in @xcite
, we introduced the notion of weak group actions on morse - complexes , for a generic morse - smale function @xmath50 , which is not @xmath2-invariant .
it can be also applied in this setting , and later we will use this idea to define the equivariant fukaya category for an exact symplectic manifold . another way is to work with a @xmath2-invariant morse function which is not necessarily morse - smale , and use the multi - valued perturbation scheme such as kuranishi structure(@xcite ) .
we will use this approach in the general case of floer theory to achieve equivariant transversality .
we also need @xmath221 for this method . from now on
, we assume that @xmath164 on @xmath116 is of morse - smale type , and leave more general case to one of the above two methods .
we spell out how to set up the @xmath2-equivariant novikov chain complex of @xmath51 under assumption [ assum : x_0fixed ] .
we denote by @xmath222 the set of critical points of @xmath164 .
we will use the normalization @xmath85 defined in the introduction .
[ def : chain ] we define the novikov chain complex of @xmath41 with respect to @xmath51 , @xmath223 as follows : the graded @xmath79-vector space @xmath224 is defined by @xmath225 where @xmath226 is the condition that for each @xmath227 , the following set is finite : @xmath228 the boundary operator @xmath229 on @xmath224 is defined by counting gradient flow lines of @xmath164 in @xmath116 as in @xmath2-morse theory .
note that the deck transformation group @xmath124 preserves the set @xmath230 , and this gives @xmath231 a @xmath127}$]-module structure , @xmath232 } \times cn_\ast ( m;\eta ) \to cn_\ast ( m;\eta).\ ] ] more precisely , consider a generator @xmath135 of @xmath127}$ ] .
i.e. an element @xmath135 of @xmath233 .
then the scalar multiplication of @xmath135 to @xmath234 is simply the action image @xmath235 .
then , @xmath127}$]-module structure on @xmath224 is given as follows .
as @xmath236 is defined by @xmath237 and @xmath135 sends @xmath238 to @xmath239 , we have an isomorphism ( denoted by the same letter ) @xmath240 we have an induced @xmath2-action on @xmath224 , which is @xmath127}$]-linear .
the differential @xmath229 is @xmath2-equivariant . as @xmath164 being @xmath2-invariant , @xmath2 acts on the set @xmath230 .
thus , each @xmath37 induces an isomorphism @xmath241 as before .
hence , we obtain a @xmath2-action on generators of @xmath224 over @xmath127}$ ] .
note that @xmath2 and @xmath124 actions commute ( since their actions on @xmath116 commute ) and the lemma follows .
the proof of @xmath2-equivariancy follows from and the definition of @xmath229 .
we call the complex @xmath224 equipped with the above @xmath2-action , the @xmath2-equivariant novikov complex of @xmath242 . its @xmath2-invariant part @xmath243 ( which is a @xmath127}$]-module ) is called the novikov complex of @xmath244,{\overline{\eta}})$ ] where @xmath245 is the induced 1-form on @xmath44 $ ] . here , the @xmath127}$]-module structure of @xmath243 is given by taking the @xmath2-invariant part of ( recall that the @xmath2-action is @xmath127}$]-linear ) : @xmath246 , and the boundary @xmath229 is @xmath127}$]-linear(@xcite ) . in conclusion , we get @xmath247 which is a module over @xmath127}$ ] .
what actually happens in taking the @xmath2-invariant part is as follows . for a critical point @xmath248 of @xmath164 ,
consider the isotropy group @xmath249 .
if the @xmath194-action on @xmath250 is orientation preserving , then this provides a non - trivial element in the @xmath2-invariant part .
if some element @xmath37 of @xmath194 reverses the orientation of @xmath250 , then exactly the half of @xmath194 reverses the orientations of @xmath238 .
thus the sum of @xmath2-action images of @xmath251 becomes zero by the cancellation , and hence do not contribute to the @xmath2-invariant part of the novikov complex .
we refer readers to @xcite for more details on related phenomenon .
in this section , we construct a @xmath2-novikov complex in the general case without assumption [ assum : x_0fixed ] .
the main difficulty is that the construction of covering spaces ( as spaces of homotopy classes of paths from a chosen base point ) is not compatible with the @xmath2-action .
indeed , the @xmath2-action may move the chosen base point to another point . for this , we will introduce the notion of _ energy zero subgroup _ @xmath53 , and define its action on novikov complexes . if @xmath252 , we can define equivariant and orbifolded theory from this @xmath53 action .
but in the general case that @xmath253 , we need to use a different novikov ring , what is called @xmath254 .
in fact , the most natural way to view it is to consider the orbifold fundamental group of the global quotient orbifold @xmath44 $ ] and the orbifold analogue of the novikov covering construction .
we will explain how @xmath53 , and @xmath254 arise in this orbifold setting .
hence , if @xmath253 , we can not define @xmath2-equivariant theory , but we can define @xmath2-orbifolded theory , by taking @xmath53-invariant parts of the novikov complex , with novikov ring given by @xmath254 .
we first take a generic point @xmath97 such that the isotropy group at @xmath100 of the @xmath2-action is trivial .
that is , @xmath255 for all @xmath256 .
as the @xmath2-action on @xmath41 is effective , such @xmath100 always exists .
let @xmath51 be a @xmath2-invariant morse 1-form @xmath51 on @xmath41 and consider the universal covering space @xmath134 and novikov covering @xmath257 obtained by @xmath258 where @xmath259 for @xmath260 .
now , we take the group action into account .
note that the @xmath2-action on @xmath41 does not induce a @xmath2-action on @xmath116 : consider @xmath261 with @xmath262 .
then , @xmath183 is an equivalent class of paths from @xmath100 to @xmath170 .
the naive @xmath2-action sends @xmath183 to @xmath263 , which is a homotopy class of paths from @xmath264 to @xmath265 .
this path is not an element of @xmath116 since it does not start from @xmath100 .
a subgroup @xmath53 of @xmath2 consists of an element @xmath35 such that there exist a path @xmath266 from @xmath100 to @xmath264 , with @xmath267 we call @xmath53 the _ energy zero _ subgroup of @xmath2 for @xmath51 and a path @xmath266 an energy zero path . to see that @xmath53 indeed forms a subgroup ,
consider @xmath268 with energy zero paths @xmath266 and @xmath269 respectively .
then , we can take an energy zero path for @xmath270 to be the concatenation @xmath271 a priori , such energy zero paths may not always exist ( i.e. @xmath272 in general ) .
for example , consider @xmath273-action defined by the rotation on @xmath274 and @xmath275 a @xmath2-invariant @xmath83-form on @xmath274 .
then , @xmath276 . by considering the energy zero subgroup @xmath53
, we can define a @xmath53-action on the novikov covering @xmath116 as below .
[ gactionnovcov ] given @xmath261 ( regarded as a class of paths in @xmath41 ) and @xmath277 with an energy zero path @xmath266 , we define the action of @xmath37 on @xmath183 as the concatenation of @xmath266 with the naive @xmath37-action image of @xmath183 : @xmath278 this gives a well - defined @xmath53-action on @xmath116 . for a different choice of an energy zero path @xmath279 for @xmath37 ,
the concatenation @xmath280 defines a loop in @xmath184 where @xmath281 for a path @xmath282 .
then , two elements @xmath283 , and @xmath284 differ by an action of @xmath184 , hence the action of @xmath53 on @xmath116 is well - defined .
also it is easy to check that @xmath285 for @xmath268 .
if one tries to define a @xmath37-action similarly for an element @xmath286 ( by taking a path @xmath266 from @xmath100 to @xmath287 ) , then it is straightforward to see that the relation @xmath288 can not hold for the action of @xmath37 . indeed , attaching @xmath266 keeps increasing ( or decreasing ) the energy .
now let us closely look into the @xmath53-action @xmath116 . from the construction , the projection @xmath111 is a @xmath53-equivariant map : @xmath289 the pull - back @xmath161 is exact by the definition of @xmath116 and hence we can choose @xmath162 so that @xmath290 as before .
we can proceed as in lemma [ invmorse ] to obtain @xmath291 is @xmath53-invariant . as we assumed that @xmath41 is connected
, we can find interesting properties of @xmath53 : @xmath53 is a normal subgroup of @xmath2 . for @xmath292 and @xmath35 ,
we want to show that @xmath293 lies in @xmath53 .
so , we need to find an energy zero path from a base point @xmath170 to @xmath294 .
set @xmath295 .
then , it suffices to find such a path from @xmath144 to @xmath296 .
choose any path @xmath229 from @xmath144 to @xmath170 and let @xmath269 be an energy zero path from @xmath170 to @xmath297 .
then , it is easy to check that @xmath298 from @xmath144 to @xmath296 has energy zero .
[ locgping ] for any @xmath299 , the isotropy group @xmath300 is a subgroup of @xmath53 .
we fix @xmath299 , and @xmath301 .
since @xmath41 is connected , we can choose a path @xmath302 from @xmath100 to @xmath144 .
then , @xmath303 defines a path from @xmath264 to @xmath304 .
then , we get a path from @xmath100 to @xmath264 by @xmath305 note that @xmath306 thus , @xmath282 has zero energy and realizes @xmath37 as an element of @xmath53 .
this implies that @xmath307 for all @xmath144 .
[ cor : getag ] if @xmath41 has a point , say @xmath308 , which is fixed by the whole @xmath2-action , then @xmath309 .
take @xmath144 to be @xmath308 , and apply lemma [ locgping ] .
[ lem : deccomm ] the @xmath53-action on @xmath116 commutes with the action of the deck transformation group @xmath310 .
the proof of lemma [ ggamma ] ( for the case with a @xmath2-fixed base point ) extends to this general case . but here , we give a geometric proof of it using the @xmath53-action .
let @xmath132 be represented by a loop @xmath136 at @xmath100 .
if @xmath266 realizes @xmath277 ( i.e. @xmath311 and @xmath312 ) , then @xmath313 realizes @xmath314 .
any point @xmath183 in @xmath116 can be thought of as a ( homotopy class of ) path from @xmath100 modulo @xmath184-action .
then , we have path representations of elements in @xmath116 as follows : ( b ) @xmath315 ( c ) @xmath316,height=192 ] 1 .
@xmath317 is represented by the concatenation @xmath318 .
@xmath319 is represented by the concatenation @xmath320 3 .
the difference between ( a ) and ( b ) is represented by @xmath321 .
+ since @xmath51 is @xmath37-invariant @xmath322 we conclude that ( c ) has energy zero , or equivalently @xmath323 since the @xmath2-action on @xmath41 is orientation preserving , it is easy to see that @xmath53 preserves the induced orientation on @xmath116 . here is a simple lemma on the relationship between isotropy groups in @xmath41 and @xmath116 analogous to lemma [ lem : isolocalgroup ] [ lem : isolocalgroup2 ] consider @xmath261 with @xmath324 .
then the isotropy group @xmath325 at @xmath183 in @xmath53 is given by @xmath326 the second equality in follows from the lemma [ locgping ] .
we only prove the first equality , here . for @xmath277 , we choose an energy zero path @xmath266 from @xmath100 to @xmath264 in @xmath41 .
then , if @xmath327 , then @xmath328 ( which is represented by @xmath329 ) has the same end point as @xmath183 .
this means that @xmath37 is an element of @xmath179 .
conversely , if @xmath37 fixes @xmath170 , and admits an energy zero path @xmath266 , then both @xmath329 and @xmath183 is a path from @xmath100 to @xmath170 in @xmath41 , and @xmath109 vanishes on @xmath330 .
thus @xmath182 and @xmath183 differ by an action of @xmath184 , which proves that @xmath331 . to define a @xmath2-equivariant novikov complex ,
let us first assume that @xmath332 , and postpone the study of general case to the next subsection .
we define the novikov chain complex as in definition [ def : chain ] , and novikov ring @xmath127}$ ] as in . by proceeding as before ,
we obtain [ thm : novchaincpx ] assume that the induced @xmath333-invariant function @xmath162 is of morse - smale type .
then , we have an induced @xmath2-action on the novikov complex @xmath334 which is @xmath127}$]-linear ( i.e. @xmath2-action commutes with @xmath127}$ ] ) .
this defines the @xmath2-equivariant novikov complex of @xmath335 . by taking the @xmath2-invariant part of @xmath336
, we get the chain complex @xmath337 which should be considered as novikov homology of the orbifold @xmath244,{\overline{\eta}})$ ] with respect to the induced 1-form @xmath245 .
this construction agrees with the previous construction of novikov complex with a @xmath2-fixed point by corollary [ cor : getag ] , and we leave the detailed check as an exercise . when @xmath338 , proper definition of orbifolded novikov complex is more involved as we need to introduce orbifold novikov ring @xmath339}$ ] .
first , we explain how the energy zero subgroup @xmath53 naturally arises from the orbifold setting .
we give a brief review on the orbifold fundamental group @xmath340\right)$ ] .
we call a continuous map @xmath341 \to m$ ] a generalized loop based at @xmath100 if @xmath342 and @xmath343 .
i.e. it is a loop up to @xmath2-action .
note that a genuine loop at @xmath100 is obviously a generalized loop .
since @xmath100 is generic , @xmath344 is uniquely determined from @xmath302 .
two generalized loops @xmath302 and @xmath345 ( based at @xmath100 ) are homotopic if * @xmath346 .
* there is a homotopy @xmath347 between @xmath302 and @xmath345 relative to end points .
we write @xmath348 for such homotopy relation .
two generalized loops can be multiplied as follows .
@xmath349 where @xmath350 denotes the concatenation of two paths as before .
the set of homotopy classes of generalized loops ( based at @xmath100 ) , denoted by @xmath340,{\overline{x}}_0\right)$ ] , has a group structure induced by .
one can also show that @xmath351 gives the inverse of @xmath352 $ ] , where @xmath131 .
we have a group homomorphism @xmath353\right ) \to g$ ] defined by @xmath354 ) = g_\alpha.\ ] ] it is easy to see that @xmath355 thus , we have the following short exact sequence of groups which is non - split in general . @xmath356\right ) \stackrel{e}{\longrightarrow } g \longrightarrow 1.\ ] ] the map @xmath357 is surjective since we assumed that @xmath41 is connected .
[ lem : orbfix ] if @xmath41 has a point , say @xmath308 , which is fixed by the whole @xmath2-action , then orbifold fundamental group can be written as a semi - direct product : @xmath358 ) \cong g \ltimes \pi_1 ( m).\ ] ] recall that the statement of the lemma is equivalent to the existence of a splitting of the extension .
thus , we need to find a homomorphism @xmath359)$ ] such that @xmath360 on @xmath2 . now ,
if @xmath2 fixes @xmath308 , we can construct @xmath361 as follows .
take a path @xmath362 \to m$ ] such that @xmath363 and @xmath364 .
for each @xmath35 , @xmath365 is a path from @xmath287 to @xmath308 .
thus , consider the concatenation of two paths @xmath366 which is a generalized loop starting from @xmath100 ending at @xmath287 .
it is easy to check that @xmath361 is a homomorphism , and @xmath367 .
it is well - known that the most of the covering theory extends to the case of orbifolds if one considers orbifold fundamental group , orbifold covering and orbifold universal covering .
( see for example takeuchi @xcite ) .
the orbifold universal covering for @xmath44 $ ] is given by the universal covering @xmath134 of @xmath41 , with the projection maps @xmath368 . and
@xmath369)$ ] becomes the deck transformation group of @xmath370 , whose action can be written as follows .
[ def : pi1act ] we define the ( left ) action of @xmath369)$ ] on @xmath134 as the concatenation of paths : for @xmath371)$ ] with @xmath372 , and for a path @xmath183 from @xmath100 to @xmath170 , as an element of @xmath134 , we define @xmath373 recall @xmath51 is a morse 1-form on @xmath41 , which is @xmath2-invariant .
we denote by @xmath374 the induced 1-form on the quotient @xmath375 .
we define @xmath376 ) \to { \mathbb{r}}$ ] as follows : @xmath377).\ ] ] this defines a group homomorphism since @xmath51 is @xmath2-invariant .
we consider the subgroup @xmath378 , the kernel of @xmath379 , which acts on the orbifold universal covering @xmath134 , and consider the quotient orbifold @xmath380 $ ] .
an analogue of novikov covering @xmath381 in section 2 would be the orbifold covering @xmath382 \to [ m / g].\ ] ] but it is cumbersome to work directly with @xmath380 $ ] .
for example , @xmath378 is discrete but may not be finite in general .
alternatively , we find another presentation of the orbifold @xmath380 $ ] , which is much easier to understand .
similarly to [ lem : orbfix ] , if there exists a @xmath2-fixed point , then we have @xmath383\right ) \cong g \ltimes { \rm{ker } } i_\eta.\ ] ] before we give the actual construction , it may be helpful to first explain the formalism .
consider the group @xmath71 acting effectively on a manifold @xmath72 , and let @xmath384 be a normal subgroup of @xmath71 .
then , @xmath384 , @xmath71 , and the quotient group @xmath385 form an exact sequence @xmath386 note that the @xmath71-action on @xmath72 induces the @xmath387-action on the quotient space @xmath388 because @xmath384 is a normal subgroup .
also the quotient satisfies @xmath389 thus instead of @xmath390 , we will consider @xmath388 and its quotient by the group @xmath385 , and the benefit is that @xmath388 , @xmath387 are simpler than @xmath72 and @xmath71 ( also in our case below @xmath388 will be still a manifold ) .
now , we put @xmath391 , and @xmath392 and consider the following diagram of subgroups of fundamental groups .
@xmath393 & { \rm{ker } } i_{\eta } \ar[d]_{\leq } \ar[r ] & { \rm{ker } } i_{\eta}^{orb } \ar[d]_{\leq } \ar[r]^{\qquad e_{\eta } } & g_{\eta } \ar[d]_{\leq } \ar[r ] & 1 \\ 1 \ar[r ] & \pi_1 ( m , x_0 ) \ar[r ] \ar[d ] & \pi_1^{orb } ( [ m / g ] ) \ar[r]^{\qquad e } \ar[d ] & g \ar[r ] \ar[d ] & 1 \\ 1 \ar[r ] & \gamma \ar[r ] & \gamma^{orb } \ar[r ] & g /g_\eta \ar[r ] & 1 } \ ] ] we define @xmath53 to be the quotient group @xmath394 .
@xmath53 can be identified with a subgroup of @xmath2 via the map @xmath395 ) / \pi_1 ( m ) = g.\ ] ] hence , we have @xmath396 in the formalism above . from the diagram
, we can correspondingly construct a diagram of covering spaces as follows : @xmath397^{id } \ar[d ] & { \widetilde{m}}_{univ } \ar[d ] \\ { \widetilde{m}}={\widetilde{m}}_{univ } / { \rm{ker } } i_{\eta } \ar[r]^{/g_\eta } \ar[d ] & [ { \widetilde{m}}_{univ } /{\rm{ker } } i_{\eta } ^{orb } \ar[d ] ] \\ m \ar[r]_{/g } & [ m / g ] } \ ] ] here @xmath398 , which is the novikov covering space considered in the previous section . from the discussion above , the identity becomes @xmath399 \cong [ { \widetilde{m}}_{univ}/{{\rm{ker } } i_{\eta}^{orb}}].\ ] ] note that @xmath116 is a manifold ( it is a covering of @xmath41 ) , and we have a presentation of the orbifold which involves simpler group @xmath53 than @xmath378 .
this provides an explanation of @xmath53 , and should extend to the general orbifold case . in the case
when @xmath253 , there is a subtle problem with novikov rings as examined from now on .
the deck transformation group for the novikov cover @xmath380 \to [ m / g]$ ] is given by the quotient group @xmath400 ) / { \rm{ker } } i_{\eta}^{orb},\ ] ] while the deck transformation group for @xmath125 is given by @xmath401 ) / { \rm{ker } } i_{\eta}.\ ] ] from the diagram , we see that the difference of two groups is exactly given by @xmath402 .
we write the associated novikov rings as @xmath127}^{orb}$ ] and @xmath127}$ ] respectively as in .
the inclusion @xmath403 gives rise to a map @xmath127 } \to \lambda_{[\eta]}^{orb}$ ] between the ( compeletions of ) group rings .
the novikov complex is defined in the following way . from the previous construction
, we have the novikov complex @xmath334 for @xmath116 with novikov coefficients @xmath127}$ ] and the @xmath53-action on it .
now , we can replace the novikov coefficients by @xmath404 } } \lambda_{[\eta]}^{orb},\ ] ] which still has a @xmath53-action .
[ def : obnoveta ] we define the orbifolded novikov complex of @xmath244 , { \overline{\eta}})$ ] to be @xmath405 } } \lambda_{[\eta]}^{orb } \big)^{g_\eta}\ ] ] for example , let us consider the case that @xmath2-action is free .
the following proposition shows that the complex in definition [ def : obnoveta ] is the right novikov complex for the quotient space of the free action .
if @xmath2 acts on @xmath41 freely , then we have @xmath339 } \cong \lambda_{[{\overline{\eta}}]}$ ] and the following isomorphism of @xmath406}$]-modules @xmath407 } } \lambda_{[\eta]}^{orb } \big)^{g_\eta } \cong cn_\ast ( m / g ; \bar{\eta})\ ] ] since @xmath2 acts freely on @xmath41 @xmath408 ) \cong \pi_1 ( m / g ) , \quad { \rm ker } i_{\eta}^{orb } \cong { \rm ker } i_{\bar{\eta}}.\ ] ] this gives the desired identification of @xmath339}$ ] and @xmath406}$ ] .
note that the @xmath53-action on @xmath116 is also free .
one can see from the diagram that the novikov covering space of @xmath375 with respect to the morse @xmath83-form @xmath245 is exactly the quotient of @xmath116 by the free @xmath53-action .
therefore , @xmath409 is naturally isomorphic to @xmath410 which is essentially the morse chain complex of @xmath411 with the novikov coefficients .
@xmath412 + * ( 1 ) * let us first drop the assumption that @xmath41 is connected from theorem [ thm : novchaincpx ] .
denote connected components of @xmath41 by @xmath413 @xmath414 .
i.e. @xmath415 . let us call @xmath413 and @xmath416 , _
@xmath2-related _ if there exist @xmath417 and @xmath35 such that @xmath418 .
this defines an equivalence relation .
we divide the index set @xmath419 into @xmath420 using this equivalent relation so that connected components @xmath421 are @xmath2-related for any @xmath422 .
to define a group action , we need to make the following assumption .
if we denote by @xmath423 the subgroup of @xmath2 whose elements preserve the connected component @xmath424 , then we require that @xmath425 hence , for any @xmath426 and @xmath427 , there exists an energy zero path @xmath282 connecting @xmath170 and @xmath428 . as an initial step ,
consider the construction of the morse - novikov complex for the subcollection @xmath421 for a fixed @xmath429 .
now , we choose base points @xmath430 so that they lie in a single @xmath2-orbit .
we can consider novikov - morse theory on each connected component , with base points given by @xmath431 s .
then , from the assumption , any @xmath37-image of @xmath431 for @xmath89 can be connected to one of @xmath432 by energy zero path , say @xmath266 .
then , we can define the group action as before .
namely , we concatenate @xmath266 to the front of the image under the naive group action . also , one can see that the deck transformation group of one component is isomorphic ( by conjugation ) to that of the other component , hence the resulting novikov rings can be identified , which we denote as @xmath433 .
one can take the direct sum of novikov complexes @xmath434 so that it admits a natural @xmath2-action .
now , let us illustrate how to add various complexes for different @xmath429 s .
for this , we need to introduce the universal novikov ring to properly compare various novikov rings of different subcollections .
then , we will take completed direct sum over all different subcollections to define the novikov complex ( this is similar to what has been done for novikov floer theory in @xcite , where the case of several connected components of a path space are considered ) .
the universal novikov ring is defined by @xmath435 and the ring homomorphism from @xmath127 } \to \lambda$ ] by @xmath436 for each novikov complex @xmath437 with @xmath2-action , we can introduce the new novikov complex of @xmath41 as @xmath438 } } \lambda\ ] ] which admits a @xmath2-action also .
+ * ( 2 ) * we have considered a global quotient orbifold @xmath44 $ ] , and have seen how to set up its novikov complex by working on @xmath116 observing @xmath53-actions .
we can generalize the construction to any effective orbifold if we just use orbifold terms , but it will not be explicit as in the global quotient case .
as we will not use it in this paper , we briefly outline the construction . consider a compact connected oriented orbifold @xmath439 with a closed morse 1-form @xmath51 on @xmath439 .
it is straight forward that @xmath51 defines a homomorphism from @xmath440 .
then , as before its kernel @xmath441 corresponds to the orbifold covering @xmath442 on which @xmath443 is an exact @xmath83-form . note that @xmath444 is no longer a smooth manifold , but rather an orbifold .
consider the deck transformation group @xmath445 , and take the ( one - sided ) completion of the group ring of @xmath446 to define @xmath339}$ ] .
let @xmath447 be a morse function on the ( orbifold ) novikov cover which integrates @xmath161 .
if @xmath50 is of morse - smale type , we apply the construction in @xcite ( for general orbifolds ) to obtain the orbifold morse complex of @xmath447 with @xmath339}$]-coefficients .
@xmath448 has a natural @xmath339}$]-module structure and the boundary operator @xmath229 which counts gradient trajectories is @xmath339}$]-linear .
we may call the resulting complex the novikov complex of @xmath449 .
we recall the setup of floer - novikov theory briefly ( following @xcite ) , and also the notion of orientation spaces as well as some gluing formulas for them , which are well - known constructions . in the last subsection
, we will introduce the spin @xmath124-equivalence ( refining @xmath124-equivalence ) to associate the canonical spin structure to each generator of floer - novikov complexes .
let @xmath28 and @xmath54 be lagrangian submanifolds of a symplectic manifold @xmath4 , transversally intersecting with each other .
let us assume that both @xmath28 and @xmath54 are connected , compact , oriented and also spin for simplicity . consider the space of paths @xmath450 \to m \,\
, | \,\ , l(0 ) \in l_0 , l(1 ) \in l_1\}.\ ] ] we denote by @xmath451 its connected component containing @xmath452 .
we define an @xmath83-form @xmath302 on @xmath453 as follows : for @xmath454 , @xmath455 } \omega(l'(t ) , \xi(t ) ) dt.\ ] ] the @xmath83-form @xmath302 is closed but not necessarily exact and hence , one needs to find a suitable novikov covering for @xmath302 .
first let us consider the universal covering @xmath456 of @xmath457 .
a path in @xmath458 is a continuous map @xmath459 \times [ 0,1 ] \to m$ ] satisfying @xmath460 and @xmath461 where @xmath462 . if @xmath463 for @xmath464 , the homotopy class of such @xmath465 gives a point in the universal covering space @xmath456 lying over @xmath429 .
denote by @xmath466 .
given two elements @xmath467 and @xmath468 of @xmath457 , we can glue these to get a loop @xmath469 in @xmath451 .
so we have a loop @xmath470 \to m\ ] ] which satisfies the lagrangian boundary conditions @xmath471 we can define its symplectic area @xmath472 and maslov index @xmath473 . the maslov index @xmath474 is defined as usual by the difference of the maslov indices of the loops in the lagrangian grassmannian , where the loops are obtained from two boundary components of @xmath45 after a symplectic trivialization of the pull - back bundle over @xmath45 .
this gives a homomorphism @xmath475 the following definition of the @xmath124-equivalence between two paths in @xmath451 is due to @xcite .
[ def : gammaeq ] @xmath467 and @xmath476 are said to be @xmath124-equivalent if @xmath477 the set of @xmath124-equivalence classes forms a covering @xmath478 associated to the the following subgroup of @xmath479 : @xmath480 hence the deck transformation group @xmath481 of the covering @xmath482 is given by the quotient @xmath483 . by the definition of @xmath124-equivalences ,
the homomorphisms @xmath484 and @xmath474 give rise to @xmath485 defined by @xmath486 and @xmath487 . here , the element @xmath488 is represented by a map @xmath489 \to m$ ] satisfying above .
note that @xmath490 is an abelian group .
indeed gives an injective homomorphism onto its image by the definition of @xmath124-equivalence .
define @xmath491 whose element is a formal sum @xmath492\ ] ] with @xmath493 .
it should satisfy the condition that the set @xmath494 is finite for any @xmath45 . from the construction , @xmath495 is exact ( where @xmath47 is the covering map in ) .
the floer action functional @xmath496 on @xmath478 is defined by @xmath497 ) = \int w^{\ast } \omega\ ] ] for an element @xmath498 $ ] in @xmath478 .
then , the direct computation shows that @xmath499 i.e. the floer action functional is a morse function on the novikov cover which integrates @xmath500 . in short ,
floer - novikov theory is morse theory of @xmath496 on @xmath478 with novikov coefficients @xmath501 .
critical points of @xmath496 are given by @xmath502 \in { \widetilde{\omega } } ( l_0 , l_1;l_0)$ ] where @xmath503 is a constant path at @xmath504 .
like in @xcite , we work with the universal novikov ring @xmath505 and its subring @xmath506 as coefficient rings for floer cohomology groups : @xmath507 @xmath508 recall for a single lagrangian @xmath5 , the associated novikov ring is defined by @xmath509 \ , \right| \ , \ # \ { \beta | a_\beta \neq 0 , e(\beta ) < c \ } \,\,\mbox{is finite for each } \,\,c \right\ } .\ ] ] the homomorphism @xmath510 given by @xmath511 \right ) = \sum_{\beta } a_\beta t^{\int_\beta \omega}\ ] ] identifies @xmath512 with a subring of the universal novikov ring .
we use the same identification as to transfer @xmath513 into the universal novikov coefficients .
we recall orientation spaces for lagrangian floer theory .
the cauchy - riemann ( cr for short ) operators associated to generators of the floer - novikov complex were introduced in @xcite for bott - morse lagrangian floer theory .
the index spaces ( definition [ def : orspace ] ) of these operators will be called orientation spaces .
the orientation spaces are keys to analyze fredholm indices , and orientations in the lagrangian floer theory .
as so , it is natural to expect that they play a important role in understanding group actions on lagrangian floer theory .
we fix a reference path @xmath24 , and specify pre - fixed choices of data in what follows . for the chosen reference path @xmath20 \to m$ ]
, we write @xmath514 , @xmath515 . along the chosen reference path
, we choose a fixed path @xmath516(@xmath517 ) , where @xmath518 is an element of oriented lagrangian subspace in @xmath519 whose end points are given by @xmath520 and @xmath521 .
we denote @xmath522 } \{t\ } \times \lambda_{l_0}(t ) \to [ 0,1].\ ] ] before we explain additional choices needed to specify spin structure of the lagrangian bundle , we recall the definition of the associated cr operator . consider a critical point @xmath502
\in cr(l_0 , l_1 ; l_0)$ ] for @xmath188 a point in the lagrangian intersection @xmath523 .
we choose a representative @xmath524 ( instead of its @xmath124-equivalence class ) .
we want to define the orientation space @xmath525 at the end .
the map @xmath465 produces a path @xmath526 of lagrangian subspaces in @xmath527 from @xmath528 to @xmath529 by concatenating @xmath516 and lagrangian paths in top and bottom edge of @xmath530 .
more precisely , for @xmath531 ) \cup ( [ 0,1 ] \times \{0,1\ } ) \stackrel{homeo}{\cong } [ 0,1],\ ] ] we get a path @xmath532 such that , @xmath533 as drawn in figure [ [ 0,1 ] ] . ,
height=153 ] since @xmath534 ^ 2 $ ] is contractible , we can trivialize @xmath535 @xmath536 ^ 2 \times t_p m\ ] ] so that @xmath537 is a path in the oriented lagrangian grassmannian of @xmath538 .
put @xmath539 for @xmath532 , one defines a cr operator by @xmath540 whose lagrangian boundary condition comes from @xmath537 . here , @xmath541 is the banach space of @xmath542 maps @xmath543 with weighted norms : 1 .
@xmath544 , where we use @xmath545 2 .
@xmath546 , where @xmath547 3 .
@xmath548 .
it is well known that @xmath549 is a fredholm operator .
see @xcite for more details .
finally , one assigns an orientation space to each generator @xmath550 $ ] as follows .
[ def : orspace ] the orientation space @xmath551 for a pair @xmath552 is defined as @xmath553 the canonical orientation on the moduli space associated to the cr operators with lagrangian boundary conditions is determined from the spin structures of the lagrangian sub - bundles .
later , we will deal with isomorphisms between various orientation spaces , and to find a canonical isomorphism , we need to also specify spin structures of the lagrangian bundles on the boundary components . without these
, there would be an ambiguity of signs whenever we try to compare , identify and glue orientation spaces .
signs are indeed very crucial information ( for example different signs for the group action change the invariant sets ) . as in chapter 8 of @xcite
, we fix the following choices of spin data . also we choose and fix a trivialization @xmath554 \times \mathbb{r}^n \to \widetilde{\lambda}_{l_0 } $ ] , which also provides a trivialization of @xmath555 and @xmath556 .
the latter trivialization @xmath557 gives an embedding @xmath558 into the fiber of @xmath559 at @xmath560 .
we choose one of its lifts @xmath561 between two possible choices .
the trivialization @xmath562 of @xmath563 also can be lifted , and we make a choice @xmath564 \times spin(n ) \cong p_{spin}(\lambda_{l_0}).\ ] ] now , we can use @xmath565 to glue @xmath566 at @xmath567 and @xmath568 .
namely , we glue @xmath569 and @xmath570 using @xmath571 at @xmath560 for @xmath18 . we emphasize that with these pre - fixed data , we have the canonical spin structure of @xmath537 for the generator @xmath552 : for @xmath572 with @xmath573 , we have @xmath574 and for @xmath575 with @xmath576 , we have @xmath577 , and for @xmath518 for @xmath578 , we have @xmath579 . at @xmath580 ( resp .
@xmath581 ) , we glue @xmath579 and @xmath574 ( resp .
@xmath577 ) by @xmath582 ( resp .
@xmath583 ) .
we need to recall some elementary facts on determinant ( index ) spaces of cr operators .
see ( * ? ? ?
* section ( 11c ) ) for more explanations .
first , we consider the gluing of cr operators on @xmath584 and the one on holomorphic strips , and investigate the relation between orientation spaces before and after the gluing . let @xmath585 be given by @xmath584 , and let @xmath586 be a strip @xmath587 $ ] .
consider a @xmath32-holomorphic strip connecting two intersection points @xmath188 and @xmath195 and the induced lagrangian sub - bundle data @xmath588 along @xmath589 , which give rise to a weighted cr operator @xmath590 .
the standard gluing theorem for the weighted cr operators @xmath591 for @xmath585 , and @xmath590 for @xmath586 says that @xmath592 where @xmath593 and by @xmath594 , we mean the lagrangian sub - bundle data along @xmath595 obtained by concatenating that of @xmath596 backwards , that of @xmath585 , and that of @xmath597 .
( see ( a ) of figure [ fig : sss ] . )
hence , the determinant line bundle @xmath598 for the holomorphic strip is obtained as a tensor product ( compare with ) @xmath599 and @xmath586 ( b ) gluing of @xmath600 and @xmath601,height=163 ] in the discussion below we need another type of gluing result .
let @xmath600 be a disc with a trivial symplectic bundle @xmath602 and a lagrangian sub - bundle @xmath603 ( which is not necessarily trivial ) along @xmath604 and consider @xmath605 and a trivial symplectic bundle @xmath606 and a lagrangian sub - bundle @xmath607 along @xmath608 which is fixed sub - bundle for @xmath609 as in . consider boundary marked points @xmath610 , and @xmath611 and an identification @xmath612 which identifies @xmath613 with @xmath614 .
then determinant spaces of the associated cr operators satisfy @xmath615 where @xmath616 with a lagrangian sub - bundle data along @xmath617 obtained by concatenating that of @xmath618 backwards , that of @xmath600 counter - clockwise , and that of @xmath619 .
( see ( b ) of figure [ fig : sss ] . ) here , the new term @xmath620 appears because the elements of the kernel of cauchy - riemann problem on @xmath600 and @xmath601 should have the same boundary value in @xmath621 at @xmath622 .
later , we will need the following lemma of seidel .
* lemma 11.17)[lem : se11 ] let @xmath623 be a loop in the lagrangian grassmannian . then
a choice of pin structure determines an isomorphism @xmath624 one way to interpret this lemma is to say that the choice of pin structure gives rise to a canonical orientation of the determinant space ( relative to @xmath625 ) .
this lemma will be used in for a lagrangian loop @xmath623 of @xmath604 so that we have a canonical isomorphism between @xmath626 and @xmath627 .
let us briefly review the argument in @xcite for lemma [ lem : se11 ] in the case of a maslov index zero loop .
first consider @xmath623 to be a constant loop with a trivial spin structure . then
, the kernel is identified with @xmath625 whereas the cokernel is trivial . for non - constant @xmath623 of index zero with a trivial spin structure ,
we consider a homotopy from @xmath623 to a trivial one to obtain the identification @xmath628 if we equip @xmath623 with a non - trivial spin structure , the identification is changed by reversing the sign . we have defined the closed @xmath83-form @xmath302 on @xmath453 , and considered the novikov covering @xmath629 by using @xmath124-equivalences on bounding surfaces of paths to define action functional @xmath630 . now
, we introduce a new notion of spin-@xmath124-equivalences
. this will be a finer equivalence relation than the @xmath124-equivalence , and hence it will have a possibly bigger deck transformation group , which results in the bigger novikov ring @xmath631 .
but later , we will reduce the coefficient ring to the usual universal novikov ring with help of a homomorphism @xmath631 to @xmath632 . in this process ,
@xmath124-equivalent , but not spin-@xmath124 equivalent elements will be identified with opposite signs . hence , the floer complex is essentially the same complex as that of @xcite .
spin structures are essentially taken into account also in the definition of floer complexes in @xcite , but the usage of spin @xmath124-equivalence here gives more canonical and explicit way of merging the spin structures of lagrangian bundle data at critical points into the floer complexes .
namely , @xmath124-equivalences are not very natural to assign orientations in the sense that @xmath124-equivalent generators with different spin structures on the associated lagrangian bundle should rather be thought to give the same generator _ with opposite sign_. in @xcite , the choice of spin structures is carried as extra data , but we would like to combine it with the @xmath124-equivalence by defining a finer equivalence relation .
we emphasize that what to be done is a reformulation of orientation analysis of @xcite which is written very carefully in section 8.1.3 .
introduction of spin-@xmath124-equivalences provides a way to understand the novikov complex and its orientation scheme at once , and we explain how to adapt orientation scheme of @xcite into this setting .
detailed constructions are to be given from now on .
we first define the spin-@xmath124-equivalence . if two pairs @xmath467 and @xmath476 have the same end points @xmath429 , then induced lagrangian paths @xmath633 and @xmath634 have the same end values .
they can be glued to each other by first inverting @xmath634 and identifying @xmath635 with @xmath636 for @xmath18 . as a result
, we obtain a lagrangian bundle over a circle @xmath274 , which we denote by @xmath637 .
their spin structures also can be glued too .
note that @xmath637 is a lagrangian sub - bundle over a rectangle . on the horizontal edges
, it restricts to the tangent bundle of lagrangian submanifolds with given spin structures , @xmath638 and @xmath639 . on vertical edges ,
it becomes @xmath516 over the base path @xmath24 with the spin structure @xmath579 .
how to glue these spin structures at vertices by @xmath571 were explained in the paragraph after .
this provides the desired canonical spin structure on @xmath637 . if @xmath467 and @xmath476 are @xmath124-equivalent , the maslov index of @xmath637 vanishes , and hence the trivial spin structure of @xmath637 makes sense ( see the explanation before assumption [ as : spin ] for the definition of trivial spin structure ) .
[ def : spingamma ] @xmath467 and @xmath476 are said to be _
spin-@xmath124-equivalent _ if they are @xmath124-equivalent , and the induced spin structure on @xmath637 over @xmath274 is a _
trivial spin structure_. the set of spin-@xmath124-equivalence classes is denoted by @xmath640 .
the spin-@xmath124 equivalence is linked to the covering @xmath640 in the same way that we had the @xmath124-equivalence from a covering @xmath478 .
in addition to @xmath641 , we have a homomorphism @xmath642 , which is defined by 0 for the trivial spin structure on @xmath643 with @xmath644 , and @xmath83 otherwise .
in fact , by using lemma [ lem : se11 ] , we can extend @xmath645 to the following map @xmath646 let @xmath647 \to m$ ] be a cylinder representing an element of @xmath648 i.e. the boundary @xmath649 maps to @xmath28 and the boundary @xmath650 to @xmath54 by @xmath45 .
we cut the cylinder @xmath651 $ ] along @xmath652 $ ] to obtain a disc @xmath653 , and choose a path of lagrangian subspaces from @xmath654 to @xmath655 which connects corresponding spin structures also .
in this way , we get a lagrangian loop along @xmath656 with a spin structure from @xmath45 . with help of lemma [ lem : se11 ] , we can determine whether the canonical orientation of the associated determinant space equals the given orientation of @xmath657 or not .
if it is equal , then we define @xmath645 of @xmath45 to be zero , and one otherwise .
note that @xmath658 does not depend on the choice of paths connecting @xmath654 and @xmath655 since this path contributes to @xmath645 twice after cutting .
the covering @xmath640 corresponds to the following subgroup of @xmath659 : @xmath660 we denote by @xmath661 the group of deck transformations of the covering @xmath662 it is not hard to check that @xmath663 is a normal subgroup of @xmath664 and hence , the deck transformation group is given by the quotient group @xmath665 . note that @xmath666 is also an abelian group since it is a quotient group of the abelian group @xmath667 .
the novikov ring for the coverings @xmath668 is given as below .
here , @xmath79 is a commutative ring with unity which includes @xmath80 ( we need rational coefficients for multisections ) .
[ def : novl01 ] @xmath669 is the set of all formal sums @xmath670,\ ] ] such that @xmath671 and @xmath672 we write @xmath673 compare it with the novikov ring @xmath501 defined in subsection [ subsec : fntheory ] . the homomorphism @xmath674 @xmath675 is defined as the map induced from the natural projection @xmath676 with an additional multiplication of @xmath645 defined in .
later , we will use the coefficient ring @xmath501 ( or the universal novikov ring ) using this homomorphism @xmath674 .
now , new ( naive ) generators of the floer complex are given by @xmath677 \in \left .
{ \widetilde{\omega } } ( l_0 , l_1;l_0)^{sp } \ , \right| \ , p \in l_0 \cap l_1 \right\},\ ] ] where @xmath503 is the constant path at @xmath188
. if an element @xmath678 is represented by the loop @xmath45 of @xmath679 , then the action of @xmath680 on @xmath681 is defined by @xmath682 = [ c \star w , l_p].\ ] ] from this action , one can naturally define a @xmath683$]-module which is generated by elements in @xmath684 .
the actual definition of the cochain complex should also involve the orientation spaces associated to the generators in .
more precisely , consider the formal sum @xmath685)=k } a_{[w , l_p ] } x_{[w , l_p]}\ ] ] where @xmath686 } \in r$ ] and @xmath687 } \in \left|\theta_{[w , l_p]}^- \right|_r $ ] .
for the precise definition of @xmath688}^-$ ] , see subsection [ oriline ] .
spin-@xmath124-action on orientation spaces is defined using the isomorphism of orientation spaces in .
we omit the details , and refer readers to section [ gactfloernov ] where similar operation on orientation spaces are explained carefully .
let @xmath689 \in \left .
cr(l_0 , l_1;l_0)^{sp } \ , \right| \ , a_{[w , l_p ] } \right\ } \neq 0.\ ] ] we call the sum @xmath690 a floer cochain of degree @xmath691 if @xmath692 \,|\ , \mathcal{a } ( [ w , l_p ] ) \leq \lambda \ } ) ) < \infty\ ] ] for any @xmath693 .
we denote the set of all floer cochains of degree @xmath691 by @xmath694 and we define @xmath695 from the @xmath680-action @xmath684 , we get a natural @xmath696-module structure of @xmath697 .
the floer complex of the pair @xmath27 is defined as @xmath698 using the homomorphism @xmath674 of .
we will use the universal novikov coefficients by considering ring homomorphism @xmath699 in many applications such as morphism spaces of fukaya categories .
this will enable us to compare various of modules arising from lagrangian floer theory over the single coefficient field .
we will revisit the @xmath696-module structure on @xmath697 with more details in subsection [ galpha ] and discuss the new novikov ring which fits into orbifold floer cohomology in subsection [ subsec : newnov ] .
consider @xmath2-invariant lagrangian submanifolds and suppose further that they admit @xmath2-invariant spin structures .
although spin structures are @xmath2-invariant , group actions on the frame bundles of lagrangians do not lift in general to the one on spin bundles .
for example , consider a reflection @xmath700 , and its lift @xmath701 .
then , @xmath702 , but we have @xmath703 where @xmath704 is the rank of @xmath705 ( * ? ? ?
* lemma 11.4 ) .
hence , the @xmath3-action given by @xmath384 on @xmath706 induces the action on @xmath707 only up to signs , and the sign error indeed has an important geometric origin . for the general group actions
, we will regard these sign differences as a group cohomology class , and call them spin profiles . before presenting the definition of spin profiles
, we briefly recall basic facts on spin structures for reader s convenience .
let @xmath708 be the clifford algebra .
there is a natural embedding @xmath709 and vectors satisfy the relation @xmath710 for the unit @xmath711 .
then , @xmath712 is defined as the multiplicative subgroup generated by @xmath713 , and @xmath714 the subgroup of @xmath707 whose elements are given by products of even number of generators in @xmath715 . both of homomorphisms @xmath716 and
@xmath717 are induced by the twisted adjoint action on @xmath706 , sending @xmath718 for each generator @xmath719 .
a spin structure on @xmath5 is a principal @xmath720 bundle @xmath721 which is a fiberwise double cover of the principal orthonormal frame bundle @xmath722 .
let @xmath723 be the covering map such that @xmath724 .
@xmath725 acts on the set of isomorphism classes of spin structures on @xmath5 and the action is free and transitive .
thus , it is an affine space modeled on @xmath725 .
in fact , milnor @xcite has shown that a spin structure of a vector bundle on a simplicial complex can be understood as a trivialization of the vector bundle on the 1-skeleton , which can be extended to the 2-skeleton .
then , the action of @xmath726 can be interpreted as changing the trivialization of the vector bundle along the non - trivial loops in the 1-skeleton .
for example , homotopy classes of trivializations of a rank @xmath727 vector bundle on @xmath534 $ ] ( with fixed trivialization at the end points ) could differ by a loop of @xmath728 , but trivial and non - trivial loops have different lifts into @xmath720 .
we are particularly interested in the vector bundle on a circle .
if the vector bundle @xmath729 is spin , then there are two choices of spin structures .
however , there is no preferred choice unless we know that the vector bundle @xmath730 can be extended to @xmath731 with @xmath732 .
if @xmath730 has extension , there is a unique spin structure on @xmath733 which extends to @xmath731 .
we call this preferred choice the _ trivial spin structure_. in our setting , any loop in the lagrangian grassmannian of maslov index zero is indeed contractible .
for such a loop , we have a choice of the trivial spin structure .
let us first make the following necessary assumption .
[ as : spin ] let @xmath5 be a connected spin ( orientable ) lagrangian submanifold with a choice of spin structure and suppose the @xmath2-action on @xmath5 is orientation preserving .
we assume @xmath2 preserves the isomorphism class of the given spin structure .
i.e. a pulled - back spin structure by @xmath37 is isomorphic to the original one for each @xmath35 .
such spin structures are said to be @xmath2-invariant .
we do not consider the case that spin structure is not @xmath2-invariant . recall that @xmath5 is said to be relatively spin , if there exist a class @xmath734 such that the restriction of @xmath735 to @xmath5 is zero , and hence there exist a bundle @xmath736 on 3-skeleton of @xmath41 , such that @xmath737 becomes spin .
floer homology between two relatively spin lagrangian submanifolds can be defined only if they become relatively spin with the same @xmath736 . with @xmath2-actions
, we can work with @xmath738 so that there exist a @xmath2-bundle @xmath736 on 3 skeleton of @xmath41 with @xmath737 being spin and @xmath2-invariant .
we leave further details to the readers , and consider the case that @xmath5 is spin from now on . given a @xmath2-action on @xmath5 ,
let us denote by @xmath739 its action @xmath740 for @xmath89 and @xmath741 , and let a @xmath2-invariant metric be given on @xmath41 .
we denote the orthonormal frame bundle of @xmath5 by @xmath742 , which has an induced @xmath2-action denoted by @xmath743 .
in fact , @xmath744 provides an isomorphism @xmath745 let us fix a spin structure of @xmath5 satisfying the above assumption [ as : spin ] .
we would like to lift @xmath743 to a map @xmath746 between spin bundles @xmath747 such a lift exists by the assumption [ as : spin ] and the standard covering theory argument .
indeed , an explicit isomorphism of bundles over @xmath748 @xmath749 from can be viewed as a lift of @xmath750 in .
note that the lift is not unique , as it can be multiplied by @xmath751 to produce another one .
we always choose the lift @xmath752 for the identity element @xmath753 to be the identity map on @xmath754 . for @xmath755
, we have a natural map given by the composition @xmath756 but it does not necessarily equal @xmath757 , but possibly differs by multplication of @xmath758 if the group action on the frame bundle does not lift to the one on the spin bundle .
note that this difference @xmath759 is locally constant function on @xmath741 .
[ def : spf ] the function @xmath760 is defined implicitly by the relation @xmath761 where @xmath762 means the multiplication by @xmath758 .
if @xmath5 has several connected components , we make the following assumption which enables us to get a constant @xmath763 , not depending on connected components
. see ( ii ) of subsection [ subsec : genspfs ] below for more discussions on this issue .
[ ass : ldisconnspf ] if @xmath5 is not connected ( but @xmath764 is connected ) , then we assume that component - preserving subgroups of @xmath2 are the same for every connected components of @xmath5 and normal in @xmath2 .
we assume further that each connected component has the same spin profile with respect to the action of the component - preserving subgroup .
we will show that @xmath760 defines an second group cohomology class which is independent of choices of @xmath765 s .
thus , for example , if we have an induced group action on spin bundles , @xmath763 is cohomologous to @xmath766 .
@xmath763 defines group cohomology class in @xmath0 , which we call a _
spin profile _ of @xmath5 .
the function @xmath763 depends on the choices of the lifts @xmath767 , but the resulting group cohomology class @xmath768 $ ] is independent of these choices . we first show that @xmath763 is indeed a cocycle for the group cohomology @xmath769 .
let @xmath770 be the set of all liftings of @xmath750 for every @xmath35 .
clearly , @xmath770 forms a group and the map @xmath771 sending @xmath772 is @xmath773 to @xmath83 .
the kernel of @xmath771 consists of the liftings of the identity , which are @xmath774 .
therefore we get the following exact sequence of groups @xmath775 the choice of lifts @xmath776 for each @xmath89 defines a section @xmath777 by setting @xmath778 ) , which may _ not _ be a group homomorphism .
recall that @xmath779 is assumed to be the identity .
as @xmath780 is abelian , it naturally admits a @xmath2-module structure by conjugation : for @xmath781 we write this conjugation action as @xmath782 , which is independent of the choice of @xmath562 . from general theory of group extension , extensions @xmath770 as in
are classified by a cohomology class in @xmath783 given by so called the factor set @xmath784 = \sigma(g ) \sigma(h ) \sigma(gh)^{-1}.\ ] ] since @xmath785 : p_{spin } ( l ) \to p_{spin } ( l)$ ] covers the identity from @xmath786 to itself , it should be either @xmath774 .
thus , the factor set is indeed a map @xmath787 : g \times g \to { \mathbb{z}}/ 2{\mathbb{z}}\ ] ] which exactly agrees with @xmath763 in the definition [ def : spf ] .
the following lemma is standard , whose proof is omitted .
( see @xcite . )
we have 1 . @xmath788=[e , g ] = 0 $ ] for all @xmath35 , 2 .
@xmath785^f - [ fg , h ] + [ f , gh ] - [ f , g ] = 0 $ ] for all @xmath789 . here , we used additive notation for @xmath790 to emphasize that it is abelian . from the lemma
we get a cohomology class @xmath791 $ ] in @xmath792 .
now we show that the class does not depend on the choice of a section @xmath562 .
let @xmath793 be another section ( which is also regarded as a lifting of each @xmath37 in @xmath2 to an isomorphism @xmath794 covering @xmath795 ) .
then , there exists @xmath796 such that @xmath797 for all @xmath35 .
the factor set @xmath791'$ ] corresponding to @xmath793 is @xmath798'&= & \beta(g ) \sigma(g )
\beta(h ) \sigma(h ) \sigma(gh)^{-1 } \beta(gh)^{-1}\\ & = & \beta(g ) + \sigma(g ) \beta(h ) \sigma(g)^{-1 } + \sigma(g ) \sigma(h ) \sigma(gh)^{-1 } + \beta(gh)^{-1}\\ & = & [ g , h ] + \beta(h)^g - \beta(gh ) + \beta(g).\end{aligned}\ ] ] therefore , @xmath791 $ ] and @xmath791'$ ] are cohomologous and give the same cohomology class @xmath799 \in h^2 ( g ; { \mathbb{z}}/ 2{\mathbb{z}})$ ] .
once a @xmath2-invariant spin structure ( satisfying assumption [ as : spin ] ) is fixed , the spin profile of @xmath5 is uniquely determined as seen in the previous subsection .
we now analyze the dependence of the spin profiles @xmath799 $ ] on the choice of spin structures on @xmath5 .
[ prop : diffspinspf ] let @xmath1 and @xmath801 be two @xmath2-invariant spin structures and denote the resulting spin profiles by @xmath802 $ ] and @xmath803 $ ] , respectively .
then , there exists an exact sequence @xmath804 ; { \mathbb{z}}/2)\ ] ] such that the first map sends the difference of two spin structures @xmath805 to the difference of two spin profiles @xmath802- [ { \rm spf_l^t}]$ ] .
the proof will occupy the remainder of this subsections .
if @xmath5 is spin , the fibration @xmath806 induces the following short exact sequence @xmath807 the inclusion of the fiber induces the first injection @xmath808 .
one can check that this is in fact a central extension .
let @xmath809 be a homomorphism , which is a section of @xmath188 , i.e @xmath810 .
[ lem : spinseciso ] a choice of spin structures on @xmath5 is equivalent to a choice of a section @xmath1 of the map @xmath188 of . if we are given a spin structure @xmath754 on @xmath5 , then from the diagram @xmath811\ar[d ] & \pi_1(p_{spin } ( l ) ) \ar[r]^{\quad q } \ar[d]^{\pi } & \pi_1 ( l)\ar[r]\ar[d]^{= } & 1 \\ 1 \ar[r ] & { \mathbb{z}}/2 { \mathbb{z}}\ar[r ] & \pi_1 ( p_{so } ( l ) ) \ar[r]^{\quad p } & \pi_1 ( l ) \ar[r ] & 1 , } \ ] ] @xmath812 is a desired section .
note that @xmath813 conversely , suppose there exists a section @xmath814 of .
then , the image of @xmath1 is an index 2 subgroup of @xmath815 , which determines a 2-fold covering of @xmath786 .
denote this covering by @xmath754 and let @xmath730 be a general fiber of @xmath816 .
@xmath817\ar[d ] & \pi_1(p_{spin } ( l ) ) \ar[r]^{\quad q } \ar[d]^{\pi } & \pi_1 ( l)\ar[r]\ar[d]^{= } & \pi_0 ( f ) \\ 1 \ar[r ] & { \mathbb{z}}/2 { \mathbb{z}}\ar[r ] & \pi_1 ( p_{so } ( l ) ) \ar[r]^{\quad p } & \pi_1 ( l ) \ar[r ] & 1 } \ ] ] from the construction , @xmath195 in the diagram above is an isomorphism and hence @xmath818 .
then , ( * ? ? ?
* theorem 1.4 ) tells us that @xmath754 is indeed a spin structure on @xmath5 .
a choice of such a section geometrically means a choice of trivialization of @xmath819 along each based loop in @xmath820 .
if @xmath821 is another section , then their difference @xmath822 is a group homomorphism from @xmath823 to @xmath790 .
i.e an element of @xmath824 now , the bundle @xmath825 is @xmath2-equivariant , and hence we can consider an induced map of orbifold fundamental groups @xmath826 \to \pi_1^{orb } [ l / g].\ ] ] note that @xmath827 is surjective . indeed given any path @xmath282 from @xmath100 to @xmath287 for @xmath35
, @xmath828 is a trivial @xmath558-bundle over @xmath282 .
thus , if @xmath115 is a base point in @xmath786 lying over @xmath100 , we can find a path from @xmath115 to @xmath829 which covers @xmath282 .
the following lemma is straightforward .
[ lem : spgpdia ] we have the commutative diagram of group homomorphisms as below .
@xmath830 & 1 \ar[d ] & 1 \ar[d ] & \\ 1 \ar[r ] & { \mathbb{z}}/2 { \mathbb{z}}\ar[r]\ar[d ] & { \mathbb{z}}/2 { \mathbb{z}}\ar[r]\ar[d ] & 1 \ar[r]\ar[d ] & 1\\ 1 \ar[r ] & \pi_1 ( p_{so } ( l ) ) \ar[r]\ar[d]_{p } & \pi_1^{orb } \left [ p_{so } ( l ) / g \right ] \ar[r]\ar[d]_{p^{orb } } & g\ar[r]\ar[d]^{= } & 1\\ 1 \ar[r ] & \pi_1 ( l ) \ar[r]\ar[d ] \ar@/_/[u]_{s } & \pi_1^{orb } [ l / g ] \ar[r]\ar[d ] & g \ar[r]\ar[d ] & 1\\ & 1 & 1 & 1 & \\ } \ ] ] analogously to lemma [ lem : spinseciso ] , we show that the splitting of the second column in the above diagram is equivalent to having a @xmath2-action on the spin bundle induced by the section @xmath1 of . in other words ,
the obstruction to lift the @xmath2-action on @xmath786 to @xmath754 lies in the second group cohomology class determining the group extension in the second column .
[ obst_porb ] fix a spin structure on @xmath5 which is induced by a section @xmath831 of .
the @xmath2-action on @xmath5 ( and @xmath786 ) can be lifted to @xmath754 if and only if there exists a section @xmath832 of @xmath827 which fits into the following diagram .
@xmath833^{s } \ar[r ] & \pi_1^{orb } [ l / g ] \ar[d]^{s^{orb}}\\ \pi_1 ( p_{so } ( l ) ) \ar[r ] & \pi_1^{orb } \left [ p_{so } ( l ) / g \right ] } \ ] ] where horizontal maps come from natural exact sequences @xmath834 \to g \to 1\ ] ] and @xmath835 \to g \to 1.\ ] ] firstly , suppose there is a section @xmath832 .
we consider the universal cover @xmath836 of @xmath786 which also can be regarded as a covering of @xmath837 $ ] . since @xmath836 is simply connected , it is the orbifold universal cover of @xmath838 $ ] and hence @xmath839 $ ] acts on it .
@xmath840 $ ] and its subgroup @xmath823 also act on @xmath836 via @xmath841 \right ) \leq \pi_1^{orb } \left [ p_{so } ( l ) /g \right]$ ] and @xmath842 $ ] .
we take the quotient of @xmath836 by @xmath820-action so that the resulting quotient space is exactly the spin bundle @xmath754 associated to @xmath1 .
then , clearly @xmath843 / \pi_1 ( l ) \cong g$ ] acts on @xmath754 . on the other hand , if we have a @xmath2-action on @xmath754 which makes @xmath816 and @xmath844 @xmath2-equivariant , then we get a group homomorphism @xmath845 \to \pi_1^{orb } [ l / g]$ ] and @xmath846 \to \pi_1^{orb } \left [ p_{so } ( l ) /g
\right]$ ] .
the first map should be an isomorphism since the only element possibly contained in the kernel is a loop in the fiber and the fiber is simply connected .
thus , @xmath847 \ar[dr]^{\cong } \ar[rr ] & & \pi_1^{orb } \left [ p_{so } ( l ) /g
\right ] \ \\ & \pi^{orb } \left [ p_{spin } ( l)/g \right ] \ar[ur ] & } \ ] ] gives a desired section . to prove proposition [ prop : diffspinspf ] , we need a slightly different commutative diagram , which is obtained from the above .
the following lemma is given for this purpose , whose proof is elementary and omitted .
let @xmath848 be the image of the composition of inclusions @xmath849 $ ] .
if the section @xmath1 corresponds to the @xmath2-invariant spin structure on @xmath5(see lemma [ lem : spinseciso ] ) , then @xmath848 is a normal subgroup of @xmath850 $ ] . consider the central column of the diagram in lemma [ lem : spgpdia ] , and divide the 2nd and the 3rd rows by @xmath848 and @xmath823 respectively to obtain the following diagram @xmath851 & 1 \ar[d ] & 1 \ar[d ] & \\ 1 \ar[r ] & 1 \ar[r]\ar[d ] & { \mathbb{z}}/2 { \mathbb{z}}\ar[r]\ar[d ] & { \mathbb{z}}/2 { \mathbb{z}}\ar[r]\ar[d ] & 1\\ 1 \ar[r ] & s(\pi_1 ( l ) ) \ar[r]\ar[d]_{\cong } & \pi_1^{orb } \left [ p_{so } ( l ) / g \right ] \ar[r]\ar[d]_{p^{orb } } & g_s ' \ar[r]\ar[d ] & 1\\ 1 \ar[r ] & \pi_1 ( l ) \ar[r]\ar[d ] & \pi_1^{orb } [ l / g ] \ar[r]\ar[d ] & g \ar[r]\ar[d ] & 1\\ & 1 & 1 & 1 & .\\ } \ ] ] note that the 2nd column is independent of the choice of spin structure @xmath1 .
however , the 3rd column depends on the choice of @xmath848 , and only makes sense if @xmath1 is a @xmath2-invariant spin structure ( so that @xmath848 is a normal subgroup ) .
recall that the obstruction to lift @xmath2-action to the spin bundle lies in the existence of a section of @xmath827 .
the bottom left square diagram of shows that such a section exists at least on the subgroup @xmath820 of @xmath843 $ ] .
therefore , the obstruction descends to a group extension in the last column .
( see ( * ? ? ?
* ( ex 6.6.4 ) ) . )
therefore , the extension of @xmath2 in the last column gives rise to the second cohomology class @xmath852 \in h^2(g,{\mathbb{z}}/2)$ ] , the spin profile of the spin bundle determined by @xmath1 .
for a different @xmath2-invariant spin structure @xmath853 , we get a similar diagram for the extension @xmath854 , and a possibly different spin profile @xmath855 \in h^2(g,{\mathbb{z}}/2)$ ] .
the discussion in the first paragraph of the proof of lemma shows that @xmath856 genuinely acts on @xmath754 , but its quotient @xmath2 may not .
we recall from ( * ? ? ?
* ( ex 6.6.4 ) ) that from the 2nd and the 3rd columns of the diagram canonically induce a homomorphism @xmath857 ; { \mathbb{z}}/2)$ ] which sends the cohomology class of the extension of the 3rd column to the one of the 2nd column .
it implies that the cohomology class @xmath852 $ ] maps to the extension class of the 2nd column in @xmath858 ; { \mathbb{z}}/2)$ ] under this induced homomorphism .
but , the image of @xmath852 $ ] is independent of the choice of spin structure @xmath1 .
this proves that the image of @xmath852 $ ] and @xmath855 $ ] coincide under this homomorphism . from (
* ( 6.8.3 ) ) , we have the following exact sequence called _ lyndon - hochschild - serre spectral sequence _ from the 3rd row of the diagram : @xmath859 ; { \mathbb{z}}/2).\ ] ] therefore , the difference of two spin profiles @xmath852- [ { \rm spf}_l^t]$ ] which maps to zero under the last homomorphism , is in the image of @xmath860 . as both @xmath1 and @xmath801 are @xmath2-invariant spin structures , the difference of two spin structures @xmath822 lies in @xmath860 . the first map in sends it to the difference of two spin profiles @xmath852- [ { \rm spf}_l^t]$ ] .
this proves the proposition [ prop : diffspinspf ] .
in addition , we explain how @xmath822 associates @xmath861 and @xmath862 in a more geometric way . for each @xmath35 , choose a path @xmath266 from @xmath100 to @xmath287 and a trivialization of @xmath863 i.e. an orthonormal frame of @xmath819 along @xmath266 .
then , the frame gives a path connecting @xmath864 and @xmath865 in @xmath742 .
the composition @xmath866 ( representing the element @xmath867[\gamma_h][\gamma_{gh}]^{-1}$ ] in @xmath868 $ ] ) is a loop based at @xmath100 on which a trivialization of @xmath819 is fixed .
if this trivialization agrees with @xmath869 , then we define @xmath785_t=1 $ ] and @xmath785_t = 0 $ ] otherwise .
@xmath870_t$ ] is exactly the factor set representing @xmath801 in the cohomology .
now , it is not difficult to check that the factor set corresponding to @xmath1 is given by @xmath784_s = ( s - t)(l_{g , h})+ [ g , h]_t\ ] ] where @xmath871 is the evaluation of an @xmath83-cocycle @xmath872 at @xmath873 . *
( i ) * we finish the section with a short discussion on orientation reversing actions .
suppose that @xmath5 is oriented , and a half of @xmath2-action on @xmath5 is orientation reversing .
then , @xmath874 has opposite orientation compared to @xmath819 for an orientation reversing @xmath37 .
we recall from @xcite the following well - known fact on an isomorphism between a vector bundle and itself with the reversed orientation .
let @xmath875 be an oriented vector bundle with transition functions @xmath876 on a cover @xmath877 of @xmath71 .
choose maps @xmath878 , and define the new transition maps by @xmath879 .
then a new vector bundle @xmath880 is obtained which is isomorphic to @xmath730 via the maps @xmath881 .
different choices of @xmath881 yield isomorphic bundles .
similarly , by choosing @xmath882 , we can find an isomorphism from spin structures @xmath883 to @xmath884 .
this provides an isomorphism between sets of spin structures of @xmath730 and @xmath880 . back to our case , consider an oriented lagrangian submanifold @xmath5 and its spin structure @xmath885 .
[ as : spin2 ] we require that the spin structure @xmath886 ( of @xmath887 ) is isomorphic to the given spin structure @xmath885(of @xmath5 ) for each orientation reversing @xmath888 in the sense described above . for each @xmath37
, we choose an isomorphism @xmath889 then , the spin profile of @xmath5 is defined by as well .
seidel @xcite used an isomorphism between pin structures @xmath890 the additional factor @xmath891 is not necessary for us since we only deal with orientable lagrangian submanifolds here and hence @xmath891 is a trivial bundle . + * ( ii ) * here , we explain more details on spin profiles when @xmath5 is not connected ( but a @xmath2-orbit ) .
in such a case , we have made assumption [ ass : ldisconnspf ] . as observed before
, @xmath763 is a locally constant function , but not necessarily a constant function when @xmath5 is not connected , since signs in may also depend on connected components of @xmath5 .
we do not know how to handle the general case , but we add more arguments under the assumption [ ass : ldisconnspf ] together with [ as : spin disconn ] below .
first of all , from assumption [ ass : ldisconnspf ] there exists a normal subgroup @xmath892 of @xmath2 such that it preserves each connected component of @xmath5 , and @xmath893 parametrizes the set of components .
[ as : spin disconn ] the following exact sequence splits @xmath894 for the rest of the paper , we will only deal with disconnected lagrangian submanifolds satisfying the above assumption also .
so , one can express @xmath2 as a semi - direct product @xmath895 where @xmath896 . for later use ,
let @xmath897 determine the semi - direct product @xmath895 .
the main advantage here is that the quotient group @xmath898 also can be viewed as a subgroup of @xmath2 and hence , acts on @xmath5 purely by permuting connected components . in particular
, we can label connected components of @xmath5 by @xmath899 .
we fix a @xmath892-invariant spin structure on the connected component @xmath900 corresponding to @xmath901 , and pull it back to the component @xmath902 corresponding to @xmath903 by @xmath904 . as before we take a lift of @xmath892-action on the spin bundle on @xmath900 which gives rise to a spin profile @xmath905 . likewise
, we get a spin profile @xmath906 for each connected component @xmath902 associated to @xmath907 .
assumption [ ass : ldisconnspf ] implies that @xmath908 for all @xmath909 . for @xmath910 and @xmath911 , note that the connected component containing @xmath265 is labeled by @xmath912
so , @xmath913 and @xmath914 .
now we define a map @xmath915 by the following diagram : @xmath916 \ar@{=}[d ] & p_{spin } ( l)_{g\cdot x } \ar@{=}[d]\\ p_{spin } ( l)_{x_0 } \ar[r ] & p_{spin } ( l)_{\phi_a ( g_0 ) \cdot x_0 } } \ ] ] where we abbreviate @xmath917 to @xmath918 to emphasize that the second row only depends on the choice of lift of @xmath892-action to the spin bundle on @xmath900 . under assumption
[ ass : ldisconnspf ] , the above choice of lifts produces @xmath763 which does not depend on connected components of @xmath5 .
the condition @xmath908 implies that @xmath919 is invariant under the conjugation action of @xmath302 on @xmath892 .
the rest of the proof follows easily by the direct computation and is omitted .
analyzing component - preserving subgroups is subtle in general since this involves a study of conjugacy relation among subgroups in @xmath2 .
for example , two component - preserving subgroups @xmath920 and @xmath921 are always conjugate to each other , but possibly conjugated by more than one element in @xmath2 .
we have only dealt with the case when @xmath2 can be decomposed nicely into component preserving " and permuting " subgroup so that there is no complicated issue on conjugacy classes .
we postpone more general discussion to @xcite .
the floer action functionals are single - valued for exact symplectic manifolds , and hence novikov theory is not needed in such cases .
equivariant fukaya categories can be constructed in a relatively simple way ( than the general case ) , and in fact this only requires a choice of a group cohomology class in @xmath6 .
we remark that seidel defined equivariant fukaya category for @xmath922 , and it corresponds to the non - trivial group cohomology class @xmath38 in our language . in this section
, definitions of equivariant lagrangian branes and equivariant fukaya categories for exact symplectic manifolds will be provided .
we will use notations and constructions by seidel in @xcite to which we refer readers for more details . at the end
, we consider @xmath2-invariant parts of morphisms in these categories , which defines orbifolded fukaya categories ( or , its first approximation as we should add the theory of bulk deformation by twisted sectors ) .
let @xmath4 be an exact symplectic manifold with an action of a finite group @xmath2 .
suppose that we have a @xmath2-invariant quadratic volume form @xmath923 on @xmath41 , and a @xmath2-invariant almost complex structure @xmath924 .
let @xmath925 be a square phase map of @xmath51 defined by @xmath926 .
a lagrangian brane in the sense of @xcite is a triple @xmath927 where @xmath5 is an exact lagrangian submanifold with a grading @xmath928 satisfying @xmath929 and a pin structure @xmath930 on it . for simplicity , we assume that @xmath5 is oriented , and @xmath930 is a spin structure on @xmath5 .
let @xmath1 be a group cohomology class @xmath6 .
[ def : seqvbrane ] a triple @xmath931 is called an _ @xmath1-equivariant brane _ if 1 .
@xmath5 is preserved by the @xmath2-action on @xmath41 ; 2 .
@xmath932 is @xmath2-invariant ; 3 . the isomorphism class of @xmath930 is @xmath2-invariant and has a spin profile @xmath1 .
we review each condition .
if @xmath5 is not preserved by @xmath2-action , then one can consider its equivariant family @xmath933 as an equivariant immersion from @xmath934-copies of @xmath5 to @xmath41 .
for simplicity , we require that each intersection @xmath935 is either empty or equal to @xmath5 for all @xmath35 .
condition on @xmath2-grading can be met as follows . if @xmath5 is connected , then @xmath936 is a locally constant function .
thus , @xmath937 on @xmath5 , but since @xmath938 we have @xmath939 .
that is , @xmath302 is automatically @xmath2-invariant .
if @xmath5 has several connected components , we first fix a grading on one component which is invariant under the action of the subgroup preserving the given component , and then define the gradings on other components by the group action .
as revealed before , a spin structure @xmath930 has an associated spin profile .
if it is not equal to @xmath1 , one may try to change @xmath930 to another spin structure and match the associated spin profile with @xmath1 making use of the proposition [ prop : diffspinspf ] .
however , it is not possible in general and @xmath940 is , in fact , a critical condition .
the setup of floer cochain complexes in @xcite briefly goes as follows .
let @xmath941 be two exact lagrangian branes such that @xmath28 and @xmath54 intersect transversely at @xmath942 .
( in fact , transversal intersection conditions are not necessary as we are able to choose perturbation data @xmath943 instead .
we return to this point in a moment . ) then , the pair of linear lagrangian branes ( @xmath944 ) at @xmath144 is given by @xmath945 their index and orientation space are defined as in ( * ? ? ?
* ( 11.25 ) ) and denoted by @xmath946 given a regular floer datum , the graded floer cochain group is defined as @xmath947 ( see ( * ? ? ? * ( 12.16 ) ) . ) if one changes the pin structure @xmath948 by multiplying @xmath949 for @xmath944 , then the resulting pin structure is isomorphic to the original one , but the orientation space @xmath950 is altered by the multiplication of @xmath951 . a floer datum in @xcite for a pair of objects @xmath27 consists of @xmath952 , c^\infty_c(m,{\mathbb{r } } ) ) , \qquad j_{01 } \in c^{\infty}([0,1 ] , \mathcal{j}),\ ] ] where @xmath953 is the space of compatible almost complex structures on @xmath41 .
they are required to satisfy that @xmath954 is transverse to @xmath54 where @xmath955 is the time one map of the hamiltonian vector field of @xmath956 . given @xmath35 , one can push forward the floer datum to get @xmath957 in an obvious way we are now ready to define @xmath1-equivariant fukaya categories ( modifying the construction of fukaya categories by seidel ) .
[ thm : exeqfuk ] let @xmath4 be an exact symplectic manifolds . for @xmath8 , we define the _ @xmath1-equivariant fukaya category _ as follows .
objects of the @xmath1-equivariant fukaya category are @xmath1-equivariant branes .
a morphism between two @xmath1-equivariant branes @xmath958 and @xmath959 is given by @xmath10 , on which @xmath2 acts linearly .
there exist @xmath11-operations @xmath960 which are compatible with the @xmath2-action .
i.e. for @xmath961 @xmath962 the definition above should include a @xmath2-equivariant flat vector bundle for each @xmath1-equivariant brane , but we omit it for simplicity and postpone it to section [ sec : orbibundle ] . on the technical level , we should be able to handle the problem of equivariant transversality involved in the @xmath11-operations on fukaya categories .
assuming transversality for a moment , we show that @xmath2-actions on morphisms can be defined canonically if both lagrangian branes have the same spin profile @xmath1 .
[ lem : transassume ] consider two @xmath1-equivariant branes @xmath941 .
suppose there exists a @xmath2-invariant regular floer datum for these branes
. then , there exists a natural @xmath2-action on the floer cochain group @xmath963 .
the equivariance of perturbation data implies that for each @xmath35 , the pulled - back of each equivariant brane @xmath964 under the @xmath37-action is isomorphic to @xmath964 itself for @xmath18 .
hence , it is easy to see that for the intersections @xmath942 and @xmath965 , @xmath966 the precise isomorphism on determinant lines @xmath967 has to be considered carefully .
recall that determinant line is defined from a lagrangian path @xmath537 from @xmath968 to @xmath969 ( from the grading of each lagrangian ) whose spin structure is determined from the spin structure of each lagrangian . since grading is @xmath2-invariant , @xmath37-action image of the lagrangian path @xmath537 for @xmath950 can be used to define @xmath970 .
here @xmath971 has an induced canonical spin structure from @xmath28 and @xmath54 ( not from the @xmath37-action ) .
this defines an isomorphism @xmath972 but the ( canonical ) spin structure for @xmath970 from @xmath28 and @xmath54 at @xmath973 can be compared to the spin structure of @xmath950 via group action , from the choice of a lift @xmath974 for each @xmath37 ( @xmath18 ) . since the lifts @xmath975 for @xmath35
do not define a group action due to sign error , the maps @xmath976 do not define a group action on @xmath977 , either .
instead , we have @xmath978 however , both @xmath26 have the same spin profile @xmath979 by the definition of @xmath1-equivariant branes . hence , the sign error cancels out and the maps @xmath980 define a @xmath2-action on the floer cochain group . in what follows , we shall describe how to take care of equivariant transversality issue .
indeed , one can adapt the following algebraic technique introduced by seidel in the exact case .
* lemma 4.3)[hompert ] let @xmath981 be an @xmath982-category with an action of a finite group @xmath2 , and @xmath983 , the associated cohomology level category .
suppose that we have another graded linear category @xmath45 with a @xmath2-action , and an equivalence @xmath984 , which is @xmath2-equivariant .
then , there is an @xmath982-category @xmath985 with @xmath986 isomorphic to @xmath45 , which carries an action of @xmath2 , and an equivariant @xmath982-functor @xmath987 such that @xmath988 .
seidel observed that one can make @xmath2 act freely on the set of objects of @xmath981 via introducing additional labels to objects in @xmath985 with elements of a group @xmath2 ( possibly making several copies of the same object ) .
now , it is much easier for @xmath981 to become @xmath2-equivariant .
then , the homological perturbation type lemma above is applied to obtain an equivariant @xmath11-structure on the category @xmath985 .
let us spell out how to use this lemma to construct the @xmath11-operations in the @xmath1-equivariant fukaya category without a transversality assumption .
first , we define the @xmath11-category @xmath981 and a free @xmath2-action on it .
an object of @xmath981 is given by @xmath989 for @xmath35 .
here , @xmath990 is an @xmath1-equivariant brane of @xmath41 and @xmath991 denotes an @xmath1-equivariant brane given by @xmath992 .
( @xmath5 and @xmath993 are @xmath2-invariant as usual . )
a morphism between @xmath994 and @xmath995 is @xmath996 we fix the perturbation datum @xmath997 of @xmath998 for each @xmath35 and use the push forward datum @xmath999 for an arbitrary pair @xmath1000 @xmath11-operations on @xmath981 are defined in a standard way with aid of similarly chosen perturbation data for higher operations @xcite .
[ lem : gactionond ] @xmath11-category @xmath981 admits an action of @xmath2 . on the object level , @xmath35 sends @xmath1001 to @xmath1002 .
the @xmath2-action on the set of objects is obviously free due to the additional index @xmath135 in @xmath1001 .
now we clarify how @xmath2 acts on the morphism spaces . for simplicity ,
let us look at the following @xmath37-action @xmath1003 and other cases can be dealt in the same way .
let @xmath523 transversally intersect at @xmath144 and denote @xmath1004 by @xmath973 .
then , we have @xmath1005 from the @xmath2-invariant gradings .
then , the group action @xmath1006 is defined in an obvious way .
recall that perturbation data for the left and the right hand sides are @xmath1007 and @xmath1008 respectively .
moreover since the spin structure for the brane @xmath1009 is defined as the pulled - back one via @xmath691 , the @xmath37-action is compatible with @xmath11-operations .
therefore , the map @xmath1010 defines a group action on @xmath981 .
no ambiguity arises since @xmath37 always sends an object to another .
we do not have to identify @xmath1011 and @xmath930 for @xmath981 as they are now spin structures for different objects @xmath1012 and @xmath1013 in @xmath981 .
thus , @xmath981 can actually be enlarged to include equivariant branes with various spin profiles . according to lemma
[ hompert ] , it suffices to define a graded linear category @xmath45 with a @xmath2-action and a @xmath2-equivariant equivalence @xmath1014 in order to define the @xmath1-equivariant fukaya category .
the graded linear category @xmath45 is a refined version of donaldson - fukaya category in the sense that objects are given with gradings and pin structures .
[ donald ] fix @xmath8 .
the linear category @xmath45 is defined as follows .
objects of @xmath45 are @xmath1-equivariant branes @xmath1015 s .
a morphism between @xmath958 and @xmath959 is given by @xmath1016 .
the composition of morphisms is defined by the triangle product as usual .
note that morphisms are given by the floer cohomology groups , rather than floer cochain complexes .
our next task is to endow @xmath45 with a group action .
we carry out the construction of a group action on @xmath45 , making use of the idea of weak group actions on chain complexes .
it is employed in @xcite to deal with the morse complex of a non - invariant function on a global quotient and we , in fact , proved that the weak group action induces a strict group action on the morse homology . here , we need a slight generalization since the product structure is also required to be compatible with the @xmath2-action at least in a weak sense .
[ condhomopert ] there exist a @xmath2-action on the linear category @xmath45 and a @xmath2-equivariant functor @xmath1017 , where @xmath1018 is the cohomology level category of @xmath981 .
the sign issue arising from spin profiles can be handled as in lemma [ lem : transassume ] , so we will discuss the question on perturbation data only .
recall that the well - definedness of donaldson - fukaya category uses the following canonical isomorphisms .
let @xmath1019 be two perturbation data for a pair @xmath941 .
from now on , we omit @xmath32-terms for notational simplicity . an one parameter family @xmath1020
connecting two such data gives rise to a continuation map @xmath1021 from the lagrangian floer complex for @xmath956 to that for @xmath1022 , which induces a canonical isomorphism between floer cohomology groups @xmath1023 and @xmath1024 .
given three perturbation data @xmath956 , @xmath1022 and @xmath1025 , the continuation map from @xmath956 to @xmath1025 is homotopic to the composition of two continuation maps , one from @xmath956 to @xmath1022 and the other from @xmath1022 to @xmath1025 .
this implies that the canonical isomorphisms on cohomology groups are compatible with compositions .
we consider a naive @xmath2-action on floer complex which is a chain map , and compose it with the continuation map @xmath1026 from some family @xmath1027 from @xmath1028 to @xmath956 : @xmath1029 then , @xmath1030 is clearly a chain map , and we claim that it defines a weak action on the chain complex @xmath1031 , in the following sense : @xmath1032 the proof of the claim goes as follows .
first , one can check that the family @xmath1033 defines a continuation homomorphism @xmath1034 where @xmath1026 is a continuation defined by @xmath1027 .
thus , @xmath1035 the composition @xmath1036 of two continuation homomorphisms is chain homotopic to the continuation map @xmath1037 via homotopy @xmath1038 and we define @xmath1039 to obtain the formula .
this shows that the @xmath2-action is well - defined on morphism spaces of the category @xmath45 .
now , we show that the composition is compatible with the @xmath2-action .
naive @xmath2-action gives the following commutative diagram : @xmath1040^{\qquad
\qquad m_2 } \ar[d]^g & cf(l_0 , l_2;h_{o2 } ) \ar[d]^g \\ cf(l_0 , l_1 ; g(h_{01 } ) ) \times cf(l_{1 } , l_2 ; g(h_{1,2 } ) ) \ar[r]^{\qquad \qquad m_2^g } & cf(l_0 , l_2 ; g ( h_{02 } ) ) . } \ ] ] in order to move back from the second line to the first line , we consider continuation homomorphisms @xmath1041 and @xmath1042 , all of which we denote by @xmath1043 for simplicity .
it is not hard to show that there exist @xmath1044 ( the second component of @xmath11-homomorphism(functor ) between two different perturbation data ) satisfying the following relation : @xmath1045 this shows that the product is well - defined in donaldson - fukaya categories independent of perturbation data , and also implies its compatibility with the @xmath2-action when applied to @xmath1046 .
the functor from @xmath1017 can be easily constructed using the isomorphisms @xmath1047 from the pull - back spin bundles to the original one .
from now on , we drop the exactness assumption and discuss the @xmath2-floer - novikov theory of a general symplectic manifold ( see section [ sec : fntheory ] ) .
as observed in @xmath2-novikov theory , we need to find a subgroup of _ energy zero _ elements .
we will call this set @xmath22 of energy zero elements of @xmath2 , which is an analogue of @xmath53 in @xmath2-novikov theory in section [ sec : general base ] . for each energy zero element
, we will assign a spin structure on the lagrangian bundle data associated to it , using the pre - fixed isomorphisms @xmath1047 .
the composition of energy zero elements will reveal the necessity of the condition that two lagrangian should have the same spin profiles to define a group action on their floer cochain complex .
let @xmath28 and @xmath54 be @xmath2-invariant lagrangian submanifolds of @xmath41 , transversally intersecting with each other .
assume that both @xmath28 and @xmath54 are connected , compact , oriented and they admit spin structures which are preserved by the @xmath2-action in the sense of assumption [ as : spin ] . in case of orientation reversing action , we assume that both @xmath28 and @xmath54 have orientation reversing @xmath2-actions and work under assumption [ as : spin2 ] .
we set the notations as follows : * @xmath1048 : the action of @xmath35 on @xmath17 , * @xmath1049 : the induced action of @xmath37 on the frame bundle of @xmath17 , * @xmath1050 : the specified lift of @xmath1049 on the spin bundle of @xmath17 + ( or @xmath1051 ) for @xmath18 .
the choice of @xmath1050 above gives rise to spin profiles @xmath1052 of @xmath17 , correspondingly .
we choose a base path @xmath24 of the space of paths @xmath1053 so that @xmath1054 for @xmath1055 .
( this is not essential , but simplifies expositions ) .
the naive @xmath2-action on @xmath1053 will be written as @xmath1056 for @xmath1057 .
consider the connected component @xmath451 containing @xmath1058 .
note that @xmath1059 may not lie in @xmath451 .
hence , we take a subgroup preserving the connected component as follows . [
def : compreserve ] we define the subgroup @xmath1060 of @xmath2 as @xmath1061 we will reduce it further to the smaller subgroup , @xmath22 of energy zero elements in @xmath1060 . for @xmath1062 , we have a surface connecting @xmath24 and @xmath56 .
namely , there exists a surface @xmath1063 \times [ 0,1 ] \to m$ ] , such that @xmath1064 the symplectic area and the maslov index of @xmath465 are defined as follows .
the symplectic area of @xmath465 is simply defined by @xmath1065 . for the maslov index of @xmath465
, we construct a loop of lagrangian subspaces along @xmath1066 ^ 2)$ ] in a canonical way .
recall that we have the pre - fixed lagrangian path @xmath1067 along @xmath24 .
then , @xmath1068 is a lagrangian path along @xmath56 from @xmath1069 to @xmath1070 .
( we use the orientation reversal @xmath1071 if @xmath37 is orientation reversing ) . along @xmath1072 for @xmath1073 , we consider @xmath1074 for @xmath1075 $ ] .
the concatenation of the above paths produces a bundle @xmath1076 on @xmath1077 ^ 2 $ ] which is a lagrangian sub - bundle of @xmath1078 on @xmath534 ^ 2 $ ] the maslov index of @xmath1076 is well - defined , and denoted by @xmath1079 .
we will require @xmath55 to have zero symplectic area and zero maslov index , but given a general @xmath1062 , such @xmath55 does not necessarily exist .
[ def : galpha ] we define @xmath1080 elements of @xmath22 are called _ energy zero _ elements .
given @xmath1081 , there can be several choices of @xmath55 with zero symplectic area and zero maslov index .
however , all @xmath55 s satisfying @xmath1082 are @xmath124-equivalent to each other .
( see definition [ def : gammaeq ] for the @xmath124-equivalence relation . )
@xmath22 forms a subgroup of @xmath2 .
let @xmath37 and @xmath135 be elements of @xmath22 .
then , there are @xmath1083 and @xmath1084 whose energies and maslov indices are both zero .
we make a concatenation @xmath1085 then , @xmath1086 connects @xmath24 and @xmath1087 and hence , @xmath1088 is accompanied with the path @xmath1089 satisfying @xmath1090 since @xmath496 and @xmath1091 are invariant under @xmath2-action .
now , we will associate a canonical spin structure of @xmath1076 to each @xmath55 so that we can define the spin-@xmath124-equivalence relation among @xmath55 s as in [ def : spingamma ] .
we will see that there may be at most two spin-@xmath124-equivalence classes for a single @xmath124-equivalence class .
recall that @xmath1092 } = \lambda_{l_0}$ ] has a preferred trivialization @xmath554 \times \mathbb{r}^n \to \widetilde{\lambda}_{l_0 } $ ] and that the spin structure @xmath579 for @xmath516 is fixed by one of its lifting @xmath1093 \times spin(n ) \cong p_{spin}(\lambda_{l_0})$ ] . over @xmath534 \times \{0\}$ ]
@xmath534 \times \{1\}$ ] ) we have @xmath1094 ( resp .
@xmath1095 ) .
gluing of two spin structures at @xmath1096 is done by @xmath1097 for @xmath18 as explained in .
it remains to specify spin structure of @xmath1076 along @xmath1098 $ ] and how to glue it to @xmath1099 s .
let us assume for a moment that the @xmath37-action preserves orientations of @xmath28 and @xmath54 .
by @xmath37-action , the reference path @xmath24 is mapped to @xmath56 .
then , the lagrangian path @xmath1100 over @xmath56 from @xmath1069 to @xmath1070 can be trivialized by @xmath1101 : @xmath1102 \times \mathbb{r}^n \cong g(\widetilde{\lambda}_{l_0}).\ ] ] we take @xmath1103 as a spin structure @xmath1104 . from @xmath1105 , we get an induced isomorphism @xmath1106 \times spin(n ) \cong p_{spin}(g(\widetilde{\lambda}_{l_0})).\ ] ] we need to define an analogue of @xmath1107 at @xmath1108 to identify two spin bundles there .
( @xmath1109 for @xmath18 . ) in fact , @xmath1110 comes into play to glue @xmath1104 with @xmath569 at @xmath1108 .
note that @xmath1111 defines a natural embedding , which we again denote by @xmath1112 consider its composition with @xmath1113 : @xmath1114 we see that @xmath1115 glues @xmath569 and @xmath1104 at @xmath1108 .
( see figure [ spinwg ] ) therefore , we have described the canonical spin structure for @xmath1076 , which depends on the pre - fixed choice of @xmath1116 .
if @xmath37 is an orientation reversing action for both @xmath28 and @xmath54 , we instead use the orientation reversal @xmath1117 along @xmath56 , whose end points can be identified with @xmath1118 and @xmath1119 .
let @xmath1120 \times so({\mathbb{r}}^n ) \to { \widetilde{\lambda}}_{l_0}$ ] be the trivialization of the frame bundle of @xmath1067 induced from @xmath562 . if @xmath1121 denotes @xmath706 with the opposite orientation , then @xmath1122 leads to the trivialization @xmath1123 \times so(({\mathbb{r}}^n)^{op } ) \to { \widetilde{\lambda}}_{l_0}^{op}$ ] .
take an isomorphism between @xmath1124 and @xmath1125 , given by an element @xmath1126 .
, height=96 ] from the extension @xmath1127 \times spin({\mathbb{r}}^n ) \stackrel{\cong}{\to } p_{spin}({\widetilde{\lambda}}_{l_0})$ ] , we have @xmath1128 \times spin(({\mathbb{r}}^n)^{op } ) \stackrel{\cong}{\to } p_{spin}({\widetilde{\lambda}}_{l_0}^{op})$ ] which are isomorphic to the former by multiplying a lift @xmath1129 .
we also have @xmath1130 \times spin(({\mathbb{r}}^n)^{op } ) \stackrel{\cong}{\to } p_{spin}(g{\widetilde{\lambda}}_{l_0}^{op})$ ] .
we next take @xmath1131 as the spin structure of @xmath1071 , and glue @xmath569 and @xmath1132 using @xmath1133 .
so far , we have described the canonical spin structure associated to the lagrangian sub - bundle @xmath1076 over the boundary of @xmath55 .
as the bundle lies over @xmath1077 ^ 2 \cong s^1 $ ] , there are only two possible spin structures .
for each @xmath55 , we associate a number @xmath1134 which is zero if the canonical spin structure is trivial and @xmath83 otherwise . to handle orientation spaces in defining a @xmath22-action in the next section
, we should not only consider @xmath55 but also with its @xmath645 value , i.e. @xmath1135 ( or @xmath58 for short ) .
lastly , we discuss the relation between group operations and @xmath645 .
we want to compare the product @xmath1136 and @xmath1137 of the concatenation @xmath1089 .
[ prop : compadm ] we have @xmath1138 , height=192 ] for @xmath37 and @xmath135 in @xmath22 , we choose paths @xmath55 and @xmath1139 of energy and index zero , respectively .
they are equipped with lagrangian bundles @xmath1076 and @xmath1140 on their boundaries and also the spin structures @xmath1141 obtained by @xmath765 and @xmath1142 respectively .
the concatenation of @xmath55 and @xmath1139 gives @xmath1143 up to @xmath124-equivalence .
@xmath1089 has two canonical spin structures .
one is the canonical spin structure associated to the surface @xmath1089 which is used to define @xmath1137 .
the other is induced from spin structures of @xmath55 and @xmath1139 described as follows .
consider the image @xmath1144 of @xmath1139 and the pull - back bundle @xmath1145 ( on @xmath1144 ) with respect to the @xmath37-action .
if we restrict @xmath1145 to horizontal edges , it coincides with @xmath1146 .
we apply the isomorphism @xmath1113 over the whole horizontal edges ( figure [ cocycsp ] ) so that we obtain a new spin structure over the boundary of @xmath1144 which is isomorphic to @xmath1145 .
let us denote the resulting spin structure as @xmath1147 .
note that the restriction of @xmath1147 to horizontal edges is given by the restriction of the spin structure @xmath1148 on @xmath17 . since @xmath1147 is isomorphic to @xmath1145 on @xmath1144 @xmath1149
we look into the gluing at each vertical ends of @xmath1147 more carefully . on the vertical edges of @xmath1144
, we have @xmath56 on the left and @xmath1150 on the right .
@xmath1147 agrees with @xmath1151 ( resp .
@xmath1152 ) over @xmath56 ( resp .
@xmath1150 ) with a trivialization @xmath1153 \times spin(n ) \cong p_{spin}(g(\widetilde{\lambda}_{l_0}))\ ] ] ( resp .
@xmath1154 \times spin(n ) \cong p_{spin}(gh ( \widetilde{\lambda}_{l_0 } ) ) $ ] ) .
the horizontal and the vertical part of @xmath1151 at the end of @xmath56 is glued by @xmath1155 . on the other hand , @xmath1156 is applied to glue the horizontal and the vertical component of @xmath1151 at the end of @xmath1150 .
( see the right vertical edge of figure [ cocycsp ] . )
the spin structure @xmath1147 can be glued to @xmath1157 along @xmath56 since the spin structure @xmath1157 over the right vertical edge have exactly the same trivializations and gluings as that of @xmath1147 over the left vertical edge .
denote the glued spin structure by @xmath1158 . by construction , it coincides with the canonical spin structure for @xmath1089 if we replace @xmath1159 in by @xmath1160 for gluing on the right vertical edge .
hence , the total difference between these two spin structures are precisely the product of @xmath1161 for @xmath18 ( see ) , and we have proved the claim .
note that if @xmath28 and @xmath54 have the same spin profiles , then differences between @xmath1158 and @xmath1162 at two vertices of the right vertical edge cancel out , which will allow us to define group actions in the next section .
in this section , we will define @xmath22-action ( definition [ def : galpha ] ) on the floer novikov complex of @xmath1163 when spin profiles of @xmath28 and @xmath54 coincide .
[ as : spin3 ] we assume that @xmath1164 the assumption can be weakened to only require that @xmath1165= [ \textnormal{spf}_{l_1}]$ ] .
in fact , if they define the same class in @xmath1166 , then one can modify one of them to fulfill assumption [ as : spin3 ] by changing the lifts @xmath746 s appropriately .
we emphasize again that this condition is necessary to define a group action on the floer complex . for @xmath1081
, we will use the bounding surface @xmath55 connecting the base path @xmath24 and @xmath56 of zero symplectic area and zero maslov index , but such @xmath55 might carry a non - trivial canonical spin structure .
if @xmath60 is @xmath766 for all @xmath37 , then we define @xmath22 action on the covering space @xmath640 without sign corrections , and it gives rise to an action on the floer complex
. however , in general , one should consider @xmath1167 " .
as the action of @xmath1168 " on the covering space does not make sense , we can not define the action of @xmath1167 directly on the covering space .
we will first make an assumption that @xmath1169 for all @xmath35 ( assumption [ ass : sp0 ] ) to give a clear view of the action of @xmath22 on the covering space , and later in subsection [ galpha ] , construct the precise @xmath23-action on the floer - novikov complex without assumption [ ass : sp0 ] .
actual @xmath22 action should be considered as an action between the orientation spaces ( see subsection [ oriline ] ) of the floer - novikov complex and these are spaces that actions of the sign factors @xmath1170 can be made .
let us make the following simplifying assumption just in this subsection .
[ ass : sp0 ] for each @xmath1081 , we can find @xmath55 with @xmath1171 recall that we have defined the universal covering space of the path space @xmath1172 by @xmath478 in subsection [ subsec : fntheory ] , and refined it to get a smaller covering @xmath1173 by introducing spin @xmath124-equivalence relation in subsection [ subsec : spingamma ] .
the @xmath2-action on @xmath41 gives a naive @xmath2-action on path spaces such that @xmath1174 for @xmath1175 , where @xmath1176 .
since this naive @xmath2-action changes the base path from @xmath24 to @xmath56 , it does _ not _ give the correct @xmath2-action for floer - novikov complex .
alternatively , we define the @xmath22-action on @xmath1177 in the following way . for @xmath1081 , we take @xmath55 satisfying .
then , the @xmath22-action on @xmath1177 is defined by first taking the naive @xmath37-action and then attaching @xmath55 .
we define the action of @xmath22 , @xmath1178 , by @xmath1179 ) \mapsto [ w_g \star ( g ( w ) ) , g ( l ) ] = : g \cdot [ w , l].\ ] ] ( see figure [ fig : galphaact ] . )
-action on @xmath1180,height=172 ] provides a well - defined @xmath22-action on @xmath1180 one may choose a different @xmath1181 satisfying to define @xmath37-action , or choose a different representative @xmath1182 $ ] of @xmath498 $ ] , but it is not hard to see that these provide the same map up to spin @xmath124-equivalence . note that we have proved proposition [ prop : compadm ] under the assumption [ as : spin3 ] , which implies that if @xmath55 and @xmath1139 satisfy , @xmath1183 also satisfies .
this shows that indeed defines a group action .
now , we prove an analogue of lemma [ lem : deccomm ] .
the covering map @xmath1184 is @xmath22-equivariant .
moreover the action of @xmath1185 and that of deck transformation group @xmath1186 on @xmath1180 commute with each other .
the projection map sends @xmath1187 $ ] to @xmath1059 , hence the first claim follows .
suppose that an element @xmath1188 is represented by a loop @xmath45 at @xmath24 in the path space @xmath836 .
then , @xmath1189= [ w_g \star ( g ( c ) ) \star ( g ( w ) ) , g ( l ) ] .\ ] ] on the other hand , @xmath1190= [ c \star w_g \star ( g ( w ) ) , g(l)].\ ] ] their difference is represented by the cylinder @xmath1191 note that @xmath1192 and @xmath1193 have the opposite energies , maslov indices , and @xmath645-values .
hence has the trivial spin structure as well as vanishing energy and maslov index , which implies @xmath1194 = c \cdot g \cdot [ w , l].\ ] ] let us review the floer - novikov setting briefly . from the construction , @xmath1195 is exact for the covering map @xmath668 in .
the floer action functional @xmath496 on @xmath640 is defined by @xmath1196 ) = \int w^{\ast } \omega$ ] for @xmath498 $ ] in @xmath640 .
then , it is not difficult to check that @xmath496 is well - defined and @xmath1197 the following lemma demonstrates that @xmath630 is compatible with the @xmath22-action .
@xmath412 1 .
@xmath630 is invariant under @xmath22-action on @xmath1180 .
the set @xmath1198 of critical points of @xmath496 consists of the pairs @xmath550 $ ] where @xmath503 is the constant path with @xmath504 .
the set @xmath1198 is invariant under the action of @xmath666 and @xmath22 .
the @xmath22-invariance of @xmath496 follows from the @xmath2-invariance of @xmath1199 and the definition of @xmath22-action .
in particular , the set of critical points of @xmath496 is preserved by @xmath22 .
the standard variation argument as in @xcite shows that @xmath1198 consists of @xmath550 $ ] s which directly implies that it is invariant under @xmath666 .
let @xmath504 and suppose that the constant path @xmath503 lies in the component @xmath453 .
we consider the relation between @xmath22 and the isotropy group @xmath194 , which is an analogue of lemma [ locgping ] .
[ locgpingfl ] for @xmath504 , we have @xmath1200 in general , if there exists a path from @xmath28 to @xmath54 which is fixed by @xmath35 , then @xmath37 belongs to @xmath22 . if @xmath1201 , then @xmath1062 since it preserves the connected component of @xmath1202 containing @xmath503 . to see that @xmath1081 , take @xmath1203 , where @xmath1204 is a bounding surface from @xmath24 to @xmath503
then , @xmath1205 is a bounding surface from @xmath56 to @xmath1206 . setting @xmath1207
, we get a bounding surface from @xmath24 to @xmath56 which has vanishing energy and maslov index .
the general case can be proved similarly .
we denote by @xmath1208}$ ] the isotropy group at @xmath498 \in { \widetilde{\omega } } ( l_0 , l_1;l_0)$ ] for the @xmath22-action . also , we denote by @xmath1209}$ ] the set of elements of @xmath22 which either act trivially on the spin @xmath124-equivalence class @xmath498 $ ] or send @xmath498 $ ] to a @xmath124(but not spin @xmath124)-equivalent @xmath1182 $ ] .
obviously @xmath1208 } \subset ( g_{\alpha})_{\pm[w , l]}$ ] .
analogously to the lemma [ lem : isolocalgroup2 ] , one might expect that @xmath1210}=g_p$ ] , but unfortunately it is not true in general .
in fact , we are not interested in the isotropy group @xmath1208}$ ] , but rather in @xmath1209}$ ] .
if @xmath550 $ ] are spin @xmath124-equivalent to @xmath1211 $ ] , their orientation spaces will be identified in a canonical way .
( see lemma [ cylinder ] . )
if they are @xmath124-equivalent but not spin @xmath124-equivalent , the identification of orientation spaces are changed by reversing the sign .
we see that in either case , @xmath1209}$ ] sends the orientation space of @xmath550 $ ] to itself . in this setting
, we have the following lemma . [ comparelocgp ]
we have @xmath1212}=g_p.\ ] ] for @xmath1201 , take @xmath1213 as in the proof of lemma [ locgpingfl ] .
we claim that @xmath1214}$ ] if @xmath1215 .
observe that the @xmath37-action on @xmath550 $ ] is nothing but @xmath1216 $ ] .
since @xmath1217 which is spin @xmath124-equivalent to @xmath465 , we have proved the claim . if @xmath1218 , we should use @xmath1219 to define an action of @xmath1220 , but we can not define such an action directly . from assumption [ ass : sp0 ] , there is certain @xmath1181 satisfying which , then , is @xmath124-equivalent but not spin-@xmath124-equivalent to @xmath55 .
note that @xmath1221 is @xmath124-equivalent to @xmath1222 , and hence to @xmath465 .
thus , @xmath1223}$ ] .
the other direction @xmath1224}\subset g_p$ ] is easy to check .
the @xmath22-action on the orientation spaces are crucial ingredient for defining a group action on floer theory of lagrangian intersections .
this signifies that as in the morse case , we do _ not _ want the group action on the set of critical points of the novikov - floer complex , but rather on the set of the orientation spaces associated to critical points .
so far , we have considered the @xmath22-action on the novikov covering @xmath1225 under assumption [ ass : sp0 ] . from now on
, we drop the assumption , and define @xmath22 action directly on the complex @xmath1226 in .
the key part of the construction is how to define an action on the orientation spaces associated to critical points .
we first go back to the definition [ def : orspace ] , and show that the orientation space @xmath688}^-$ ] can be assigned to each spin @xmath124-equivalence class @xmath550 \in cr(l_0 , l_1 ; l_0)^{sp}$ ] without ambiguity . [ cylinder ] orientation bundles @xmath688}^-$ ] for different representatives of @xmath550 $ ] can be canonically identified . hence , the assignment @xmath688}^-$ ] to each @xmath550 \in cr(l_0 , l_1 ; l_0)^{sp}$ ] is well - defined .
let two different representatives @xmath1227 and @xmath552 represent the same spin-@xmath124-equivalence class .
regarding @xmath465 and @xmath1228 as paths in @xmath453 , @xmath1228 is homotopic to @xmath1229 .
thus , @xmath1228 can be thought of as the concatenation of @xmath1230 and @xmath465 .
let @xmath1231 be the associated lagrangian bundle data of @xmath1230 along the boundary .
note that by the definition of spin @xmath124-equivalences , @xmath1230 has energy zero , and @xmath1232 .
furthermore , the associated canonical spin structure is trivial . to relate @xmath551 and @xmath1233 , we consider gluing of @xmath1234 to @xmath1235 at a point .
( see figure [ cylori ] . ) here , the domain of the @xmath1236-operator associated with @xmath1231 is replaced by @xmath1237 ( instead of a rectangle ) to use gluing theorems in subsection [ subsec : gluingthm ] more conveniently . and
@xmath1235,height=144 ] the index formula tells us that @xmath1238 for some @xmath1239 . from lemma [ lem : se11 ] with the trivial spin structure on @xmath1231
, we have an identification @xmath1240 combining and , we get the canonical identification @xmath1241 which is denoted by @xmath688}^-$ ] . now , we explore the group action on the orientation space @xmath688}^-$ ] associated to @xmath550 \in cr(l_0,l_1;l_0)^{sp}$ ] .
it will be soon revealed that the group action below is only well - defined if the spin profile of @xmath28 and @xmath54 are the same .
[ def : aact ] given @xmath1242 , take @xmath55 satisfying @xmath1082 .
then , the action on floer complex @xmath1243 is defined by the linear map @xmath1244 on generators @xmath1245}^-
\mapsto \theta_{[w_g \star g(w),l_{g(p)}]}^-.\ ] ] @xmath1246 is defined as a composition @xmath1247}^- \;\;\stackrel{a^{\theta}_g}{\to}\;\ ; \theta_{[g(w),l_{g(p)}]}^-\;\ ; \stackrel{\textrm{gluing}}{\to } \;\;\theta_{[w_g \star g(w),l_{g(p)}]}^- \stackrel{(-1)^{sp(w_g)}}{\to}\;\;\theta_{[w_g \star g(w),l_{g(p)}]}^-.\ ] ] ( see figure [ gactori ] . ) , height=288 ] the action defined is independent of choice of @xmath55 .
observe that we multiply @xmath1170 if @xmath55 has a non - trivial spin structure .
let us closely look into each part of the map .
we only explain the case when @xmath22 is orientation preserving .
one can proceed similarly for the other case , which we omit .
+ * ( 1 ) * the first map @xmath1248 is the map given by the naive @xmath2-action .
namely , taking @xmath37-action to @xmath550 $ ] and the associated lagrangian bundle @xmath1249 , we get a generator @xmath1250 $ ] and the lagrangian bundle @xmath1251 along @xmath1252 . then
, we consider two cauchy - riemann problems on @xmath584 with lagrangian boundary conditions given by @xmath1249 and @xmath1251 in @xmath1253 and @xmath1254 respectively .
note that the naive @xmath37-action sends the bundle pair @xmath1255 to @xmath1256 .
this induces an action of @xmath37 between two cauchy - riemann problems ( @xmath32 is @xmath2-invariant ) , and in particular , one can send kernel and cokernel of @xmath1235 to those of @xmath1257 .
they define a map @xmath1258 from @xmath1259 to @xmath1260 by definition [ def : orspace ] . + * ( 2 ) * the second map _ gluing _ is needed from @xmath2-novikov theory , since @xmath1258 image gives a path from @xmath56 not @xmath24 .
hence , we restricted to the subgroup @xmath22 , and consider the gluing map of determinant spaces . to define this map
, we need to specify the choices of spin structures . for this , we need to define the spin structure of @xmath1251 carefully .
the desired spin structure of @xmath1251 should have the following property .
we already have a prescribed spin structure on @xmath1261 . hence ,
if we glue @xmath1261 , with @xmath1251 , we will obtain one spin structure on the concatenation @xmath1262 .
but @xmath1222 also has another spin structure induced from the spin structures of @xmath28 and @xmath54 .
we would like to have these two spin structures being equal to each other .
for this purpose , we define the spin structure of @xmath1251 as follows . along top and bottom
edges of @xmath584 we have spin structures induced from @xmath1095 , and @xmath1094 .
since @xmath1263 lies over the circular boundary of @xmath584 for @xmath1264 , it is natural to consider @xmath1265 along this region .
then , we glue it to @xmath1099 making use of @xmath1266 for @xmath18 as before .
observe that the circular boundary part of @xmath584 has the same spin structure as the right vertical edge of @xmath55 .
what we obtain after gluing @xmath55 and @xmath1264 is indeed the canonical spin structure of @xmath1262 ( produced by the spin structure of @xmath28 , @xmath54 , @xmath1267 and @xmath1268 ) .
this is because contributions @xmath1266 appear twice in the gluing above ( once in the spin structure of @xmath55 ( right edge ) and next in the spin structure of @xmath1264 ( left edge ) ) , and hence they cancel out .
we are now ready to define the second map @xmath1269 .
@xmath1269 is an isomorphism @xmath1270}^-\;\ ; \to \;\;\theta_{[w_g \star g(w),l_{g(p)}]}^-.\ ] ] obtained from the gluing of @xmath55 and @xmath1264 at a point using . to be specific , we take @xmath1271 ( instead of @xmath534 ^ 2 $ ] ) and @xmath1272 , and @xmath1273 and @xmath1274 , and @xmath1275 and @xmath1276 .
+ * ( 3 ) * the last component , a multiplication by @xmath1170 is needed to have a well - defined group action .
namely , we want the group action to be independent of @xmath55 chosen . in case that @xmath60 is non - trivial , the gluing ( last step in the figure ) in fact has a hidden contribution of the sign factor @xmath1170 from lemma [ lem : se11 ] : @xmath1277 ( for some @xmath1278 ) .
namely , we can identify @xmath1279 with @xmath1280 " .
therefore , from , we obtain an canonical isomorphism @xmath1281 hence , to cancel out this effect , we have added a multiplication by @xmath1170 . + as explained in * ( 2 ) * , the resulting ( glued ) spin structure on @xmath1262 is the canonical spin structure , induced from those of @xmath28 , @xmath54 and @xmath516 .
therefore , @xmath1246 gives a map between orientation spaces which are determined by canonical spin structures of @xmath550 $ ] and @xmath1282 $ ] . to see that @xmath1283 defines a group action on the floer complex , notice that @xmath1284 , and both @xmath1285 and @xmath1286 are maps between orientation spaces which are determined by canonical spin structures .
the former map is twisted by @xmath1287 whereas the latter map is twisted by @xmath1288 . from the proposition [ prop : compadm ] ,
two twistings coincide precisely when spin profiles of @xmath28 and @xmath54 agree with each other .
lastly , the @xmath22-action on the floer cochain complex is the linear extension of @xmath22-action on @xmath1289 and this is compatible with what actually happens in geometry from the identity .
( see definition [ def : novl01 ] for the precise definition of @xmath696 . )
[ prop : galphacf ] @xmath22-action on @xmath1290 is @xmath696-linear .
we present an enlarged novikov ring @xmath1291 for orbifold floer cohomology which is an analogue of @xmath127}^{rob}$ ] in subsection [ subsec : orbfund ] .
heuristically , @xmath1292 contains elements in @xmath1293 as there exists a path from @xmath24 to @xmath56 for @xmath1294 which gives rise to a genuine loop of paths between @xmath1295 $ ] and @xmath64 $ ] which increases ( or decreases ) the energy non - trivially .
consider the set of homotopy classes of generalized " paths in @xmath451 which begin at @xmath24 and end at @xmath56 for @xmath35 .
then , we take the quotient of it by the set of classes which can be represented by energy , maslov and @xmath645-value zero paths .
( @xmath1296 and @xmath1297 are extended to generalized paths in the obvious way . ) denote this quotient by @xmath1298 . in particular
, @xmath1298 contains an element represented by a path @xmath465 whose end point is @xmath56 for @xmath1299 .
there is an exact sequence @xmath1300 analogously to the third line of the diagram .
@xmath1301 is defined by the completion of the group ring of @xmath1298 over @xmath79 . by mimicking the construction in definition [ def : obnoveta ] ,
the @xmath22-invariant part @xmath1302 admits a structure of @xmath1301-module .
note that @xmath1301 has the @xmath220-grading induced by @xmath1297 and there is a natural inclusion @xmath1303 which fits into the diagram @xmath1304 \ar[d ] & \lambda_{nov } ( r)\\ \lambda_{orb}^r(l_0 , l_1;l_0)^{sp } \ar[ur ] & } \ ] ] where @xmath1305 is obtained by taking the energy and the maslov index , and with an additional multiplication by @xmath1306 .
we omit further details as this new coefficient ring will not be visible if we use the universal novikov field . from now on , we will only use @xmath1307 as our coefficient .
lagrangian floer theory is built upon subtle data of @xmath32-holomorphic discs and strips . in order to have good moduli spaces of such maps ,
an appropriate perturbation scheme is required .
furthermore , we need equivariant transversality for the sake of group actions on floer theory , and it is difficult to achieve both equivariance and transversality at once using simple methods .
we have shown that the algebraic technique due to seidel can be hired to overcome this problem in section [ sec : eqfuex ] for exact symplectic manifolds and lagrangian submanifolds .
note that the bubbling off of discs does not occur in this case .
seidel developed a domain - dependent perturbation scheme for @xmath32-holomorphic strips and polygons which gives rise to the desired @xmath11-relations .
for the general case , however , disc bubblings do occur and the domain dependent perturbation is not good enough .
fukaya and ono @xcite introduced a kind of an abstract perturbation scheme so called kuranishi structure .
fukaya - oh - ohta - ono @xcite have developed it further to achieve transversality .
note that kuranishi structure is , in fact , a tool designed for attaining equivariant transversality more systematically with help of multi - sections .
hence , adding a finite group action to kuranishi structure is not presumed to cause more difficulty .
we will explain how to achieve equivariant transversality in this setup .
we refer readers to @xcite , ( * ? ? ?
* appendix a ) , @xcite for the definition and properties of kuranishi structure .
what we will carry out in this section is roughly as follows .
suppose a space @xmath1308 admits a kuranishi structure and a @xmath2-action .
we shall show that there is an induced kuranishi structure on @xmath1309 whose multi - valued perturbation gives rise to @xmath2-equivariant multi - sections on @xmath1308 if chosen suitably .
let us explain in more detail .
first we briefly explain the terms .
* a1.1 ) a kuranishi neighborhood of a point @xmath188 in a compact metrizable space @xmath72 is a quintuple @xmath1310 such that 1 .
@xmath1311 is a finite dimensional smooth manifold ( possibly with boundary or corner ) ; 2 .
@xmath1312 is a finite dimensional real vector space ; 3 .
@xmath1313 is a finite group acting smoothly and effectively on @xmath1311 and linearly on @xmath1312 ; 4 .
@xmath1314 is a @xmath1313-equivariant map @xmath1315 ; 5 .
@xmath1316 is a homeomorphism to a neighborhood @xmath1317 of @xmath1318 .
given a pair of neighborhoods @xmath1310 and @xmath1319 of @xmath1318 and @xmath1320 , a coordinate change ( * ? ? ?
* a1.3 ) is denoted as @xmath1321 . here
, @xmath1322 is an injective homomorphism , and @xmath1323 is an @xmath1324-equivariant smooth embedding from @xmath1325-invariant open neighborhood @xmath1326 of the origin in @xmath1327 to @xmath1311 , and @xmath1328 is a compatible equivariant embedding of vector bundles @xmath1329 .
these are required to satisfy additional properties which we omit .
note that a coordinate change may exist only in one direction .
* a1.5 ) a kuranishi structure on @xmath72 assigns a kuranishi neighborhood @xmath1310 for each @xmath1318 and a coordinate change @xmath1330 for each @xmath1320 such that 1 .
@xmath1331 is independent of @xmath188 , called virtual dimension of @xmath72 ; 2 . for @xmath1332
, there exists @xmath1333 satisfying @xmath1334 if trying to perturb @xmath1335 to make it transverse to zero section , one needs to construct perturbations inductively . since a coordinate change may exist only in one direction , a clever choice of subcollection of kuranishi neighborhoods
is required .
this subcollection should be ordered for inductive perturbations , and cover @xmath72 .
such a choice is called a _ good coordinate system_. ( see ( * ? ? ? * a.1.11 ) . )
one can always choose a good coordinate system , whose detailed construction has been given in part ii of @xcite .
the proof goes by the induction on the dimension of @xmath1311 , and in each dimension , one glues several kuranishi neighborhoods to obtain bigger kuranishi neighborhoods and finally get a good coordinate system .
there is a notion of tangent bundle of a space @xmath72 with a kuranishi structure , and also a notion of orientations on it , which we omit .
we now recall multi - sections of a kuranishi structure . for @xmath1310 , consider the product of @xmath429 copies of @xmath1312 , denoted as @xmath1336 , and endow it with the action of symmetric group @xmath1337 .
denote the quotient space by @xmath1338 , which has an induced action of @xmath1313 .
there exists a @xmath1313-equivariant map @xmath1339 which sends @xmath1340 $ ] to @xmath1341.\ ] ] ( * ? ? ?
* a1.19 ) an @xmath727-multisection @xmath1 of @xmath1342 is a @xmath1313-equivariant map @xmath1343 .
it is said to be _
liftable _ if there exist @xmath1344 such that its composition with @xmath1345 is @xmath1 . here
@xmath1346 needs not be @xmath1313-equivariant .
each @xmath1347 is called a branch of @xmath1 .
liftable multisections will be considered always . given an @xmath727-multi - section , we obtain an @xmath1348-multi - section by composing it with @xmath1349 map .
an @xmath727-multisection @xmath1 is equivalent to an @xmath1350-multisection @xmath1351 if their induced @xmath1352-multisections are the same .
we identify equivalent multi - sections from now on .
a lifted multisection is said to be transversal to zero if each of its branches is transversal to the zero section .
compatibilities of multi - sections can be handled in the same way as in @xcite and we omit the details .
a family of multisections @xmath1353 is said to converge to a kuranishi map @xmath1354 as @xmath1355 if there exists @xmath727 such that @xmath1353 is represented by an @xmath727-multisection @xmath1356 and @xmath1356 converges to a representative of @xmath1 .
* a1.23)[lem : multisecch ] suppose that a good coordinate system of the kuranishi structure over @xmath72 is given , and that the kuranishi structure has a tangent bundle .
then , there exists a family of multisections @xmath1357 such that it converges to @xmath1358 and @xmath1359 s are transversal to @xmath766 for all @xmath1360 .
a notion of group actions on a space with kuranishi structure is introduced in ( * ? ? ? * a1.3 ) .
[ def : aukr ] let @xmath72 be a space with kuranishi structure .
a homeomorphism @xmath1361 is called an automorphism of kuranishi structure if the following holds : for @xmath1318 and @xmath1362 , there exist kuranishi neighborhoods @xmath1363 of @xmath188 and @xmath1364 with a group isomorphism @xmath1365 , a @xmath1366-equivariant diffeomorphism @xmath1367 and a @xmath1366-equivariant bundle isomorphism @xmath1368 which covers @xmath1369 .
they are required to satisfy @xmath1370 and @xmath1371 for an induced homeomorphism @xmath1372 from @xmath1369 .
these data should be compatible with the coordinate changes of kuranishi structure as in ( * ? ? ?
* a1.47 ) .
we write @xmath1373 for an automorphism of kuranishi structure .
an automorphism @xmath1373 is said to be conjugate to @xmath1374 if @xmath1375 and if there exists @xmath1376 for each @xmath188 satisfying @xmath1377 [ def : gpkr ] consider an action of a finite group @xmath2 on @xmath72 and suppose @xmath72 is compact .
it is called an action on a space with kuranishi structure if each @xmath89-action from @xmath72 to itself lifts to an automorphism @xmath1378 of the kuranishi structure , and the composition @xmath1379 and @xmath1380 is conjugate to @xmath1381 .
the following lemma of @xcite is essential in our application .
* a1.49)[lemxg ] if a finite group @xmath2 acts on a space @xmath72 with kuranishi structure , then the quotient space @xmath1382 has a kuranishi structure .
if @xmath72 has a tangent bundle and the action preserves it , then the quotient space has a tangent bundle .
if @xmath72 is oriented and the action preserves the orientation , then the quotient space has an orientation .
we only give a brief sketch , here .
let @xmath35 and @xmath1318 .
we put @xmath1383 .
take a kuranishi neighborhood of @xmath188 @xmath1384 such that @xmath1311 is @xmath194-invariant .
we want to extend @xmath1313 using @xmath194 to obtain another finite group @xmath1385}$ ] which acts on @xmath1311 . then , @xmath1385}$ ] will serve as a local group for a kuranishi neighborhood of @xmath1386 $ ] in @xmath1382 .
we assume that the @xmath194-action on @xmath1311 is effective , and otherwise , it can be made effective by adding a finite dimensional representation to @xmath1311 and @xmath1312 .
both @xmath194 and @xmath1313 can be considered as a subgroup of the group of diffeomorphisms of @xmath1311 .
the extension @xmath1385}$ ] is the group generated by these two subgroups @xmath1387 in @xmath1388 . from definition [ def : gpkr ] , there exist @xmath1389 for @xmath1390 and @xmath1391 in @xmath194 such that @xmath1392 also from definition [ def : aukr ] , we have @xmath1393 for @xmath1201 and @xmath1394 .
this implies that the extension group @xmath1385}$ ] satisfies the following exact sequence @xmath1395 } \to g_p \to 1.\ ] ] the @xmath1385}$]-action lifts to an action on @xmath1396 , and @xmath1397 induces @xmath1398}:s_p^{-1}(0 ) /\gamma_{[p ] } \to x / g$ ] .
thus , we get a kuranishi neighborhood @xmath1399 } , \psi_{[p ] } , s_p)$ ] of @xmath1386 \in x / g$ ] .
this lemma will be used to find a compatible good coordinate system and equivariant sections thereof .
first , let us consider a good coordinate system on @xmath72 given by a partially ordered set @xmath1400 , which parametrizes a collection of kuranishi neighborhoods covering @xmath72 .
( technically , we require that this system additionally satisfy properties in ( * ? ? ?
* definition 5.3 ) . )
namely , given @xmath1401 we have a corresponding kuranishi neighborhood @xmath1402 , and partial order describes directions of embeddings .
we say that a good coordinate system on @xmath72 is compatible with the @xmath2-action on @xmath72 if the following holds : 1 . for @xmath1401 and @xmath89 , there exist @xmath1403 such that @xmath1379 sends the kuranishi neighborhood for @xmath1404 to that of @xmath1405 ( this defines a map @xmath1406 such that @xmath1407 ) ; 2 . the map @xmath1379 preserves the partial order of @xmath1408 .
we owe the following lemma to kenji fukaya .
[ lem : fukaya ] let @xmath72 be a space with kuranishi structure together with a @xmath2-action on it . then
, there exist a good coordinate system compatible with the @xmath2-action on @xmath72 . a good coordinate system is chosen by the induction on the dimension of @xmath1311 and by gluing several kuranishi neighborhoods ( in order to get a bigger one satisfying desired properties ) . from the previous lemma
, @xmath1382 is also a space with kuranishi structure , and hence admits a good coordinate system .
then , we glue the corresponding charts in @xmath72 in the same way we construct the good coordinate system for @xmath1382 .
this produces a desired good coordinate system on @xmath72 which is compatible with the @xmath2-action .
we now explain how to choose multi - sections .
if @xmath1409 is an automorphism of a space with kuranishi structure @xmath1409 and @xmath1 is a multi - section , then we can pull back @xmath1 by @xmath1409 in an obvious way .
write the pulled - back multi - section by @xmath1410 .
a multisection @xmath1 is said to be @xmath2-equivariant if for each @xmath35 the corresponding automorphism of kuranishi structure @xmath1379 sends @xmath1 to itself .
if a multisection @xmath1 is @xmath2-equivariant , the zero set of @xmath1 admits the obvious @xmath2-action .
equivariant multi - sections can be constructed from a multi - section for @xmath1382 .
virtual fundamental chains obtained in this way admit natural @xmath2-actions , and evaluation maps become @xmath2-equivariant as well .
let @xmath72 be a space with kuranishi structure equipped with @xmath2-action and fix a @xmath2-compatible good coordinate system .
then the family of multisection @xmath1357 in lemma [ lem : multisecch ] can be chosen to be @xmath2-equivariant .
this is again done by choosing a multi - section for @xmath1382 first .
we examine how to do it in local charts .
let @xmath1399 } , \psi_{[p ] } , s_p)$ ] be a kuranishi neighborhood for @xmath1386 \in x / g$ ] . given a multi - section @xmath1411 ( which is @xmath1385}$]-equivariant by definition )
, the same multi - section @xmath1412 can be regarded as a multi - section for the kuranishi neighborhood @xmath1310 of @xmath1318 since it is @xmath1313-equivariant also .
the @xmath1385}$]-equivariance implies that the multi - section @xmath1412 is @xmath194-equivariant .
we finally deal with transversality of evaluation maps from moduli spaces .
we want to choose multi - sections suitably so that evaluation maps become submersive .
fukaya , oh , ohta and ono @xcite defined lagrangian floer theory on de rham model using continuous families of multi - sections , and constructed smooth correspondences using such gadgets .
it is straightforward to modify the construction of smooth correspondences by using a continuous family of multi - sections which fits into our @xmath2-equivariant setting .
let us illustrate it for the local chart , and leave details to the reader . consider a chart @xmath1310 from lemma [ lemxg ] with @xmath1385}$ ] as before .
let @xmath1413 be a submersion .
a vector space @xmath1414 is chosen to have sufficiently large dimension so that there exists a surjective bundle map @xmath1415 the extended space @xmath1414 is given the trivial @xmath1385}$]-action .
we put @xmath1416 it is no longer @xmath1385}$]-equivariant , so we take ( family of ) multi - sections @xmath1417\ ] ] where @xmath1385 } = \{\gamma_1 , \cdots , \gamma_g\}$ ] . then , @xmath1418 is transversal to @xmath766 and the restriction of @xmath1419 to @xmath1420 is a submersion . for @xmath1421 , we can choose @xmath1422 and make the related construction done in @xmath37-equivariant way .
recall from @xcite that smooth correspondence is defined locally as follows .
let @xmath1423 be a compactly supported smooth differential form on @xmath1311 , which is @xmath1424-invariant .
we put smooth measure @xmath1425 on @xmath1426 so that @xmath1427 .
( for @xmath1421 , we take @xmath1428 . )
[ localcorres ] we define @xmath1429 where @xmath429 is the number of branches of @xmath1418 .
this is a @xmath194-equivariant map .
we can put various @xmath1414 s compatibly in the good coordinate system by using almost the same technique as in @xcite .
let @xmath5 be a ( relatively ) spin lagrangian submanifold in a symplectic manifold @xmath4 .
the floer cochain complex between @xmath5 and itself has a structure of an @xmath11-algebra as observed in @xcite .
if the finite group @xmath2 acts on @xmath41 and preserves @xmath5 , then we can construct a group action on the @xmath11-algebra of @xmath5 which is compatible with @xmath11-operations .
the construction does not involve spin profiles and @xmath2-novikov theory .
we only need equivariant transversality in the previous section with a little bit of algebraic manipulation .
we will mainly discuss the latter in this section .
the actual construction of @xmath11-algebra involves a long sequence of technical constructions .
we do not reproduce the whole steps here , but only indicate how to adapt constructions in @xcite in our setting , omitting most of the steps which can be done by straightforward modification of theirs .
we remark that the ( @xmath1430-)cohomology of the @xmath11-algebra is the floer cohomology @xmath1431 , and in particular , spin profiles automatically match .
this provides a explanation on why the spin profile condition is not necessary in this case .
moreover , we do not have to reduce the @xmath2-action to the action of the energy zero subgroup @xmath22 .
the reason behind it will be discussed in section [ sec:12 ] .
suppose that a finite group @xmath2 acts on @xmath41 effectively , preserving the symplectic form @xmath1199 , almost complex structure @xmath32 , a spin lagrangian submanifold @xmath5 and its isomorphism class of spin structure .
hence , @xmath2 acts on the moduli spaces of stable @xmath32-holomorphic discs bounding @xmath5 , and the local kuranishi model for a stable map is identical under group action . as long as two spin structures @xmath1432 and @xmath885 are isomorphic , the induced orientations on the moduli spaces of discs are the same , and the choice of the particular isomorphism @xmath746 do _ not _ play a role , here .
thus , the following proposition follows from the construction of @xcite and @xmath2-kuranishi structure built in the previous section .
let @xmath1433 , and @xmath1434 .
* theorem 2.1.29 ) the moduli space @xmath1435 of stable @xmath32-holomorphic discs with boundary on @xmath5 with @xmath691 boundary marked points ( cyclically ordered ) and @xmath429 interior marked points has a kuranishi structure with an action of @xmath2 in the sense of definition [ def : gpkr ] .
we can choose family of @xmath2-equivariant multisections @xmath1436 for which the moduli space has a virtual fundamental chain of dimension @xmath1437 .
let @xmath1438 be the boundary evaluation map .
then , the @xmath37-action induces an isomorphism @xmath1439 with commuting diagram @xmath1440_{(a_g ) _ * } \ar[rr]^{\phi_{\beta , k } } & & l \times \cdots \times l \ar[d]_{a_g } \\ { \mathcal m}_{k+1}(l,\beta ) \ar[rr]^{\phi_{\beta , k } } & & l \times \cdots \times l } \ ] ] where @xmath90 in the right column is the diagonal action of @xmath37 . for convenience ,
we work on the de rham model of an @xmath11-algebra , whose construction is described in @xcite . a de rham model @xmath11-algebra has been extensively used in the toric setting recently , where lagrangian submanifolds are @xmath1441-orbits .
@xmath1441-equivariant evaluation maps are automatically submersive , and hence was directly used to define the push - forward of differential forms .
consider the de rham complex @xmath1442 of @xmath5 . a multi - linear map @xmath1443 is defined as follows : @xmath1444 where @xmath1445 is an integration along fibers if @xmath1446 is a submersion . in general , it is the @xmath2-equivariant smooth correspondence described in the previous section .
( see the diagram below . )
@xmath1447_{ev_0 } \mathcal{m}_{k+1}^{main } ( l;\beta)^{\mathfrak{s } } \ar[rr]^{\phi_{\beta , k } } & & { \overbrace{l\times \cdots \times l}^{k } } } \ ] ] we put @xmath1448 as shown in @xcite , @xmath1449 defines an @xmath982-structure on @xmath1442 .
i.e. the family of operations satisfying @xmath1450 where @xmath1451 . note that if @xmath1452 and @xmath41 are manifolds with @xmath2-actions , and if @xmath1453 is a @xmath2-equivariant submersion , then the integration along fiber @xmath1454 is well - defined on the cochain level , and is @xmath2-equivariant .
therefore , we obtain the following lemma .
the operations @xmath1455 are @xmath2-equivariant in the sense that @xmath1456 where @xmath1457 due to the problem running out of kuranishi neighborhood " ( * ? ? ?
* section 7.2 ) , one first construct a @xmath2-equivariant @xmath1458-algebra of the lagrangian submanifold @xmath5 for a large @xmath1459 .
then , the desired @xmath982-algebra is obtained by further applying algebraic formalism as follows .
( see ( * ? ? ?
* theorem 7.2.27 ) for more details . ) in what follows , we will use the ( tensor ) coalgebra language which much reduces the notational complexity .
@xmath1460 gives rise to the coderivation @xmath1461 , and @xmath1462 on the tensor coalgebra @xmath1463 .
we write @xmath1464 , and @xmath1465 .
an @xmath11-homomorphism @xmath50 induces a cohomomorphism @xmath1466 between coalgebras .
one can average an @xmath11-homomorphism to make it @xmath2-equivariant in the following way .
[ avgahom ] let @xmath1467 be an @xmath982 homomorphism between two @xmath982 algebras @xmath1468 and @xmath1469 with @xmath2-actions .
the average @xmath1470 is defined by @xmath1471 then , @xmath1470 gives a @xmath2-equivariant @xmath982 homomorphism .
@xmath1470 is @xmath2-equivariant by construction .
we need to prove that @xmath1472 . from
, @xmath1473 commutes with the @xmath37-action . also , @xmath1473 commutes with @xmath1474 since @xmath50 is an @xmath11-homomorphism .
hence , the lemma follows .
averaging of @xmath11-operations of an non-@xmath2-equivariant @xmath11-algebra does not produce an @xmath11-algebra .
the lemma below follows immediately .
[ compavg ] @xmath1475 and @xmath1476 provided that @xmath1477 is @xmath2-equivariant ( i.e. @xmath1478 ) .
we introduce the notion of @xmath982-homotopies in the existence of the group action .
let @xmath1479 , \{m_k\})$ ] be an @xmath982-algebra over @xmath79 on which a finite group @xmath2 acts and suppose that each @xmath1460 is @xmath2-equivariant .
( @xmath1480 $ ] means the degree shifting . )
the @xmath11-homotopy is defined by employing a model of @xmath1481 $ ] . when @xmath1482 , the @xmath982-algebra @xmath1483 \otimes r[t , dt]$ ] can be used as a model .
( see @xcite for extended @xmath11-operations on @xmath1483 \otimes r[t , dt]$ ] . ) a @xmath2-action on @xmath1483 \otimes r[t , dt]$ ] is defined by giving a trivial @xmath2-action on @xmath1484 $ ] factor .
i.e. the @xmath2-action on @xmath1483 \otimes r [ t , dt]$ ] is @xmath1484$]-linear .
it is easy to see that @xmath982 operations on @xmath1483 \otimes r[t , dt]$ ] are @xmath2-equivariant .
recall that we have an @xmath982 homomorphism @xmath1485 \otimes r[t , dt ] \to c[1]$ ] given by @xmath1486 all other higher components of @xmath1487 are defined to be zero .
@xmath1487 is obviously @xmath2-equivariant and so is @xmath1488 .
two ( @xmath2-equivariant ) @xmath982 homomorphisms @xmath1489 are said to be ( @xmath2-)homotopic to each other , if there exists a ( @xmath2-equivariant ) @xmath982 homomorphism @xmath1490 $ ] such that @xmath1491 and @xmath1492 .
[ equivhpty ] suppose that @xmath45 and @xmath1493 are both equipped with @xmath2-equivariant @xmath982 structures and @xmath1494 is @xmath2-equivariant .
if @xmath361 and @xmath1495 are homotopic , then so are @xmath361 and @xmath1496 .
moreover , one can take a @xmath2-equivariant homotopy between @xmath361 and @xmath1496 .
let @xmath347 be a homotopy between @xmath361 and @xmath1495 .
then , by taking hat " and averaging the equations @xmath1491 and @xmath1492 , we get @xmath1497 we are now ready to state the @xmath2-equivariant version of the @xmath982-whitehead theorem .
let @xmath1498 be a @xmath2-equivariant weak homotopy equivalence between two filtered @xmath982 algebras @xmath1468 and @xmath1469 with @xmath2-actions .
then , there exists a quasi - inverse @xmath1499 of @xmath1477 which is also @xmath2-equivariant .
usual whitehead theorem for @xmath982-algebras in @xcite produces a quasi - inverse @xmath1500 of @xmath1477 . by averaging @xmath1501 as in definition [ avgahom ]
, we obtain the @xmath982 homomorphism @xmath1502 which is @xmath2-equivariant .
since @xmath1477 is @xmath2-equivariant , we may apply lemma [ compavg ] so that @xmath1475 and @xmath1476 .
note that @xmath1503
. then , by lemma [ equivhpty ] , we get @xmath2-equivariant homotopies @xmath1504 in order to construct @xmath11-algebra from @xmath1458 ( or @xmath1505-algebra in the filtered case ) the following theorem is essential .
* theorem 7.2.72 ) ) let @xmath1468 be a filtered @xmath1505 algebra and @xmath1469 a filtered @xmath1506 algebra with @xmath1507 , both of which are assumed to be @xmath2-equivariant .
let @xmath1508 be a @xmath2-equivariant filtered @xmath1505 homomorphism which gives rise to a filtered @xmath1505 homotopy equivalence .
then , there exist a @xmath2-equivariant filtered @xmath1506 algebra structure on @xmath1468 extending the given @xmath1505 algebra structure and a @xmath2-equivariant filtered @xmath1506 homotopy equivalence @xmath1509 extending the given filtered @xmath1505 homotopy equivalence @xmath1501 .
@xmath1510_{a_{n',k ' } } & & \\ & & c_2\\ c_1^{(n , k ) } \ar[uu ] \ar[urr]^{a_{n , k}}_{\mathfrak{h } : \simeq } & & } \ ] ] we just mention how to modify the original proof of @xcite , which deals with the obstruction class of extending @xmath1505 structure to the next level . its @xmath2-equivariant version will be explained only for the unfiltered case for simplicity .
given an @xmath1458 structure with @xmath1511 , the subsequent operation @xmath1512 should be defined to satisfy the next @xmath11-equation @xmath1513 thus , @xmath1512 may be regarded as a @xmath229-cocycle in the cochain complex @xmath1514 , c[1 ] ) , \delta = [ m_1,\cdot]).\ ] ] in addition , if @xmath1458 algebra is @xmath2-equivariant , the left hand side of defines a @xmath2-equivariant map from @xmath1515 $ ] to @xmath1483 $ ] , and hence lies in the subcomplex @xmath1516 , c[1 ] ) , \delta)$ ] .
note that @xmath2-equivariant @xmath1458 homotopy equivalence @xmath1501 preserve the obstruction class @xmath1517 $ ] , and that the obstruction vanishes if the target of @xmath1501 has a @xmath2-equivariant @xmath1518 structure with @xmath1519 .
therefore , we can choose a @xmath2-equivariant @xmath1512 which satisfies for @xmath1468 . in a similar manner ,
the original construction of @xcite can be modified to include the @xmath2-equivariance as above . following the rest of @xcite
, we obtain there exist a @xmath2-equivariant @xmath11-algebra @xmath1520 of a spin lagrangian submanifold @xmath5 , and its @xmath2-invariant part again becomes an @xmath11-algebra , which will be called the @xmath11-algebra of @xmath68 $ ] in @xmath44 $ ] .
the @xmath2-equivariance implies that if @xmath1521 are all @xmath2-invariant , then so is @xmath1522 .
therefore , we obtain an operation @xmath1523 simply by restricting @xmath1524 to the set of @xmath2-invariant differential forms on @xmath5 .
this gives rise to a well - defined @xmath982-structure on @xmath1525 . within this setting
, we may define @xmath2-bounding cochain as a bounding cochain which is @xmath2-invariant .
@xmath2-bounding cochains define boundary deformations of @xmath11-algebra of @xmath68 $ ] .
the average of a bounding cochain does not become a @xmath2-bounding cochain .
the @xmath11-algebra of @xmath68 $ ] constructed here is only concerned with smooth @xmath32-holomorphic discs with boundary on @xmath5 . in order to obtain richer orbifold theory
, one should additionally consider orbi - discs or bulk deformations by twisted sectors , which are introduced by the first author and poddar @xcite in the case of toric orbifolds .
we consider equivariant @xmath982-bimodules and consider lagrangian floer homology of a pair .
we first review the @xmath11-bimodule structure briefly . recall that a boundary operator @xmath1527 was defined in @xcite by @xmath1528 )
= \sum _ { \mu ( [ w',l_q ] ) - \mu ( [ w , l_p ] ) = 1 } \ # \mathcal{m } ( l_1 , l_0 ; [ w , l_p ] , [ w',l_q ] ) \,\ , [ w ' , l_q].\ ] ] here , @xmath1529 , [ w',l_q])$ ] is the moduli space of @xmath32-holomorphic strips @xmath1204 such that @xmath1530 = [ w',l_q]$ ] .
the orientations of these moduli spaces are already well - studied by @xcite .
[ ori ] @xcite suppose that a pair of lagrangian submanifolds @xmath27 is ( relatively ) spin .
then for any @xmath1531 , the moduli space @xmath1529 , [ w',l_q])$ ] of connecting orbits in lagrangian intersection floer cohomology is orientable .
furthermore , orientations on @xmath688}^-$ ] and @xmath1532}^-$ ] and related spin structures canonically determine the orientation on @xmath1533 . in general , disc bubbling phenomenon produce non - trivial @xmath1534 term , and hence @xmath1535 .
the general algebraic structure one obtains is in fact @xmath11-bimodule over @xmath11-algebras of @xmath28 and @xmath54 .
recall that @xmath11-bimodule @xmath41 over @xmath11-algebras @xmath1536 are given by sequence of maps @xmath1537 satisfying an @xmath11-bimodule equation .
such an @xmath11-bimodule is called @xmath2-equivariant , if @xmath2 acts on @xmath41 and @xmath1538 linearly and each @xmath1539 satisfies @xmath1540 now , suppose @xmath28 and @xmath54 admit @xmath2-actions inherited from the one on the ambient symplectic manifold . assume further that the spin structures on @xmath28 and @xmath17 are @xmath2-invariant .
i.e. they are isomorphic to their pull - back under any element of @xmath2 .
then , the spin bundle on @xmath17 gives a spin profile @xmath1541 for @xmath18 .
assume @xmath25 .
since we will only consider the action of the energy zero subgroup @xmath22 of @xmath2 for @xmath28 and @xmath54 , the condition @xmath25 can be weakened to the agreement of @xmath1542 and @xmath1543 when restricted to @xmath1544 . as studied in section [ sec : ku ] , the union of all moduli spaces @xmath1529 , [ w',l_q])$ ] with a fixed symplectic energy , has a kuranishi structure with a @xmath22-action .
therefore , @xmath1545 is @xmath22-equivariant and the homology of @xmath1546 admits a natural @xmath22-action if @xmath1547 .
[ def : flwithout ] with the setting as above , the @xmath22-invariant part of the homology is called the lagrangian floer cohomology of the pair @xmath1548 , [ l_1/g];l_0)$ ] and will be written by @xmath1549 , [ l_1/g]).\ ] ] we take the sum over all possible @xmath24 to define lagrangian floer cohomology of the pair @xmath1548 , [ l_1/g])$ ] . if @xmath338 , we introduced orbifold novikov ring @xmath1550 in subsection [ subsec : newnov ] , and explained that @xmath1550 is not visible if we use the universal novikov field . as in the previous section ,
it is not hard to adapt the construction of @xcite ( or @xcite for the de rham version ) to define @xmath22-equivariant @xmath11-bimodules for the pair @xmath27 . again using equivariant kuranishi perturbations , and adapting the construction of @xcite as in the previous section
, one can prove [ prop:12.4 ] we have a @xmath22-equivariant filtered @xmath11-bimodule @xmath1551 with @xmath1552 over @xmath22-equivariant @xmath11-algebras of @xmath28 and @xmath54 .
now , let us consider the case that @xmath54 is obtained by hamiltonian diffeomorphism of @xmath28 , i.e. @xmath1553 for a time - dependent hamiltonian @xmath347 .
if @xmath1554 \to { \mathbb{r}}$ ] is not @xmath2-invariant , the corresponding hamiltonian perturbation produces a @xmath2-equivariant lagrangian immersion which will be studied in @xcite . here , we only consider @xmath1554 \to { \mathbb{r}}$ ] which are @xmath2-invariant . here
, the @xmath2-invariance of @xmath347 means that each @xmath1555 for @xmath1556 $ ] is a @xmath2-invariant function .
this implies that @xmath1557 is @xmath2-equivariant , and hence , for example , @xmath37-fixed points of @xmath41 can only move to a @xmath37-fixed points by @xmath1557 .
@xmath1558 is , then , preserved by @xmath2 .
since @xmath28 and @xmath1558 are equivariantly isotopic , one can identify their spin bundles and liftings of @xmath2-actions on spin bundles .
thus , spin profiles of @xmath28 and @xmath1558 coincide .
the energy zero subgroup @xmath1559 for the pair @xmath1560 is particularly simple .
indeed , we will show that @xmath22 becomes the entire group @xmath2 .
we first choose a candidate for energy zero path for each @xmath35 as follows .
if we define @xmath1561 by @xmath1562 then @xmath1563 is also @xmath2-invariant ( since @xmath1564 is @xmath2-equivariant . )
fix a generic point @xmath100 in @xmath28 and choose for each @xmath35 a path @xmath266 in @xmath28 from @xmath100 to @xmath264 .
we take a base path @xmath24 from @xmath28 to @xmath1558 to be @xmath1565 .
then , the strip @xmath1566 ^ 2 \to m$ ] defined by @xmath1567 gives a path from @xmath24 to @xmath56 in @xmath1568 .
we claim that this path has zero energy .
@xmath55 is an energy zero path .
we have to show that @xmath1569 ^ 2 } w_g^\ast \omega = 0 $ ] .
observe that @xmath1570 therefore , the energy of @xmath55 is given by @xmath1571 since @xmath1563 is @xmath2-invariant and @xmath1572 .
it directly follows from the lemma that with the setting as above , the energy zero subgroup for a pair @xmath1573 is the entire group @xmath2 .
recall that the @xmath11-algebra is a bimodule over itself .
so we can compare two @xmath11-bimodules @xmath1574 and @xmath1575 over @xmath11 algebra of @xmath28 .
the following proposition of @xcite can be also proved in a @xmath2-equivariant setting along the same line of their proof , and we omit the details .
suppose that @xmath28 is connected and let @xmath29 as above .
then there is a @xmath2-equivariant filtered @xmath11-bimodule quasi - isomorphism @xmath1576 in particular , this gives the following lagrangian intersection theoretic results .
if the @xmath2-invariant @xmath11-algebra of @xmath28 has a non - vanishing homology , @xmath28 and @xmath1577 have a non - trivial intersection for any time - dependent hamiltonian isotopy @xmath1578 .
we remark that seidel already observed in @xmath3-action of exact case that such a lagrangian piunikhin - salamon - schwarz morphism exists and becomes an isomorphism , as well as hamiltonian invariance of equivariant lagrangian floer homology .
( see the last paragraph of section ( 14b ) of @xcite ) .
in this section , we explain how to add @xmath2-equivariant flat vector bundles to the theory . in homological mirror symmetry
, one needs to consider rank 1 flat unitary bundles on a lagrangian submanifold . with a finite group action on the lagrangian submanifold , it is natural to introduce @xmath2-equivariant structures on such line bundles .
we also observe that we can define a natural action of a character group @xmath1579 ( which is the dual group @xmath1580 for abelian @xmath2 ) on both equivariant and orbifolded fukaya categories by twisting @xmath2-equivariant structures on flat vector bundles .
such character group action comes into play in @xmath2-equivariant homological mirror symmetry , which will be studied in the next section .
we remark that such character group action will be also discussed in our joint work with siu - cheong lau in preparation .
we will assume that @xmath1581 for simplicity from now on .
we first review the standard way of including these flat bundles into lagrangian floer theory ( without group actions ) . for a unitary bundle @xmath1582 on a lagrangian @xmath17 ( @xmath18 ) ,
the new floer complex @xmath1583 is defined by replacing each generator @xmath550 $ ] of @xmath1584 with @xmath550 \otimes { { \rm hom}}(u_0|_p , u_1|_p)$ ] .
the floer differential @xmath229 in is modified by additional contributions from holonomies .
namely , given @xmath1585 and a @xmath32-holomorphic strip @xmath1204 which maps @xmath550 $ ] to @xmath1586 $ ] , we get @xmath1587 by composing @xmath1588 with holonomies around two boundary components of @xmath1204 appropriately . consider a trivial complex vector bundle @xmath1589 of rank @xmath727 where @xmath2 acts on @xmath5
a @xmath2-equivariant structure on this bundle means the choice of a @xmath2-action on @xmath1590 , or equivalently the choice of a representation @xmath1591 .
such a homomorphism induces the diagonal @xmath2-action on @xmath1592 .
if @xmath1593 is a character of @xmath2 , then one can twist @xmath1594 by @xmath439 and obtain @xmath1595 given by @xmath1596 . in general ,
let @xmath1597 be a possibly non - trivial vector bundle on @xmath5 .
a @xmath2-equivariant structure on @xmath1598 is given by the choice of an isomorphism @xmath1599 for each @xmath35 , which satisfies the cocycle conditions .
the _ twisting of a @xmath2-equivariant structure _
@xmath1600 by a character @xmath1593 is a @xmath2-equivariant structure @xmath1601 on @xmath1598 defined by @xmath1602 here , @xmath1603 is defined as a fiberwise complex multiplication on a complex vector bundle .
this defines a character group action on the set of @xmath2-equivariant structures on a vector bundle on @xmath5 .
note that for an abelian group @xmath2 , the set of @xmath2-equivariant structures on a complex trivial line bundle are in one - to - one correspondence with the dual group @xmath1604 .
let @xmath28 and @xmath54 be two @xmath2-invariant lagrangian submanifolds and choose @xmath2-equivariant flat complex vector bundles @xmath1605 and @xmath1606 on @xmath28 and @xmath54 , respectively .
here we allow @xmath1582 to have rank greater than @xmath83 since an irreducible representation of @xmath2 may have a dimension greater than @xmath83 .
in particular , we expect that bundles of higher ranks might play a non - trivial role for non - abelian @xmath2 .
we define the floer cochain complex for the pair @xmath1607 and @xmath1608 similarly as above , but using orientation spaces . @xmath1609 } |\theta_{[w , l_p]}^-|_{\mathbb{c}}\otimes_{\mathbb{c}}{{\rm hom}}(u_0|_p , u_1|_p)\ ] ] the floer differential for is defined in a standard way as explained above .
[ gactiongbundleori ] a @xmath2-action on is defined as follows : for @xmath35 , a linear map @xmath1610 is given by @xmath1611 .
the @xmath2-action on the first factor of is exactly the same one as in section [ gactfloernov ] .
now , we consider @xmath2-invariant connections on a @xmath2-equivariant complex vector bundle @xmath1612 on @xmath5 : [ gcompconn ] a connection @xmath1613 of a @xmath2-equivariant complex vector bundle @xmath1612 on @xmath5 is said to be @xmath2-invariant , if the pull - back connection @xmath1614 is isomorphic to @xmath1613 via @xmath1615 .
more precisely , @xmath1616 where @xmath1617 is the horizontal distribution associated to @xmath1613 .
we will only consider @xmath2-invariant connections from now on .
now , consider the pair @xmath1618 together with the choice of a @xmath2-invariant connection @xmath1619 on @xmath1582 for @xmath18 .
the new floer differential involving equivariant flat bundle data is @xmath2-equivariant .
let @xmath1204 be a holomorphic strip bounding @xmath28 and @xmath54 and let @xmath188 and @xmath195 be its end points .
we denote by @xmath1620 the boundary component of @xmath1204 lying in @xmath17 .
( see figure [ equivbdiff ] . )
take an element @xmath1369 of @xmath1621 and @xmath37 of @xmath2 . 1 .
we first compute @xmath1622 : @xmath1623 sends @xmath1369 to @xmath1624 , and then , the strip @xmath1625 sends @xmath1626 to @xmath1627 where @xmath1628 means the parallel transport along @xmath282 by @xmath1619 for @xmath18 .
2 . secondly , let us compute @xmath1629 : @xmath1204 contributes to @xmath229 by sending @xmath1369 to @xmath1630 .
now , applying the @xmath37-action , we get @xmath1631 we have to show that and represent the same element in @xmath1632 .
this follows from the @xmath2-equivariance of @xmath1619 , which implies that the @xmath2-action on @xmath1582 commutes with the parallel transport for @xmath1619 .
i.e. @xmath1633 let @xmath537 be the action of @xmath1634 on equivariant vector bundles defined by the fiberwise complex multiplication .
then , @xmath537-action preserves connections in the sense that @xmath1635 note that the twisting by the character group is essentially given by the complex multiplication .
therefore , @xmath2-invariant connections still remain @xmath2-invariant after twisting equivariant structures on a bundle . for an abelian @xmath2
, one may use a homomorphism @xmath1636 ) \to u(1)$ ] to classify equivariant flat line bundles on @xmath5 and check the above discussion more rigorously .
see the related explanation after conjecture [ conj : fiberpres ] .
we summarize the construction of equivariant fukaya categories in this section , and also clarify once again the role of the group action and its dual group action .
we are considering the case that a finite group @xmath2 effectively acts on a closed oriented symplectic manifold @xmath41 . for each spin
profile @xmath1637 , the equivariant fukaya category @xmath1638 is defined as follows .
an object of @xmath1638 is given by a @xmath2-invariant lagrangian submanifold @xmath5 whose spin profile equals @xmath1 , together with a @xmath2-equivariant flat unitary bundle @xmath1598 . for each pair
@xmath1639 , we choose a reference path @xmath24 ( one in each homotopy class ) and we need to assume that energy zero subgroup @xmath22 equals @xmath2 always . then , we define the morphism to be the ( completed ) direct sum over @xmath24 of the floer complex of the pair @xmath1640 with respect to @xmath24
. there should be @xmath2-equivariant @xmath11-category operations by extending the construction in this paper to the one of fukaya @xcite .
note that we possibly have several equivariant fukaya categories corresponding to spin profiles in @xmath1641 , and each @xmath2-invariant ( spin ) lagrangian submanifold @xmath5 can belong to only one of them corresponding to its spin profile
. we may also consider @xmath302-twisted @xmath2-equivariant vector bundles for @xmath1642 .
these bundles are not @xmath2-equivariant , but their failures of @xmath2-cocycle conditions are given by @xmath302 .
we remark that there is a corresponding notion for sheaves , too ( see for example @xcite ) .
in fact , from the expression , effects of @xmath1642 and a spin profile condition in @xmath6 on lagrangian floer theory can be combined .
hence , it seems that we can enlarge the equivariant or the orbifolded fukaya category by including such objects .
we leave it for future investigation . as discussed , @xmath2 acts on morphisms of the equivariant fukaya category @xmath2-@xmath1643 . by taking @xmath2-invariant part of morphisms of @xmath2-@xmath1643
, we obtain the orbifolded fukaya category @xmath1644 , which still has an induced @xmath11-category structure . in fact , for the definition of orbifolded fukaya category , we do not need assumptions that @xmath57 .
now the character group @xmath1579 acts on objects of @xmath2-@xmath1643 ( hence also on morphisms ) , by twisting equivariant structures of @xmath2-bundles on lagrangians .
therefore , @xmath1579 acts both on the @xmath2-equivariant category and on the orbifolded fukaya category .
the original fukaya category of @xmath41 does not contain the data of @xmath2-equivariant structure of unitary bundles .
hence , one needs to take the @xmath1579-invariant part of equivariant fukaya category @xmath2-@xmath1643 to obtain information about the fukaya category of @xmath41 itself .
there is a well - known fukaya category associated to a lefschetz fibration @xmath1645 , called the fukaya - seidel category of @xmath47 @xcite , @xcite .
if the lefschetz fibration @xmath47 is @xmath2-invariant , i.e. if there is a @xmath2-action on @xmath1452 such that the fibration map @xmath47 is @xmath2-invariant , then we can define the @xmath2-equivariant fukaya - seidel category of @xmath47 and the orbifolded fukaya - seidel category of @xmath1646 \to s$ ] .
these will depend on the choice of a group cohomology class @xmath1 in @xmath6 .
the construction is more or less an easy modification of seidel s construction .
first , we review basic ingredients of fukaya - seidel categories .
( see @xcite for details . )
let @xmath1647 be an exact morse fibration where @xmath1452 admits an exact symplectic form @xmath1199 and @xmath1648 is isomorphic to @xmath1237 .
we assume that @xmath1649 in order to discuss @xmath220-grading , and assume @xmath1650 to simplify the exposition .
suppose there exists a @xmath1199-preserving @xmath2-action on @xmath1452 and @xmath47 is invariant under @xmath2 .
then , the symplectic connection defined by @xmath1199 is @xmath2-invariant , and hence a parallel transport @xmath1651 along a curve @xmath1652 \to s$ ] is a @xmath2-equivariant diffeomorphism .
let @xmath1653 be a regular value of @xmath47 and @xmath41 the fiber of @xmath47 over @xmath1653 .
if @xmath1654 is the set of critical values of @xmath47 , we choose disjoint paths @xmath1655 ( except at @xmath1653 ) from @xmath1653 to @xmath1656 for each @xmath1657 .
@xmath1658 are arranged by their cyclic order at @xmath1653 .
see figure [ distbasis ] where we set @xmath1659
. there may be several different critical points over single critical value @xmath1656 .
denote by @xmath1660 the number of critical points over @xmath1656 .
then , the parallel transport along @xmath1655 defines a set of vanishing cycles @xmath1661 . since @xmath2 acts on these @xmath1660 critical points
, @xmath2 also acts on the set of vanishing cycles @xmath1661 .
we do not need to give an ordering to the family of vanishing cycles for @xmath1655 , because they are disjoint from each other , and hence floer cohomology groups and dehn twists between them are all trivial . as usual
, we also pay attention to the lefschetz thimbles which will be denoted by @xmath1662 s .
more precisely , @xmath1662 is the union of images of the vanishing cycle @xmath1663 under the parallel transport over @xmath1655 so that @xmath1664 . since the symplectic connection as well as @xmath47 are @xmath2-invariant , the group @xmath2 acts on the union of lefschetz thimbles over each @xmath1655 , and we immediately obtain the following lemma . for each @xmath1665 and @xmath35 , we have @xmath1666 for some @xmath1667 .
namely , an action of each @xmath35 permutes thimbles over the given path @xmath1655 ( while some of thimbles can be preserved by @xmath37 ) .
gradings and spin structures on vanishing cycles are defined as follows . as
explained right after definition [ def : seqvbrane ] , we can equip @xmath1663 s with @xmath2-invariant gradings .
if @xmath1668 , then the vanishing cycles are spheres of @xmath1669 , which have canonical spin structures . in case @xmath1670 ,
vanishing cycles are circles , and we choose a spin structures of each vanishing cycle to be the one obtained from the restriction of the trivialization of the tangent bundle of the corresponding lefschetz thimble .
we remark that an object consists of a single connected component if @xmath2-action preserves @xmath1663 , and has several disjoint components otherwise .
for example , if there is only one critical point over the critical value @xmath1656 , then the vanishing cycle @xmath1671 is preserved by the @xmath2-action and is an object of the category as this critical point should be fixed by the whole group @xmath2 .
now , we define an equivariant ( or an orbifold ) version of a directed @xmath11-category @xmath1672 , so called the fukaya - seidel category . for each @xmath1 in @xmath1166
, we will have a category consisting of @xmath1-equivariant branes , which are @xmath2-invariant ( union of ) vanishing cycles with the spin profile @xmath1 .
[ def : fs ] consider a @xmath2-invariant lefschetz fibration @xmath46 , and choice of paths @xmath1673 and vanishing cycles @xmath1674 as before .
an object of both the @xmath1-equivariant fukaya - seidel category @xmath2-@xmath1675 and orbifold fukaya - seidel category @xmath1676 is given by a @xmath2-orbit @xmath1677 for each vanishing cycle @xmath1663 for some @xmath1678 whose spin profile is @xmath1 , together with a @xmath2-invariant grading and @xmath2-equivariant flat complex vector bundle on @xmath1677 .
we require that this bundle extends to a @xmath2-equivariant flat complex vector bundle on the corresponding thimble .
we denote by @xmath1679 the @xmath2-orbit of vanishing cycle @xmath1677 together with these additional data .
also , if @xmath1663 and @xmath1680 lie in the same @xmath2-orbit , then we identify @xmath1677 and @xmath1681 .
the objects of equivariant and orbifold categories are the same as above , but their morphisms will be defined differently .
we postpone the discussion of adding @xmath2-equivariant flat complex vector bundle to the construction in subsection [ equivbundlefs ] , and we first explain the construction without them .
we may suppose that all vanishing cycles ( and hence , all objects in @xmath985 ) are transversal to each other by choosing vanishing paths in general positions ( figure [ distbasis ] ) .
the cyclic order at @xmath1653 gives a partial order , on the index @xmath1665 of @xmath1663 .
[ def : fs2 ] we define morphisms between two objects @xmath1682 and @xmath1683 of the @xmath1-equivariant category @xmath2-@xmath1675 as follows . for @xmath1684 , @xmath1685 for @xmath1686 , and two objects @xmath1682 and @xmath1683
are in fact the same , then we define @xmath1687 for the remaining case , their morphisms are defined to be zero ( since they do not intersect ) .
this @xmath11-category is not strictly but partially directed in a sense . note
that the morphism space admits a @xmath2-action . for @xmath1688 ,
the @xmath2-action on their morphism is exactly the same as in section [ sec : eqfuex ] .
if @xmath1686 and two objects @xmath1682 and @xmath1683 are the same , then the @xmath2-action on @xmath1689 is defined as follows .
we first fix orientations of @xmath1690 , and the construction below depends on this choice .
since @xmath1691 is a @xmath2-orbit , each connected component @xmath1692 is mapped to @xmath1693 for some @xmath674 .
in such a case , we define @xmath37-action @xmath1694 to be @xmath1695 , where the sign is positive if the @xmath37-action preserves the pre - fixed orientations , and negative otherwise .
this defines @xmath2-action on morphism spaces . in order to construct @xmath2-equivariant @xmath11-operations on morphisms of @xmath2-@xmath1675
, we may adapt the construction in the previous section .
namely , we consider a @xmath982-category @xmath981 , whose object is @xmath1696 for each @xmath1697 and @xmath1698 .
then the set of object of @xmath981 admits a free @xmath2-action , and we can make the perturbation data @xmath2-equivariant . as seidel mentioned in ( * ? ? ?
* remark 6.1 ) , we do not have to care about the chain complexes underlying @xmath1699 by the ( partial ) directedness , but only the above case ( @xmath1700 ) is needed to be considered .
therefore , the homological perturbation lemma of seidel ( lemma [ hompert ] ) provides @xmath2-equivariant @xmath982-operations on morphisms of @xmath2-@xmath1675 .
for the orbifold fukaya - seidel category , morphisms of @xmath1676 are defined to be the @xmath2-invariant part of the corresponding morphisms of @xmath2-@xmath1675 , which has an induced @xmath982-operations . even though vanishing cycles @xmath1691 give rise to objects of directed @xmath11-category , one has to think in terms of respective vanishing thimbles denoted as @xmath1702 , which are indeed lagrangian submanifolds in @xmath41 .
we assume here that @xmath2 is abelian in order to simplify the exposition .
general cases can be handled as in section [ sec : orbibundle ] .
note that any irreducible representation of @xmath2 is 1-dimensional and hence , it suffices to consider line bundles only .
consider @xmath1703 for each @xmath1697 and the @xmath2-equivariant line bundle @xmath1598 on @xmath1691 which extends to the one on the corresponding thimble @xmath1702 .
let us denote it by @xmath1704 , which is necessarily trivial .
thus , @xmath1705 and the @xmath2-equivariant structure on @xmath1706 comes up with @xmath1707 .
suppose that @xmath1708 is a vanishing cycles with a @xmath2-equivariant line bundle @xmath1709 as above for @xmath1710 .
then , pairs @xmath1711 for @xmath1710 are objects of @xmath2-@xmath1675 and we want to define their morphism space . if @xmath1712 we define their morphism space to be as in section [ sec : fs ] using @xmath950 for each intersection points @xmath144 but with @xmath82 coefficients .
the group action on the morphism space is twisted by @xmath1713 and @xmath1714 .
i.e. @xmath1715 is defined as the original action multiplied by @xmath1716 .
this is a canonical action on @xmath1717 consistent with .
however , since @xmath1718 is canonically isomorphic to @xmath82 , we may just use @xmath950 as above .
let us consider the case @xmath1719 .
denote the connected components of @xmath1691 as @xmath1720 associated to critical points @xmath1721 . recall that @xmath1722 ( with @xmath1723 ) is given by @xmath1724
the @xmath2-action on the above morphism space has been already described in section 7 .
when they are equipped with @xmath2-equivariant line bundles , @xmath1725 is defined to be the same as , whose @xmath37-action is given by the previous @xmath37-action defined in section [ sec : fs ] with additional multiplication of @xmath1716 .
note that the @xmath2-invariant part @xmath1726 is non - trivial if @xmath2-action on the set of connected components of @xmath1691 is free .
we discuss a few examples of well - known homological mirror symmetry , but with an additional finite group action .
the homological mirror symmetry ( conjectured by kontsevich ) asserts that the derived fukaya category of a symplectic manifold ( or the derived fukaya - seidel category of a lg model ) is equivalent to the derived category of coherent sheaves of the mirror complex manifold ( or the matrix factorization category of the mirror lg model ) .
strominger - yau - zaslow approach explains such a phenomenon as a correspondence between dual torus fibrations
. we will be rather sketchy , as our motivation is to consider its relationship with group actions .
consider a lagrangian torus fibration @xmath1727 where @xmath1728 is regular torus fiber for @xmath1729 , and assume that @xmath1730 is simply connected .
suppose that the mirror is given as a landau - ginzburg ( lg ) model @xmath1731 . here , @xmath1732 is the dual torus fibration @xmath1733 where @xmath1734 is considered to be the space of flat connections on the trivial line bundle on @xmath1735 .
suppose that the mirror potential @xmath75 is defined as a lagrangian floer potential ( see for example @xcite , @xcite , @xcite ) @xmath1736 we say that an action of a finite group @xmath2 on @xmath72 is compatible with the fibration @xmath47 if the group action sends torus fibers to torus fibers .
we will assume that @xmath2 acts on @xmath1737 in a compatible way from now on .
thus each @xmath35 sends a fiber @xmath1735 into @xmath1738 for some @xmath1739 for each @xmath1740 .
we define the @xmath2-action on @xmath73 as @xmath1741 ( here , @xmath1742 . )
note that this @xmath2-action could be trivial even if the @xmath2-action on @xmath72 is effective as we will see in the example of @xmath1743 . for a @xmath2-action on @xmath72 compatible with @xmath47
, we have an induced homomorphism @xmath1744 .
this gives rise to the following exact sequence @xmath1745 we may denote @xmath1746 and @xmath1747 .
the example of @xmath1743 below is the case when @xmath1748 .
but if @xmath1749 , it is clear that @xmath2-action on @xmath73 is also effective if that on @xmath72 is .
we will see such an example in the section [ sec : cp2 ] , where the @xmath2-action on @xmath1750 rotates the moment triangle @xmath71 .
it is not hard to show that @xmath76 becomes @xmath2-invariant for the induced @xmath2-action on @xmath73 .
[ chitriv ] @xmath75 is @xmath2-invariant .
it follows from the invariance of the symplectic form @xmath1199 under the @xmath2-action : @xmath1751 we denote by @xmath1 the spin profile of torus fibers @xmath1735 for @xmath1729 .
we can formulate the homological mirror symmetry conjecture in this setting as follows : there are equivalences of derived categories , @xmath1752 where @xmath1753 is the derived @xmath1-orbifolded fukaya category of @xmath72 defined in this paper , @xmath1754 is the derived category of @xmath2-equivariant coherent sheaves on @xmath72 and @xmath1755 is the @xmath2-equivariant matrix factorization category of @xmath76 .
the character group @xmath1579 acts on both sides of by twisting equivariant structures in a compatible way . in general , for a mirror pair @xmath1756
, one may conjecture the equivalence of two categories @xmath1753 and @xmath1757 . here , note that the right hand sides do not involve spin profiles .
this is because sheaves on the right hand side are expected to be obtained as certain family version of floer homologies , and hence @xmath2-action is well - define for them ( without spin profiles ) .
note also that the right hand side is well - defined even when @xmath2 acts trivially .
suppose that an abelian group @xmath1759 acts freely on regular torus fibers .
in particular , the character group @xmath1760 is the dual group of @xmath2 .
its dual torus fibration can be understood as follows .
consider @xmath1761 ) \to g \to 1\ ] ] and take @xmath1762 to obtain the following exact sequence : @xmath1763 ) , u(1 ) ) \to y_b={{\rm hom}}(\pi_1 ( x_b ) , u(1)).\ ] ] if @xmath1764 is still a ( lagrangian ) torus , then the last map @xmath1765 is surjective .
this follows from the fact that the first and the second term in the exact sequence are isomorphic to @xmath1766 and @xmath1767 is divisible .
let us assume that @xmath1768 is again a lagrangian torus fibration from now on .
this happens for example when @xmath72 is a toric manifold with the action of a finite subgroup @xmath1759 of the torus @xmath1769 .
since @xmath1770 ) = \pi_1(x_b / g)$ ] , the middle term @xmath1771 gives the dual torus fiber of @xmath1772 , giving rise to the dual torus fibration @xmath1773 . from , it is easy to see that the dual group @xmath1760 acts on @xmath1774 and the quotient is isomorphic to the mirror manifold @xmath73 of @xmath72 .
let @xmath1775 be the quotient map via @xmath1760 .
indeed @xmath1760 acts on @xmath1774 freely and @xmath1776 is a covering whose deck transformation group is precisely @xmath1760 since the first map in is injective .
let @xmath1777 be given by the composition @xmath1778 .
@xmath1779 is @xmath1760-invariant by definition .
we define @xmath1779 by the composition because smooth holomorphic discs in @xmath1780 $ ] are in fact obtained from smooth holomorphic discs in @xmath72 .
( see @xcite for more details : we do not consider bulk - deformations by twisted sectors , here . )
the situation is summarized in the diagram below : @xmath1781_{g } \ar@{<.>}[rr]^{mirror } & & y \ar[r]^w & { \mathbb{c}}\\ [ x / g ] \ar@{<.>}[rr]_{\quad mirror } & & { \widetilde{y}}\ar[u]^{g^\ast } \ar[ur]_{{\widetilde{w}}}&. } \ ] ] namely , in this case , taking @xmath2-quotient on one side corresponds to taking a @xmath1760-covering on the other side .
we remark that such a phenomenon has been already observed for the pair of pants case by abouzaid , auroux , efimov , katzarkov and orlov @xcite .
now , it is natural to expect that @xmath1782 is the lg mirror of @xmath1780 $ ] in the sense of syz .
moreover , the @xmath384-model category for @xmath72 and the one for its quotient are expected to have a close relationship .
[ conj : fiberpres ] in the above setting , we have equivalences of derived categories @xmath1783 and @xmath1784 here , @xmath1785 is the @xmath1760-equivariant matrix factorization category . as the @xmath1760-action on @xmath1774 is free , @xmath1785 is equivalent to @xmath1786 where the equivalence from @xmath1786 to @xmath1785 is the pull - back functor by the projection @xmath1787 .
( this is a special case of proposition 2.2 in @xcite . )
hence the second part of the conjecture follows from the first part of the conjecture .
we can relate the @xmath1760-action on @xmath1774 with the @xmath1760-action on @xmath1788 in a geometric way . in view of syz ,
a point in @xmath1771 corresponds to a torus fiber @xmath1789 $ ] together with a flat line bundle over it , or equivalently a @xmath2-equivariant flat line bundle @xmath1790 over @xmath1735 .
since @xmath2 acts freely on @xmath1735 , a @xmath2-equivariant structure on the line bundle @xmath1598 is unique up to isomorphism .
however , if we apply the character group action of @xmath1791 to @xmath1792 , the flat structure , or holonomy in an orbifold sense , of @xmath1793 $ ] changes as follows . consider a generalized curve @xmath1794)$ ] with @xmath1795 to @xmath1796 where @xmath100 is the base point for @xmath1797 .
after projecting down to @xmath1798 $ ] , @xmath282 becomes a genuine loop @xmath1799 .
then , the holonomy along the loop @xmath1799 in @xmath1789 $ ] can be measured by the difference between the parallel transport along @xmath282 and @xmath1615 , both of which are linear maps from @xmath1800 and @xmath1801 if @xmath1802 is twisted by an element @xmath1791 , then the identification @xmath1615 of @xmath1800 and @xmath1801 is also twisted by @xmath1803 .
i.e. @xmath1800 and @xmath1801 are now identified by @xmath1804 . as a result ,
the holonomy around @xmath1799 in @xmath1789 $ ] is changed by @xmath1803 .
this explains the action of @xmath1760 on @xmath1771 shown in .
we provide an example which illustrate the phenomenon discussed in the previous subsection .
equip @xmath1805 with the @xmath1806-action given by @xmath1807 \to [ \rho z_0 : z_1]$ ] where @xmath1808 .
it preserves all torus fibers , and hence @xmath1809 in this example .
associated to this action , we can construct a three - fold cover @xmath1774 of the mirror @xmath1810 of @xmath72 . identifying @xmath1774 with @xmath1634 , the covering map
is explicitly given by @xmath1811 the @xmath1812-action on @xmath1774 sends @xmath465 to @xmath1813 .
the variable @xmath465 indeed comes from orbidiscs bounding torus fibers in @xmath1780 $ ] . in figure
[ cp1modz3 ] , @xmath1814 .
see @xcite or @xcite for more details .
then , in terms of a coordinate @xmath465 on @xmath1774 , the pulled - back potential @xmath1779 can be written as @xmath1815 ( for simplicity , we set the khler parameter @xmath195 to be @xmath83 . ) note that @xmath1779 is invariant under the @xmath1760-action .
consider the lagrangian brane @xmath1816 supported on the balanced torus fiber with the trivial holonomy in @xmath1780 $ ] .
in @xcite , it is shown that the syz transformation of @xmath1816 is the following matrix factorization of @xmath1779 .
@xmath1817 notice that @xmath773 is a critical value of @xmath1779 .
now , @xmath1818 sends @xmath1816 to the lagrangian brane supported on the same torus fiber but with the holonomy @xmath623 . in the mirror
, it sends to @xmath1819 where we used @xmath1820 .
there is aa easy example of a @xmath1760-invariant object in @xmath1821 , which can be written down only in terms of @xmath1822 : @xmath1823 using @xmath1824 , is expressed as @xmath1825 which was shown to be the mirror matrix factorization of the equator in @xmath1743 in @xcite .
in this section , we consider the projective plane @xmath1750 equipped with the @xmath1826-action which cyclically permutes three homogeneous coordinate of @xmath1827 $ ] in @xmath1750 . note that the torus fibration @xmath1828 is compatible with this @xmath1829-action .
the landau - ginzburg mirror of @xmath1750 is given by @xmath1830 with the superpotential @xmath1831 .
it is easily check that the mirror @xmath1806-action also permutes three variable @xmath170 , @xmath144 , and @xmath1832 , and hence @xmath76 is invariant under this action . in what follows ,
we identify the mirror of @xmath1750 with @xmath1833 and use @xmath1834 as the superpotential .
let us first remark on spin profiles and energy zero subgroups . for @xmath1839 , we will consider the floer homology of a torus fiber with itself , and in such a case a choice of spin profile does not affect the group action .
indeed , spin profiles for invariant lagrangians involved in this example are all zero since @xmath1840 .
also the central fiber has a fixed point of group action .
thus from lemma [ locgpingfl ] , we have @xmath1841 , and we can define @xmath2-actions on lagrangian floer homology groups . [ gequivsh1 ] a @xmath2-equivariant structure on a sheaf @xmath1842 on @xmath72 is a choice of an isomorphism @xmath1843 for each @xmath37 such that these isomorphisms satisfy the cocycle condition @xmath1844 since @xmath1845 for @xmath1846 , the @xmath2-equivariant structure on @xmath1842 gives an isomorphism @xmath1847 for each @xmath35 which covers the base action .
@xmath1848^{\theta_g ( x ) } \ar@{.>}[d ] & & \mathcal{f}_{g \cdot x } \ar@{.>}[d ] \\ x \ar@{|->}[rr]^g & & g \cdot x } \qquad \xymatrix { \mathcal{f}_x\ar[rr]^{\theta_{gh } ( x ) } \ar[dr]_{\theta_h ( x ) } & & \mathcal{f}_{(gh ) \cdot x}\\ & \mathcal{f}_{h \cdot x } \ar[ur]_{\theta_g ( h \cdot x ) } & } \ ] ] on the level of stalks , the cocycle condition can be understood more clearly shown in the right side diagram in .
we write @xmath1851 for the derived category of .
the original definition of @xmath1852 is slightly different from the one given here , but it is known that they are equivalent if @xmath2 is finite .
( see @xcite . )
note that three sheaves in proposition [ beilinsoncot ] are all @xmath1806-invariant .
we fix an equivariant structure on each of these sheaves so that the induced action on morphism spaces and are given by cyclic permutation on @xmath1861 .
let @xmath1692 ( @xmath1862 ) be the irreducible representation of @xmath1806 on which the primitive generator of @xmath1806 acts by @xmath1863 .
then , any @xmath1806-equivariant structure on @xmath1864 is isomorphic to @xmath1865 for some @xmath674 .
moreover , for @xmath1866 with @xmath1867 , @xmath1868 ( see the proof of ( * ? ? ?
* theorem 2.1 ) . )
if @xmath1869 is strictly greater than @xmath1870 , then @xmath1871 and @xmath2 permutes its basis ( which are chosen in and ) .
therefore , the rank of the right hand side is @xmath83 regardless of @xmath1872 and @xmath1873 .
however , if @xmath1874 , then @xmath1875 * ( ii ) * the mirror @xmath1806-action fixes three critical points of @xmath76 , @xmath1876 where @xmath1877 .
they are indeed the only fixed points of this @xmath1806-action .
we have vanishing cycles @xmath1878 , @xmath1879 and @xmath1880 in @xmath1881 associated to @xmath567 , @xmath568 , and @xmath1882 respectively .
each @xmath1692 is a @xmath1806-invariant lagrangian submanifold of @xmath1881 . since the @xmath1806-action on @xmath1881 is free , the action on @xmath1692 is topologically equivalent to the one generated by the @xmath1883-rotation on the circle(@xmath1884 ) and in particular , it is orientation preserving .
we equip @xmath1692 s with nontrivial spin structures ( which extend to the vanishing thimbles ) . as mentioned
, we do not need to care about spin profiles as @xmath1885 is zero and therefore , the group action lifts to the spin bundles . by projecting to the @xmath170-coordinate plane ( minus the origin ) ,
we see that each pair of vanishing cycles intersects at three points .
we denote them as @xmath1886 @xmath1887 @xmath1888 since the group action is free on @xmath1881 , it also freely permutes three points in @xmath1889 for @xmath1890 . from @xcite ,
now , we consider the equivariant vector bundles on @xmath1692 s .
recall from section [ sec : orbibundle ] that such bundles extend to the vanishing thimbles ( @xmath1891 s ) so that the equivariant structures are determined by characters in @xmath1892 .
let @xmath1893 ( @xmath1862 ) be the character that sends the primitive generator of @xmath1806 to @xmath1894 ( @xmath1895 ) .
then , the @xmath1806-equivariant fukaya - seidel category has @xmath1896 objects : @xmath1897 , @xmath1898 .
one can easily check that if @xmath1684 , the @xmath2-invariant part of @xmath1899 is of rank @xmath83 ( regardless of @xmath1872 and @xmath1873 ) since the @xmath2-action is free . if @xmath1900 but @xmath1901 , then the @xmath1806-action on @xmath1902 is nontrivial as @xmath1903 for some @xmath1904 .
therefore , there can not exist any @xmath37-invariant element other than zero .
hence we have @xmath1905 * ( i ) * in @xmath1750 , there are three nontrivial @xmath384-branes , the clifford torus @xmath1906 , together with three different holonomies , @xmath1907 , @xmath1908 and @xmath1909 , which are known to have non - vanishing floer homologies @xcite .
in fact they are expected to generate the fukaya category of @xmath1750 which has been announced by abouzaid , fukaya , oh , ohta and ono @xcite .
the central fiber @xmath1906 is preserved by the @xmath1806-action .
as the values of the ( lagrangian floer ) potential function are distinct for these three @xmath384-branes , floer homologies between them are not defined , but only floer homologies for the same pairs of objects are defined , which turn out to be isomorphic to the singular cohomology of @xmath1906 ( as modules ) .
we work with @xmath3-grading , and hence both the even and the odd degree components of their endomorphism spaces have rank @xmath773 .
let us consider @xmath1910 only , since the other cases are almost the same .
we identify @xmath1911 with the space of harmonic forms on @xmath1906 , which is generated by @xmath1912 , @xmath1913 , @xmath1914 ( and their products as well ) with the relation @xmath1915 .
the induced action on @xmath1906 sends @xmath1916 to @xmath1917 ( with @xmath1918 ) .
one can easily check that @xmath83 and @xmath1919 is invariant under this action .
however , the @xmath2-invariant part of @xmath1911 does not have any 1-forms since the only 1-form possibly @xmath2-invariant is @xmath1920 . *
( ii ) * likewise , among three critical value @xmath1921 , @xmath1922 , and @xmath1923 of the potential @xmath1924 , we only consider the category @xmath1925 .
@xmath1926 has a unique singular point @xmath1927 , which is nodal .
let @xmath1842 be the skyscraper sheaf supported at this point .
note that @xmath1842 is @xmath1806-invariant .
we compute @xmath1928 in the split - closure of @xmath1925 .
after the idempotent completion ( taking the split - closure ) , it is enough to consider the skyscraper sheaf @xmath1929 at the origin in @xmath1930 where @xmath1931 / xy$ ] .
recall from the previous subsection that the action on nearby regular fibers ( locally identified with @xmath1932 ) are induced by @xmath1933-rotations on zero sections .
thus , the continuity of the action shows that the corresponding group action on @xmath1930 ( near the origin ) is analytically equivalent to the one given by @xmath1934 .
( see figure [ egdbsing ] . )
note that we have an exact triangle @xmath1935\ ] ] consisting of @xmath1806-invariant sheaves and that @xmath384 is a free module .
therefore , @xmath1936 is isomorphic to @xmath1937 in the derived category of singularities .
the even degree part of @xmath1938 is , then , given by @xmath1939 ( here , @xmath1940 means the space of @xmath384-module homomorphisms from @xmath1937 to itself which do not factor through a free @xmath384-module . ) for @xmath1941)$ ] , we first compute @xmath1942 $ ] and @xmath1943 $ ] . from the following exact triangle @xmath1944\ ] ] we see that @xmath1945 \cong a / ax = { \mathbb{c}}[y](=\mathcal{o } ( y-{\rm axis } ) ) $ ] .
similarly , @xmath1943 \cong { \mathbb{c}}[x]$ ] .
therefore , @xmath1946 \oplus { \mathbb{c}}[y]).\ ] ] any morphism in the right hand side should send @xmath170 to a constant multiple of @xmath1947 $ ] , or it would factor through @xmath384 .
likewise , the image of @xmath144 has to be a multiple of @xmath1948 $ ] .
therefore , @xmath1949 .
note that all sheaves involved here admit natural @xmath1806-actions since they are defined by geometric terms preserved by @xmath1806 ( e.g. @xmath1950 is the ideal sheaf consisting of regular functions on @xmath1951 which vanish on @xmath144-axis ) .
these actions , of course , can be twisted further by tensoring irreducible representations of @xmath1806 , but we do not consider them here for simplicity .
all elements in are clearly invariant under the induced action on so that @xmath1952 however , @xmath1953 since the primitive generator of @xmath1806 acts on @xmath1954 and @xmath1955 by multiplying @xmath623 , while it keeps both @xmath1947 $ ] and @xmath1956 $ ] invariant .
m. abouzaid , k , fukaya , y .- g .
oh , h. ohta , and k. ono , _ quantum cohomology and split generation in lagrangian floer theory , _ in preparation .
a. adem , j. leida and y. ruan , _ orbifolds and stringy topology , _ cambridge tracts in mathematics , 171 .
cambridge university press , cambridge , 2007 .
d. auroux , l. katzarkov and d. orlov , _ mirror symmetry for weighted projective planes and their noncommutative deformations , _ ann . of math . 167
( 2008 ) , no .
3 , 867943 denis auroux , _ mirror symmetry and t - duality in the complement of the anticanonical divisor _ j. gkova geom . topol . 1 , 2007 , p.51 - 91 a. a. beilinson , _ coherent sheaves on @xmath1957 and problems of linear algebra , _ functional analysis and its applications 12.3 ( 1978 ) : 214 - 216 .
j.bernstein , v.lunts , _ equivariant sheaves and functors , _
springer lecture notes 1578 ( 1994 ) k. chan and n. c. leung , _ matrix factorizations from syz transformations , _ advances in geometric analysis , adv .
lect . math .
( alm ) 21 , int . press , somerville , ma , 2011 .
w. chen and y. ruan , _ a new cohomology theory of orbifold , _ comm .
248 ( 2004 ) , 1 31 .
mr 2104605 w. chen and y. ruan , _ orbifold quantum cohomology , _
preprint math.ag/0005198 .
w. chen and y. ruan , _ orbifold gromov - witten theory , _ orbifolds in mathematics and physics .
, 310 , amer .
soc . , providence , ri(2002 ) : 2585 .
holomorphic discs , spin structures and the floer cohomology of the clifford torus , _ imrn ( 2004 ) no .
35 1803 - 1843 c .- h .
cho and h. hong , _ orbifold morse - smale - witten complex , _ preprint , arxiv:1103.5528 c .- h .
cho and h. hong , _ floer cohomology for @xmath2-equivariant immersions , _ in preparation .
cho , h. hong and s. lee _ examples of matrix factorizations from syz , _ sigma 8 ( 2012 ) , 053 c .- h .
cho , h. hong , h.s .
shin _ on embeddings between orbifolds , _ to appear in journal of kms .
cho and y .- g .
oh , _ floer cohomology and disc instantons of lagrangian torus fibers in fano toric manifolds , _ asian j. math .
10 ( 2006 ) , no . 4 , 773 - 814 .
cho , m. poddar _ holomorphic orbidiscs and lagrangian floer cohomology of symplectic toric orbifolds , _ preprint , arxiv:1206.3994 a. d. elagin , _ semiorthogonal decompositions of derived categories of equivariant coherent sheaves , _ izvestiya : mathematics 73.5 ( 2009 ) : 893 .
a. floer , _ morse theory for lagrangian intersections , _ j. differ .
28 ( 1988 ) , 513 - 547 .
k. fukaya , _ morse homotopy , @xmath982-category and floer homologies , _ in proceedings of garc workshop on geometry and topology , ed by h. j. kim , seoul national university , korea 1993 . k. fukaya _
deformation theory , homological algebra , and mirror symmetry , _ geometry and physics of branes ( como , 2001 ) , 121209 , ser .
high energy phys .
, iop , bristol , 2003 k. fukaya _
floer homology and mirror symmetry ii , _ adv . stud . in pure math.34 2002 , 31 - 127 k. fukaya , k. ono , _ arnold conjecture and gromov - witten invariants , _ topology , vol .
38 , no . 5 , 933 - 1048 , 1999 k. fukaya , y .- g .
oh , h. ohta , and k. ono , _ lagrangian intersection floer theory - anomaly and obstruction , _ ams / ip studies in advaneced math . 46 , international press / amer .
k. fukaya , y .- g .
oh , h. ohta , and k. ono , _ lagrangian floer theory on compact toric manifolds i , _ duke math .
j. 151 ( 2010 ) , 23 - 174 .
k. fukaya , y .- g .
oh , h. ohta , and k. ono , _ lagrangian floer theory on compact toric manifolds ii : bulk deformations , _ selecta math , new series , volume 17 , number 3 ( 2011 ) , 609 - 711 k. fukaya , y .- g .
oh , h. ohta , and k. ono , _ technical details on kuranishi structure and virtual fundamental chain , _
preprint arxiv:1209.4410 m. kontsevich , _ homological algebra of mirror symmetry _ , icm-1994 proceedings , zrich , birkhuser ( 1995 ) .
r. c. kirby and l. r. taylor _ pin structures on low - dimensional manifolds , _ lms lec .
note series 151 , camb .
univ . press , ( 1990 ) 177 - 242 e. lerman and s. tolman , _ hamiltonian torus actions on symplectic orbifolds and toric varieties , _ trans .
349 ( 1997 ) 4201 - 4230 j. milnor _
spin structures on manifolds , _
lenseignement mathematique , 9 ( 1963 ) y .- g .
oh , _ floer cohomology of lagrangian intersections and pseudo - holomorphic discs i , _ comm .
pure and appl .
46 ( 1993 ) , 949 - 994 addenda , ibid , 48 ( 1995 ) , 1299 - 1302 .
i. satake , _ on a generalization of the notion of manifold , _ proc .
usa 42 ( 1956 ) , 359 - 363 .
p. seidel _
fukaya categories and picard - lefschetz theory , _ ems zurich lect . adv . math .
2008 p. seidel , _ homological mirror symmetry for the quartic surface , _ preprint .
p. seidel , _ vanishing cycles and mutation , _
european congress of mathematics .
birkhauser basel , 2001 p. seidel , _ more about vanishing cycles and mutations , _ symplectic geometry and mirror symmetry ( seoul , 2000 ) , 429 - 465 , world sci .
publ . , river edge , nj , 2001 y. takeuchi _ waldhausen s classification theorem for finitely uniformizable 3-orbifolds , _ trans .
328 , no . 1 ( 1991 ) , 151 - 200 h. b. lawson , jr and m - l .
michelsohn , _ spin geomtery , _ princeton univ . press .
38 1990 c. a. weibel , _ an introduction to homological algebra , _ cambridge studies in advanced mathematics 38 . | we construct finite group actions on lagrangian floer theory when symplectic manifolds have finite group actions and lagrangian submanifolds have induced group actions . we first define finite group actions on novikov - morse theory . we introduce the notion of a _ spin profile _ as an obstruction class of extending the group action on lagrangian submanifold to the one on its spin structure , which is a group cohomology class in @xmath0 . for a class of lagrangian submanifolds which have the same spin profiles ,
we define a finite group action on their fukaya category . in consequence ,
we obtain the @xmath1-equivariant fukaya category as well as the @xmath1-orbifolded fukaya category for each group cohomology class @xmath1 .
we also develop a version with @xmath2-equivariant bundles on lagrangian submanifolds , and explain how character group of @xmath2 acts on the theory . as an application , we define an orbifolded fukaya - seidel category of a @xmath2-invariant lefschetz fibration , and also discuss homological mirror symmetry conjectures with group actions . |
in this paper the status of heavy flavours , charm , bottom and top , is reviewed .
for the charm quark , emphasis is placed on measurements of the production rates and the differential cross sections at hera and the tevatron and of the extraction of the gluon structure function from the proton at hera using charm .
the large kinematic range at hera leads to a number of models with limits on applicability .
these models can not be directly applied to the tevatron and the work to understand charm production is not complete . for bottom physics , the current status of the lifetime ratios between charged and neutral mesons is examined in the context of the spectator model .
the production cross section at hera and the tevatron is also studied revealing an area that needs further study .
experimental methods for doing a @xmath0 mixing analysis at hadron machines are explored and the results of a variety of @xmath0 mixing studies presented .
this leads to a description of the measurement of the cp violation parameter , @xmath1 , at cdf as an indication of the propects for measurements of this value at hadron machines . a brief summary of the current knowledge and world sample of top quarks closes the report .
figure [ fig : disandwdepend ] illustrates the relevent kinematic variables in vector boson , v , production at hera for v=@xmath2 : _ s _ , the electron - proton center of mass energy , _ q _ , the photon virtuality , _ t _ , the four momentum transfer to proton , @xmath3 , the mass of dissociated proton remnants , _ w _ , the @xmath4-proton cm energy . by choosing various ranges in the kinematic variables , experimental observations can be matched to theoretical predictions for the processes expected to dominate the dynamics in the chosen region .
various descriptions of charm production and the relationship between theoretical description and experimental signature are given in table [ tab : heramess ] where the variable _ z _ , the fraction of @xmath4 momentum given to _
j_/@xmath5 , has been introduced . in each case
a @xmath2 or a @xmath6 is observed .
the first column of the table gives an additional experimental constraint .
the second column is the conventional description of the process with the third column indicating constraints on the kinematics .
@lll experimental & common name & kinematic + constraint & & restrictions + no electron & photoproduction & @xmath7 + an electron & dis & @xmath8 + other activity & inelastic or & @xmath9;@xmath10 + & resolved photoproduction & + no other activity & elastic & @xmath11;@xmath12 + no forward activity & diffractive & rapidity gaps + [ tab : heramess ] central to the description of charm production is the scale of the processes under scrutiny .
one suggested @xcite choice for the scale is @xmath13/4 $ ] .
this implies that for applications of perturbative qcd , the minimum scale is the mass of the @xmath2 .
there is heated debate over whether perturbative qcd computations of one or two gluon ladders can describe the data @xcite or whether modified models of soft pomeron exchange @xcite are more satisfactory .
one example where the perturbative qcd models describe the data is illustrated in figure [ fig : disandwdepend ] @xcite .
the data are also shown to have a @xmath14 dependence such that the cross section goes as @xmath15 with @xmath16 .
the vector dominance model and regge theory predict a smaller slope of @xmath17 and therefore do not provide a good description of the data .
further success in the description of the data can be found in the measurement of the gluon structure function using open charm production and comparing it to the measurement of @xmath18 @xcite .
this is illustrated for the h1 measurement in figure [ fig : gx ] .
the rapid rise of the gluon structure function at low @xmath19 is a prediction of its evolution in the context of the dglap equations @xcite .
however , application of perturbative qcd to the tevatron collider data does not succeed @xcite .
a colour singlet model @xcite , like that used successfully in hera , fails .
the disagreement is as large as a factor of 50 .
better agreement can be found if the cdf data are fit with the colour octet model @xcite ; however the fits of this model fail badly when applied to the differential cross sections measured in a variety of kinematic variables at hera @xcite . while it is tempting to dismiss the scale set by the charm mass as too small and the discrepancies as being no surprise , doing so would ignore the results where there is success in the predictions . assembling the successes and failures and understanding the limits of the use of perturbative qcd and its matching with soft
phenomenology remains an unsolved problem .
the production of @xmath20 quarks in hadron machines differs from the intuitively appealing @xmath21 decay mechanism resulting in two 45 gev jets having @xmath0 mesons carrying 80% of the jet energy . in hadron
machines the trigger biases the kinematics .
one trigger used by cdf is a lepton having high transverse momentum relative to the proton - antiproton collision .
the lepton can be either an electron or muon from a semileptonic decay .
the @xmath0 mesons typically have a transverse momentum of @xmath22 gev / c .
another very important trigger is a dimuon trigger with a cut on the invariant mass of the muons .
this @xmath2 trigger , so called because the invariant mass of the muons is required to be within a broad range around the @xmath2 mass , will yield @xmath0 mesons having a @xmath23 gev / c .
alternatively the @xmath2 mass may be excluded in order to get a cascade of sequential decays : @xmath24 followed by @xmath25 .
the @xmath0 mesons or baryons carry 70% of the @xmath20 quark energy but the particles may or may not be recognizable as jets .
the hadron colliders then rely on lepton tagging , displaced vertices and reconstructed decay products to perform the measurements . in this regard ,
once the trigger provides a @xmath20-enriched sample , the tools for finding @xmath20 s are similar in hadron and lepton colliders .
narrow resonances are particularly useful because they reduce the combinatoric possibilities . from such samples , a variety of measurements have been performed and in the following sections a few have been selected for exposition .
one measurement of particular significance is that of the various @xmath0 meson and baryon lifetimes @xcite .
the ratio of @xmath20 lifetimes is important because it probes the extent to which the simple spectator model is valid and the degree to which the other diagrams including @xmath14 exchange and final state interference play a role .
the measured value of the charged to neutral meson lifetime ratio is @xmath26 , well within the theoretical prediction of 1.0 - 1.1 .
thus , to a precision of a few per - cent , the spectator model is quite good and the more subtle effects are only beginning to be seen .
there is some discrepancy for the baryon lifetimes ; however , their errors remain statistically dominated and rigorous theoretical tests await the larger data samples anticipated at the next run of the tevatron . as with the charm quark , theoretical computation of inclusive bottom production cross sections
is not well understood . for the production of open @xmath20 ,
as illustrated in figure [ fig : bopencrossect ] , the situation is as follows .
the d0 collaboration @xcite finds that the shape is well described but the cross section is a factor of two to three higher than the theory predicts .
the h1 collaboration @xcite used inclusive muon production and reported a production cross section about a factor of five above theoretical prediction .
this analysis has subsequently been revised @xcite and is now only two times the prediction .
the cdf collaboration finds that the cross sections for @xmath0 @xmath27,@xmath28 @xcite , as well as inclusive muon production @xcite are all about a factor of two higher than expected in theory .
the production of @xmath29 at cdf @xcite is a factor of two to five times larger than expected , and at zeus the photoproduction rate of @xmath29 is a factor of five too high @xcite .
various @xmath20 production mechanisms at a hadron collider have very different kinematics . at tree level two @xmath20
quarks are produced with an azimuthal separation of 180 degrees .
in contrast to this a contribution from gluon splitting results in both @xmath20 quarks recoiling against a gluon @xcite , giving a small azimuthal separation of the @xmath20 and @xmath30 quarks .
if these contributions are not computed correctly , this can result in an improper determination of acceptance and hence the cross sections can be extracted improperly .
these contributions have been explored @xcite by measuring the azimuthal separation of the muon decay from the @xmath20 quark and the jet from the @xmath31 which was tagged using the impact parameter of the tracks in the jet .
the result shown in figure [ fig : bopencrossect ] indicates that while the normalisation remains incorrect , the shape seems to be correctly reproduced but with rather large errors in the sensitive region . among the measurements of rare decays ,
one of current interest is the cleo @xcite measurement of the branching ratio for @xmath32 .
this is bad news for the measurement of @xmath33 since estimates on the experimental errors assumed a branching ratio a factor of two larger .
cdf @xcite has measured the angular distributions of the decay products in @xmath34 to obtain the polarization of the @xmath35 and found that the p - odd component is small .
this is good news for identifying channels with a well - defined cp eigenstate but the results are low in statistics .
cdf has also observed @xcite the @xmath36 and the lifetime measurement indicates that @xmath37 quark decays first : @xmath38 ps .
unfortunately , full reconstruction of the @xmath36 in @xmath39 will be difficult because the lifetime measurement implies a branching ratio that is ten times smaller .
a number of precise measurements of @xmath40 mixing at lep , slc and the tevatron exist .
the fraction of mixed mesons is given by @xmath41 and is measured as a function of proper time . in order to measure mixing , the @xmath35 meson must be reconstructed or partially reconstructed and its flavour must be tagged at production and decay .
two major classes of tagging in @xmath0 mixing are same side and opposite side tagging . given a semileptonic decay on one side where the lepton satisfied a trigger , one may search for a semileptonic decay on the other side . since the other lepton is not biased by the trigger , it generally has lower momentum and is called a slow lepton .
this tag is referred to as a slow lepton tag ( slt ) .
the soft fragmentation pion in the same hemisphere as the @xmath40 meson may be identified and used to tag the initial @xmath40 state .
this is called a same side tag ( sst ) .
the charge of particles in the opposite hemisphere may be summed with a momentum or lifetime weighting to provide a jet charge tag .
the cdf group has measurements that use the following combinations of tags @xcite : @xmath6 lep / sst ; lep/@xmath42 , lep ; @xmath43/@xmath44 ; @xmath44/@xmath44 ; @xmath6 lep / lep ; @xmath45/lep .
results for these mixing measurements are shown in table [ tab : cdfbmix ] .
they agree with one another and with other measurements of @xmath0 mixing at lep and sld @xcite .
@ll tag & @xmath46 + @xmath6 lep / sst & @xmath47 ps@xmath48 + lep /@xmath42 , lep & @xmath49 ps@xmath48 + @xmath43/@xmath44 & @xmath50 ps@xmath48 + @xmath44/@xmath44 & @xmath51 ps@xmath48 + @xmath6 lep /lep & @xmath52 ps@xmath48 + @xmath45 / lep & @xmath53 ps@xmath48 + average & @xmath54 ps@xmath48 + [ tab : cdfbmix ] similar tagging methods are used in the search for @xmath55 mixing .
this is currently done by fitting the data with a fixed mixing frequency and allowing the amplitude to vary as @xmath56 .
the result presented by the lep @xmath0 mixing working group is that the world average @xmath57 ps@xmath48 at 95% cl . because a large value of @xmath58 will result in different lifetimes for the two eigenstates ,
the @xmath55 lifetime may be fitted for two components . combined results from lep and cdf
give @xmath59 40ps@xmath48 at 95% cl .
using the world average @xmath55 lifetime , this gives @xmath60 it is worth noting that cdf runii with the layer 00 upgrade can reach a value of @xmath61 .
also , the sld and lep collaborations still have analysis that take advantage of improvements in their tracking that could in fact find a value of @xmath58 before the tevatron takes data .
cdf @xcite and opal @xcite have produced measurements of the cp asymmetry where a @xmath40 meson decays to the same eigenstate .
this asymmetry is given by : @xmath62 in all cases , @xmath63 where @xmath64 is the dilution due to mistagging and the error on the cp asymmetry goes as : @xmath65 where @xmath66 is the tagging efficiency or the fraction of events tagged and @xmath67 is the dilution or the number of right , @xmath68 , minus wrong , @xmath14 , tags .
the analysis of cp asymmetry hinges on the ability to tag accurately , maximising the dilution parameter .
cdf reconstructs @xmath2 and @xmath69 with 400 events observed .
half of these are contained in the silicon tracker , the svx , so that there is a time - dependent measurement for 200 events . because the behaviour of @xmath70 is a sine function , and the largest background is at low lifetimes , most sensitivity to @xmath71 is in a lower background region at higher lifetimes . in order to obtain @xmath1 ,
the dilution must be measured .
this is done using a @xmath72 sample to obtain the dilution in the jet charge and slow lepton tags .
a lepton @xmath6 sample is used to determine the sst dilution . a boost correction from monte carlo
must be added , resulting in a systematic uncertainty @xmath73 .
the measured dilutions and contributions to the value of @xmath1 for each of the tagging methods is detailed in table [ tab : sin2b ] .
@llll tag & @xmath74 & @xmath64 & @xmath1 + @xmath75 & @xmath76 & @xmath77 & @xmath78 + @xmath79 & @xmath80 & @xmath81 + slt & @xmath82 & @xmath83 & @xmath84 + jetq & @xmath85 & @xmath86 & @xmath87 + global & & @xmath88 + [ tab : sin2b ] the resulting cp asymmetry is plotted as a function of proper time in figure [ fig : sin2b ] .
the result is that @xmath89 or @xmath90 at 93%cl .
this may be compared to the opal measurement of @xmath91 .
this is a proof of principle showing that in the hadron collider the error is predominantly statistical .
furthermore , since the systematic errors are statistics dominated , having more data will improve the measurement .
the decay of a top quark results in six jets , or four jets , a lepton and a neutrino or two jets with two leptons and two neutrinos . in each case , two of the jets are @xmath20 jets .
these three cases comprise the major signatures for top that are studied at the tevatron .
table [ tab : topsummary ] shows the three modes , the kind of signature used to distinguish each mode , the dominant source of background , the branching ratios and the number of signal and background events in each of the two detectors , cdf and d0 .
@llll mode & signature & bkgr & br + 6 jet & @xmath20,kine & qcd & 45% + cdf & 157 & 123 & + d0 & 18 & 7 & + lep + @xmath924 jet & @xmath93 & @xmath94 _ jet _ & 30% + & lep jets @xmath20 & & + cdf & 96 & 40 & + d0 & 30 & 12 & + lep + lep & @xmath93 & drell yan & 5% + & lep lep 2jet & z@xmath95 & + cdf & 13 & 4.4 & + d0 & 5 & 1.4 & + [ tab : topsummary ] a number of measurements of the top quark have been performed .
these include : * mass : @xmath96 gev / c@xmath97 @xcite ; * @xmath98 : @xmath99 pb where theoretical predictions are 4.7 - 5.5 pb @xcite ; * @xmath100 : @xmath101 @xcite ; * @xmath14 helicity ( @xmath102 ) : @xmath103 ; where the theory predicts 0.7 @xcite ; * single top : @xmath104 15.4 pb , @xmath105 15.8 pb theoretical predictions are 1.7 and 0.7 pb respectively @xcite .
aleph @xcite has reported 18 events on a background of 10 .
there is a 4.6% chance that the background fluctuation can give these observed events and they are looking forward to collecting more statistics . the standard model branching ratio is @xmath106 . * extensive kinematic studies
@xcite have shown agreement with the expectation .
values studied are leading jet @xmath107 , jet rapidity separations , etc . *
the invariant mass @xcite of the top quark pair , @xmath108 , has been studied and a limit on a @xmath109 of @xmath110 gev / c@xmath97 has been set . more generally , the limits from the data on @xmath111 for various hypothetical masses @xmath112 of a new particle that decays to @xmath113 are presented .
* a limit on branching ratios in rare top decays @xcite has been set with b(@xmath114)@xmath115 and b(@xmath116)@xmath117 .
the value of the top cross section at the tevatron is slightly high compared to theory but within statistical errors .
however , each of the published cross section values for the individual modes for each experiment shows that all but one of them is high .
recent re - analysis of the data shows that the cross section values are tending to come down to the expected values .
in surveying the present state of heavy flavours , one is first struck by the fact that the heavy quark cross sections are all too high .
the top cross section agrees with theory but the errors are large .
the @xmath37 quarks are used at hera to successfully set the scale for qcd and measure the gluon structure function ; however , a consistent explanation of the production cross sections at hera and the tevatron is not possible .
the @xmath0 meson and baryon lifetime ratios show agreement with theory and the spectator picture .
a measurement of @xmath118 mixing is coming soon with current estimates of @xmath119 .
sld and lep may use optimized tracking algorithms and a unified analysis to find the value of @xmath58 before the tevatron , where it will most certainly be measured with precision .
cp violation measurements in hadron machines are not limited by systematics and the @xmath120 machines are starting up . with the data that will be collected at the tevatron , hera , and at the @xmath121 machines ( belle , pepii , and cesr ) there will be a fantastic increase in information on heavy flavours in the next few years .
i wish to thank the organizers for putting together a very good and enjoyable workshop .
99 ryskin m ( 1993 ) 89 ryskin m , ( 1997 ) 231 brodsky s j , ( 1994 ) 3134
frankfurt l , ( 1996 ) 3194 frankfurt l , ( 1998 ) 512 donnachie a and landshoff p v ( 1992 ) 227 collaboration , adloff c ( 1999 ) 373 collaboration , breitweg j desy 99 - 101 ( accepted by ) collaboration , adloff c ( 1999 ) 21 gribov v n and lipatov
l n _ sov .
phys . _ * 15 * ( 1972 ) 438 and 675 + dokshitzer yu l _ sov .
* 46 * ( 1977 ) 641 + altarelli g and parisi g , ( 1977 ) 297 collaboration , abe f , ( 1997 ) 572 collaboration , abachi s , ( 1996 ) 239 berger e l and jones d , ( 1981 ) 1521 bodwin g t , ( 1995 ) 1125 cano - coloma b and sanchis - lozano m a , ( 1997 ) 753 results from the lep @xmath0 lifetime working group collaboration , abbott b fermilab pub-99/144-e ( submitted to ) collaboration , adloff c , collaboration , adloff c , ( 1999 ) 156 collaboration , abe f , ( 1994 ) 4252 collaboration , abe f , ( 1993 ) 2396 collaboration , abe f , ( 1995 ) 4358 collaboration , breitweg j , ( 1998 ) 432 collaboration , abe f , ( 1996 ) 1051 collaboration , d. jaffe , _ published in the proceedings of heavy flavours 8 , southampton , uk , july 1999 _ ribon a , _ the cdf collaboration , published in proceedings 13th les rencontres de physique de la valle daoste _ collaboration , abe f , ( 1995 ) 3068 collaboration , abe f , ( 1998 ) r112004 collaboration , abe f , ( 1998 ) 2432 collaboration , bauer g , _ published proceedings 13th topical conference on hadron collider physics , tata institute of fundamental research , mumbai , india , january 14 - 20 , 1999 _ results from the lep @xmath0 mixing working group collaboration , affolder t , fermilab - pub-99/225-e , submitted to collaboration , ackerstaff r , ( 1998 ) 379 apollinari g , representing the d0 and cdf collaboration , _ published proceedings 26th slac summer institute on particle physics _
gallinaro m , the cdf collaboration , _ published proceedings 14th international workshop on high energy physics and quantum field theory ( qfthep 99 ) , moscow , russia , may 27-june 2 , 1999 _
koehn p , the cdf collaboration , _ published proceedings international europhysics conference on high - energy physics ( eps - hep 99 ) , tampere , finland , july 15 - 21 , 1999 _ collaboration , antonelli m , _ published proceedings international europhysics conference on high - energy physics ( eps - hep 99 ) , tampere , finland , july 15 - 21 , 1999 _ collaboration , affolder t , ( 1999 ) 092001 production kinematics at hera and the @xmath14 dependence of the differential cross section for @xmath2 production measured at hera for a variety of @xmath122 compared qcd . ,
title="fig:",width=377 ] production kinematics at hera and the @xmath14 dependence of the differential cross section for @xmath2 production measured at hera for a variety of @xmath122 compared qcd .
, title="fig:",width=566 ] cross section in cdf and the azimuthal separations of the muon from semileptonic decay of a b and the jet from the @xmath31 at cdf.,title="fig:",width=377 ] cross section in cdf and the azimuthal separations of the muon from semileptonic decay of a b and the jet from the @xmath31 at cdf.,title="fig:",width=377 ] | current issues in the studies of heavy flavours in colliders are described with particular emphasis on experiments in which the uk is involved .
results on charm production at hera are examined and compared to those at the tevatron .
@xmath0 production rates at the tevatron as well as the status of @xmath0 lifetimes and mixing in the lep collaborations and at the tevatron are highlighted .
the measurement of @xmath1 from cdf is described as well as the most recent results on top physics at the tevatron . to for publication . |
, mobile data traffic is explosively and continuously rising due to smart phone and tablet users and it is expected that next - generation ( a.k.a . the 5th generation )
cellular networks will offer a 1000x increase in network capacity as well as a 1000x increase in energy - efficiency in the following decade to meet such an excessively high user demand @xcite .
the most prominent way to increase the network capacity is the _ network densification _
@xcite by either implementing more antennas at base stations ( bss ) called large - scale antenna system ( lsas ) @xcite or adding more small cells at hot spot areas called ultra - dense network ( udn ) @xcite . by using a lot of antennas at each bs
, the lsas can exploit massive spatial dimension to generate a sharp beam and thus it can provide higher spectral efficiency and also lower power consumption @xcite .
however , the performance of the lsas is highly limited by the accuracy of channel state information ( csi ) and varies over the geographical locations of users . on the other hand ,
the udn exploits spatial reuse obtained from deploying more small bss so that it can offer geographically uniform performance to each user due to the reduction of the access distance between users and bss @xcite , @xcite
. however , its performance is limited by uncoordinated out - of - cell interference so that an efficient interference management is a key challenge for the udn , while managing the costs in terms of the front / backhaul overhead and computational complexity for a network operator . large - scale cloud radio access network ( ls - cran ) is recently regarded as a novel wireless cellular network architecture to unify the above two architectures and is able to implement the interference handling mechanisms introduced for the long - term evolution ( lte ) or lte - advanced , such as the enhanced inter - cell interference coordination @xcite and the coordinated multi - point transmission @xcite , @xcite . in the ls - cran
, lots of bss each with a massive antenna array are connected to a virtualized central processing unit ( cpu ) via dedicated front / backhaul link and some of the baseband processing functionality of each bs is migrated to the cpu . as a result , it is expected that the performance of the ls - cran is much higher than that of a conventional network .
a natural objective of the ls - cran is to maximize the number of supporting users while preserving their quality - of - service ( qos ) requirements under practical constraints .
if the ls - cran can manage the inter - user interference and inter - bs interference ideally , the network can be transformed into a multi - access ( ul ) or broadcast ( dl ) channel without any interference @xcite . however , unfortunately in practical systems , such an ideal situation is hard to be realized due to various practical limitations so that the maximum number of supporting users with a qos requirement is carefully considered under such practical constraints .
there are three dominant practical constraints in the ls - cran .
first , the network total transmit power should be maintained at an appropriate level not only for saving power consumption ( and operating cost ) @xcite , but also for adopting cheap radio - frequency ( rf ) components in each bs and user @xcite .
second , the capacity of the front / backhaul links connecting the bss and the cpu is usually limited as pointed out in @xcite and installing dedicated high - capacity front / backhaul link for every bs causes too much cost , especially in a dense deployment @xcite .
last , the pilot resource for the ls - cran to acquire the csi and handle the interference caused from a large number of bss , antennas , and users by using the maximum ratio transmission ( mrt ) @xcite or the zero - forcing ( zf ) @xcite operation , needs to be limited .
then , a naturally raised question is that how many users can be supported by the ls - cran with a qos requirement under these three practical constraints ?
also , how do the target qos and the number of users satisfying the target qos simultaneously grow as the network size increases ?
the answers are ready to be addressed in this paper .
although the performance of an ls - cran has been widely investigated in literature @xcite , @xcite via simulations , an intuitive analytic result considering practical limitations is not available yet .
the exact distribution of the signal - to - interference - plus - noise ratio ( sinr ) of the mrt or zf operation is derived in @xcite for the case of two distributed bss and its laplace approximated distribution is derived in @xmath2 for a general ls - cran . in @xcite , a simpler form of the sinr distribution is derived by approximating the sinr as a gamma random variable . as parallel works ,
the average achievable rate of the mrt or zf operation is derived in @xcite for a single bs case , and in @xcite for the case of fully - distributed or fully - colocated antennas as a function of the bs and user locations . in order to understand the network behavior
more intuitively , stochastic geometry has been recently adopted to remove such dependency and describe important network metrics in a probabilistic way . in @xcite-@xcite ,
the sinr distributions of cooperative transmission schemes are provided by adopting stochastic geometry , but the analysis resorts on multiple numerical integrals . in @xcite , an asymptotic analysis on the tail of the sinr distribution
is provided by applying large - deviation theory and stochastic geometry . in @xcite ,
the average achievable rate of the mrt and zf successive interference cancellation is investigated and its scaling laws are obtained in a dense random network with multiple receive antennas . however , the effects of the limited transmit power , the limited front / backhaul capacity , or the limited pilot resource have not been jointly considered . as the network is densified
, the effect of erroneous csi becomes more crucial because all of users and antennas are jointly processed in the cpu so that small csi error can dramatically degrade the overall network performance . in an ls - cran , there exist three main sources causing erroneous csi : the limited transmit power or the limited pilot resource in the process of csi acquisition and the limited capacity of front / backhaul links . in @xcite , the effect of the csi acquisition error on the energy and spectral efficiency is analyzed for the dl single - cell lsas using the mrt or zf operation . in @xcite , an information - theoretic capacity bound is derived for ul cran by modeling the csi error as an additive gaussian noise . in @xcite and @xcite ,
the average achievable rate and its asymptotic analysis are obtained in multi - antenna relay networks by considering the effect of the csi acquisition errors . in @xcite , authors model and study the front / backhaul link by using stochastic geometry .
the achievable rate with various front / backhaul data compression strategies is analyzed in the ul scenario @xcite and this result is extended to the dl scenario @xcite . in @xcite , the achievable rate of a cooperative transmission scheme is derived by considering backhaul burdens .
however , these results become quite intractable if applied to the ls - cran with large number of antennas and users so that more insightful and intuitive analysis for the ls - cran is required .
the main objective of this paper is to characterize the asymptotic behavior of the ls - cran under three practical constraints : ( 1 ) the limited total transmit power , ( 2 ) the limited front / backhaul capacity , and ( 3 ) the limited pilot resource .
the major contributions of this paper are summarized as follows . *
the scaling exponents of the signal - to - interference - plus - noise ratio ( sinr ) defined and derived as a function of the numbers of bs , bs antennas , and single - antenna users , and the ul / dl transmit power for the interference - free ( if ) , mrt , and zf operations , which are stated in theorems 1 - 3 . * based on the derived scaling exponents , four fundamental regimes are distinguished according to the ul transmit power ( or the csi acquisition error ) : ( i ) the extremely high ( ul transmission ) power regime ( ) , ( ii ) the high power regime ( ) , ( iii ) the medium power regime ( ) , and ( iv ) the low power regime ( ) .
the scaling exponents of the three operations are derived according to the four regimes with insightful discussions .
* the limited front / backhaul capacity can be successfully transformed into partial associations , where each bs serves only users affordable with its front / backhaul capacity .
the analytical result shows that such limited front / backhaul capacity and pilot resource constrain the network behavior in a similar way as the limited transmit power , as stated in theorems 4 and 5 . *
in order to characterize the network - wise performance of the ls - cran , we define the number of supportable users as the maximum number of users satisfying a pre - determined sinr requirement .
then , we derive the tradeoff in the scaling exponents of them as the network size increases under the practical limitations , as stated in theorem 6 .
the reminder of this paper is organized as follows . in section
ii , the ls - cran system model is described and the scaling exponent of the sinr is defined as a network performance measure .
section iii provides some candidate operations suitable for practical implementation and the corresponding scaling exponents are analyzed by considering practical limitations to provide an intuitive understanding on the ls - cran in section iv . in section
v , the number of supportable users is defined as another important network metric and analyzed to provide an insight on the asymptotic network behavior of the ls - cran .
finally , conclusion is given in section vi .
matrices and vectors are respectively denoted by boldface uppercase and lowercase characters .
the superscript @xmath3 , @xmath4 and @xmath5 denote the conjugate , transpose and conjugate transpose , respectively .
@xmath6 , @xmath7 $ ] , @xmath8 , and @xmath9 stand for the cardinality of a set , statistical expectation , trace of a square matrix , and kronecker delta function , respectively .
also , @xmath10 and @xmath11 are the @xmath12 all one vector and all zero vector , respectively .
further , @xmath13 denotes the diagonal matrix whose @xmath14th element is @xmath15 , and @xmath16 denotes the circularly symmetric complex gaussian distribution with mean @xmath17 and variance @xmath18 . for a @xmath19 vector @xmath20 ,
@xmath21_{a = 1,b = 1}^{a , b}$ ] denotes the @xmath22 matrix @xmath23}^h},\cdots,[{\bf{x}}_{1b}^h,\cdots,{\bf{x}}_{ab}^h]^h } \right]$ ] . in this paper
, we only deal with a scaling exponent , @xmath24 , of a random measure , @xmath25 , as @xmath26 in a probabilistic sense , which is mathematically defined as follows .
the scaling exponent of @xmath25 is @xmath24 in probability if @xmath27 holds for any positive finite @xmath28 .
also a slightly abused notation @xmath29 is used if there is no finite positive @xmath28 satisfying ( [ eq_a1 ] ) for any finite @xmath24 .
let @xmath24 and @xmath30 be the scaling exponents of @xmath25 and @xmath31 , respectively .
then , the following modified order notations are used through this paper .
* @xmath32 or @xmath33 , if @xmath34 , * @xmath35 , if @xmath36 , * @xmath37 or @xmath38 , if @xmath39 , * @xmath40 , if @xmath41 , and * @xmath42 or @xmath43 , if @xmath44 .
m3 cm m7 cm & * definition * + @xmath45 , @xmath46 , @xmath47 & the numbers of bss , bs antennas , or users , respectively + @xmath48 & the network size , @xmath49 + @xmath50 & the set of bss , where @xmath51 is the location of bs @xmath52 + @xmath53 & the set of users , where @xmath54 is the location of user @xmath55 + @xmath56 & the set of bss associated to user @xmath55 + @xmath57 & the set of users associated to bs @xmath52 + @xmath58&the association range + @xmath59 , @xmath60 , @xmath61 & scaling exponents for the number of bss , bs antennas or users , respectively + @xmath62 , @xmath63 & the number of associated bss per user and its scaling exponent + @xmath64 , @xmath65 & the number of pilots and its scaling exponent + @xmath66 & pathloss exponent ( @xmath67 ) + @xmath68 & @xmath69 wireless channel between bs @xmath52 and user @xmath55 , where @xmath70 is a long - term fading , @xmath71 is a short - term fading + @xmath72 , @xmath73 , @xmath74 , @xmath74 & the ul or dl transmit power for user @xmath55 , respectively and their scaling exponents + @xmath75 , @xmath76 , @xmath77 & snr , sir or sinr of user @xmath55 when operation @xmath78 is applied , respectively and their scaling exponents + +
consider an ls - cran system as illustrated in fig .
1 ( a ) . suppose that @xmath45 bss with @xmath46 antennas and @xmath47 users with a single antenna are uniformly distributed on a finite region @xmath79 .
the sets of bss and users are denoted as @xmath80 and @xmath81 , respectively . with a slight abuse of notations , @xmath51 and @xmath54
are used as the locations of bs @xmath52 and user @xmath55 , respectively .
it is assumed that each of users is associated with neighboring bss and the set of the bss serving user @xmath55 is defined as @xmath82 where @xmath83 is the association range and the set of users associated with bs @xmath52 is denoted as @xmath84 . to guarantee each of users
is served by at least one bs , @xmath85 for all @xmath55 .
note that @xmath57 and @xmath86 are not necessarily disjoint and typical users are associated with multiple bss for being served cooperatively . in this paper , a network is called _ fully associated _
, if @xmath87 for all @xmath55 .
otherwise , the network is called _ partially associated_. the bss are connected to the cpu via high - speed dedicated front / backhaul link for enabling a cooperative transmission operation . for inter - signaling between the bss and the cpu ,
the sets of front / backhaul information are defined as @xmath88 and @xmath89 , where @xmath88 is the fronthaul information set ( from cpu to bs @xmath52 ) and @xmath89 is the backhaul information set ( from bs @xmath52 to cpu ) . the elements of @xmath88 and @xmath89 are closely related according to a specific transmission operation , which will be described in section iii . the dl frame structure operating in time - division duplex ( tdd ) mode is shown in fig .
1(b ) and has three phases as follows .
ul training ( ut ) phase , non - cooperative data transmission ( dt ) phase , and cooperative dt - phase . in the ut - phase , the instantaneous csi of each user is estimated by receiving users pilot ( or reference ) signal at each associated bs independently . after the ut - phase , the non - cooperative dt - phase is performed first and followed by the cooperative dt - phase . in the non - cooperative dt - phase , each bs separately transmits the data symbols to the associated users without any front / backhaul exchange . in the cooperative dt - phase ,
the bss jointly transmit the data symbols to the associated users with the aid of the exchanged information via the front / backhaul link . for simplicity , the cooperative dt - phase is only focused in this paper .
let @xmath90 denote the @xmath91 flat - fading dl channel vector from bs @xmath52 to user @xmath55 , which can be written as @xmath92 where @xmath93 is the short - term csi whose elements are independent and identically distributed ( i.i.d . )
@xmath94 and @xmath95 is the long - term csi depending on the path - loss and shadowing .
the long - term csi between bs @xmath52 and user @xmath55 is modeled as @xmath96 , where @xmath97 is the wireless channel path - loss exponent .
it is assumed that the short - term csi of each user remains constant within a given frame but independent across different frames , while the long - term csi does not vary during a much longer interval .
further , it is assumed that the long - term csis among all bss and users are perfectly known at the cpu through an infrequent feedback with a negligible overhead .
additionally , we assume perfect tdd reciprocity calibration so that the ul channel vector is just a transpose of the dl channel vector ( i.e. , the ul channel vector from user @xmath55 to bs @xmath52 is denoted as @xmath98 ) .
let @xmath99^t$ ] be the @xmath100 information symbol vector with @xmath101=\mathbf{i}_k$ ] , where @xmath102 denotes the information symbol for user @xmath55 .
let @xmath103^t$ ] be the @xmath104 _ global _ transmitted signal vector , where @xmath105 is the _ local _ transmitted signal vector of bs @xmath52 , given by @xmath106 where @xmath107 denotes the @xmath108 power allocation matrix and @xmath109 denotes the power allocated to user @xmath55 , @xmath110_{l=1,k=1}^{l , k}$ ] denotes the @xmath111 precoding matrix and @xmath112 denotes the @xmath69 precoding vector of user @xmath55 for bs @xmath52 , and @xmath78 denotes the cooperative transmission operation used in the ls - cran . since the bs serves the associated users only , @xmath113 for @xmath114 ( or equivalently @xmath115 ) .
note that the transmitted signal vector of bs @xmath52 can be written by @xmath116 .
then , the dl transmit power for user @xmath117 , @xmath118 , is given as @xmath119 and the total dl transmit power is given by @xmath120 .
let @xmath121^t}$ ] be the @xmath100 aggregated received signal vector , given by @xmath122 where @xmath123_{l=1,k=1}^{l , k}\right)^h$ ] denotes the @xmath124 channel matrix among all users and all bss and @xmath125^t}\sim\mathcal{cn}(\mathbf{0},\mathbf{i}_k)$ ] denotes the @xmath126 noise vector . then , the received signal at user @xmath55 can be expressed as @xmath127 where @xmath128 is the effective channel seen at user @xmath55 . the ul channel is estimated during the dedicated ut - phase and then the dl channel is obtained by the tdd channel reciprocity .
it is assumed that the length of the ut - phase is @xmath64 and there are @xmath64 orthonormal pilot signals , denoted as @xmath129 , where @xmath130 .
then , user @xmath117 transmits @xmath131 during the ut - phase of length @xmath64 , where @xmath132 is the ul transmit power of user @xmath117 and @xmath133 is the index of the pilot signal allocated to user @xmath117 .
the total ul transmit power is denoted as @xmath134 .
then , the @xmath135 received signal matrix at bs @xmath52 during the ut - phase can be written as @xmath136 where @xmath137 denotes the @xmath135 noise matrix whose elements are i.i.d .
@xmath94 . using the minimum mean - square error ( mmse ) channel estimator @xcite , the estimated short - term csi of user @xmath138 at bs @xmath52 can be written as @xmath139 where @xmath140 @xmath141 @xmath142 with @xmath143_m \sim \mathcal{cn}(0,1)$ ] . in ( [ estimated_csi ] )
, the first term is the desired user s channel , the second term is the leakage from the other users channels , called the _ pilot contamination _ ( pc ) effect due to the pilot signal reuse , and the third term is the noise part . invoking the orthogonality principle of the mmse estimator @xcite , @xmath71 can be decomposed as @xmath144 , where @xmath145 and @xmath146 are mutually independent .
note that the estimated version of @xmath147 at bs @xmath52 is given as @xmath148 for @xmath149 or @xmath150 for @xmath151 and thus the estimated version of @xmath152 is given as @xmath153_{l=1,k=1}^{l , k}\right)^h$ ] .
when operation @xmath78 is used , the signal - to - noise ratio ( snr ) , @xmath75 , the signal - to - interference ratio ( sir ) , @xmath76 , and the signal - to - interference - plus - noise ratio ( sinr ) , @xmath77 , are respectively defined as @xmath154 obviously , @xmath75 , @xmath76 , and @xmath77 are random variables depending on the realization of the short - term fading and the long - term fading ( i.e. , realization of locations of users and bss ) .
one of the main objectives of this paper is to characterize the asymptotic behavior of @xmath155 when the key network parameters such as the number of bss @xmath45 , the number of users @xmath47 , and the number of bs antennas @xmath46 are scaled up .
to do this , we define an auxiliary parameter @xmath49 as the network size ( or equivalently , the total number of antennas in the network ) and make the following relations : @xmath156 where @xmath59 , @xmath61 , and @xmath60 denote the scaling exponents of the numbers of bss , users , and bs antennas , respectively .
note that we only consider the case where @xmath157 and @xmath158 by definition .
then , the asymptotic performance of the network can be characterized as follows .
the scaling exponent of the sinr of operation @xmath78 , @xmath159 , is the order of growth of the sinr of a randomly selected user as @xmath48 increases such that , for any @xmath160 , @xmath161 manipulating ( [ eq_17 ] ) , we can also obtain @xmath162 which implies that the sinr of a randomly selected user is bounded by @xmath163 $ ] as @xmath48 increases . invoking our order notations ,
we can simply write @xmath164 and the sum - rate served by a network can also be written as @xmath165 .
our main goal in this paper is to characterize @xmath159 for various ls - cran operations under practical constraints so that how the target sinr and the number of users satisfying the target can simultaneously grow as the network size increases .
similarly , we define @xmath166 and @xmath167 as the scaling exponents of the snr and sir , respectively .
note that it is sufficient to find @xmath166 and @xmath167 and then @xmath168 is obtained since @xmath77 is equal to the harmonic mean of @xmath75 and @xmath75 , i.e , @xmath169 .
in this section , the ideal if operation is introduced as a reference system and practical cooperative operations are reviewed .
note that a comprehensive review on the operations is beyond the scope of this paper so that two well - known practical operations , mrt operation @xcite and zf operation @xcite , are focused with the three practical limitations . as a reference , the ideal if operation is considered , where the interference term in ( [ received_signal_from_bs_l ] ) is removed by genie perfectly without any cost while the desired signal power is maximized by using the if precoding matrix , @xmath170 obviously , the ideal if operation provides an upper - bound on the performance of any practical operation .
since this operation can not be realizable , @xmath171 and @xmath172 are not defined .
mrt operation tries to maximize the received signal power of a desired user without considering the effect of interference to undesired users @xcite .
this operation is regarded as a good candidate as the number of bs antennas increases due to its low - computational complexity and low - overhead requirement @xcite .
the precoder for mrt operation is given by @xmath173 since the precoder of mrt operation does not need information exchange among bss , the backhaul and fronthaul information sets can be expressed as @xmath174 , and @xmath175 , respectively .
zf operation can cancel the interference term in ( [ received_signal_from_bs_l ] ) ( perfectly , provided that the perfect channel estimation is available at the cpu ) at the expense of the desired signal power loss @xcite .
the precoder for zf operation is given by @xmath176 note that zf operation can be used only when the number of antennas in the system is larger than or equal to that of users , i.e. , @xmath177 and the estimated channel matrix , @xmath178 , has full - rank .
to construct the zf precoder , the cpu requires to know the estimated channel matrix @xmath178 so that the short - term csis of all associated users at each bs need to be conveyed to the cpu via the dedicated front / backhaul link .
thus , the backhaul and fronthaul information sets can be expressed as @xmath179 and @xmath180 , respectively .
note that it is well - known that the performance of zf operation is better than that of mrt operation if the network has sufficient transmit power .
but , if not , zf operation can be inferior to mrt operation . in this paper , we will show that zf operation is always superior or identical to mrt operation in the viewpoint of the scaling exponent of the sinr regardless of the operating transmit power . in a real - world network , there exist various practical limitations . in this paper
, we consider limitations on the total transmit power , the front / backhaul capacity , and the pilot resource .
the major merit of the ls - cran is that it can decrease the transmit power consumption so that it is suitable to a green communication .
so , the total transmit power is a key constraint in a future cellular system . in order to limit the total transmit power of uplink or downlink , similarly as in ( [ eq_16 ] ) , we make an additional asymptotic relationship as @xmath181 where @xmath74 and @xmath182 denote the scaling exponents of the ul and dl transmit powers , respectively . also , the total transmit power is @xmath183 .
since the front / backhaul information sets closely depend on the cardinality of the set @xmath184 , i.e. , the number of bss serving user @xmath117 , so that the front / backhaul link overhead can be quantified from the network association state .
define @xmath185 as the number of bss associated to user @xmath117 .
then , similarly as in ( [ eq_16 ] ) , we make an additional asymptotic relationship as @xmath186 where @xmath63 denotes the scaling exponent of the number of associated bss per user . since at most @xmath45 bss can be associated to each user , @xmath187 . due to the natural time - frequency selectivity of wireless channel
, it is required to acquire csi of users for every coherence interval .
however , the number of dedicated orthogonal pilots is strictly limited by the number of orthogonal resources in one coherence interval so that the number of pilots ( pilot resources ) , @xmath64 , should be limited appropriately . in this paper , similarly as in ( [ eq_16 ] ) , we make an additional asymptotic relationship as @xmath188 where @xmath65 denotes the scaling exponent of the number of pilot resources . since at most @xmath47 pilot sequences are sufficient to guarantee no pilot reuse , @xmath189 .
note that @xmath190 implies that the available number of pilot sequences increases as the network size increases .
in order to quantify the effect of the csi accuracy according to the ul transmit power , we consider the following four regimes . * case ( @xmath191 ) : the ul transmit power is sufficiently high so that every bs can acquire the accurate csi of a randomly selected users .
* case ( @xmath192 ) : each bs can acquire the accurate csi of a randomly selected users within a distance of @xmath193 . *
case ( @xmath194 ) : randomly selected user s csi is erroneous even at the nearest bs but is still meaningful for providing an array gain .
* case ( @xmath195 ) : randomly selected user s csi becomes quite poor even at the nearest bs so that no array gain can be provided .
+ suppose that if operation is used with a full association ( @xmath196 ) and no pilot reuse ( @xmath197 ) .
then , the scaling exponents are respectively given by @xmath198 where @xmath199 denotes the array gain and @xmath200 . please see appendix a. theorem 1 is also illustrated in figs . 2 and 3 according to @xmath74 and @xmath59 , respectively .
intuitively , the snr of if operation is composed of the three parts as @xmath201 where the first part is the dl transmit power , the second part is the densification gain which comes from the decrease of the access distance of @xmath202 and the last part is the array gain of a coherent transmission which depends on the csi accuracy and thus the ul transmit power .
[ snr behavior of if operation ] fig .
2 reveals how the ul transmit power affects on the snr behavior . in and , the full array gain ( @xmath203 ) is achieved so that an additional ul transmit power does not improve the quality of dl service in the network , i.e.
, is wasteful . in , a partial array gain ( @xmath204 )
depending on the ul transmit power is obtained so that the network total power needs to be consumed by considering both the dl transmit power and the csi accuracy . in , no array gain ( @xmath205 ) is obtained due to poor csi accuracy so that the quality of dl service becomes irrelevant to the ul transmit power and its performance is identical to the random beamforming without small - scale csis in @xcite .
3 shows the snr behavior according to the bs scaling exponent .
it turns out that additional bss ( even with smaller number of bs antennas at each bs ) are always beneficial but the slope of @xmath206 ( vs. @xmath59 ) varies according to @xmath74 .
the slope becomes @xmath207 in , @xmath208 in , and @xmath209 in or , which implies that additional bss ( while keeping the network size fixed ) become the most effective in because the additional bss improve not only the densification gain but also the array gain and the least effective in or because only the densification gain is improved .
@xmath210 suppose that mrt operation is used with a full association ( @xmath196 ) and no pilot reuse ( @xmath197 ) .
then , the scaling exponents are respectively given by @xmath211 where @xmath212 .
please see appendix b. theorem 2 is also illustrated in figs . 2 and 3 according to @xmath74 and @xmath59 , respectively .
interestingly , although @xmath213 is identical to @xmath206 , the gap between @xmath214 and @xmath213 , denoted by @xmath215 , changes according to the sign of @xmath216 . from the definition of @xmath217 ,
the interference caused in mrt operation at a randomly selected user @xmath55 can be represented as @xmath218 where the second and third parts depend on the sign of @xmath219 . when @xmath220 , one bs should simultaneously serve @xmath221 users apart by @xmath202 so that there are @xmath221 interferer whose received power is @xmath222 .
when @xmath223 , the dominant interference comes from the bs apart by @xmath224 so that the received power of the dominant interference is @xmath225 .
[ asymptotical optimality of mrt operation ] in order for mrt operation to behave as if operation asymptotically ( i.e. , @xmath226 ) , the dl transmit power should be limited as @xmath227 which gives us the following insights . * for mrt operation ,
the if optimality condition depends only on the numbers of users ( @xmath61 ) and bss ( @xmath59 ) and the dl transmit power ( @xmath228 ) , but is independent to the number of bs antennas ( @xmath60 ) and the ul transmit power ( @xmath74 ) . * when the number of antennas in a bs is much larger than the total number of users in a multi - cell network , mrt operation becomes interference - free asymptotically as expected in literature @xcite .
however , this is of little interest because the network size is too large ( or the number of users is too small ) .
instead , theorem 2 and ( [ eq_23_mrt ] ) gives more insightful if optimality condition for mrt operation being asymptotically interference - free in a multi - cell network .
@xmath210 [ sir behavior of mrt operation ] as can be seen from fig .
3 , additional bss are beneficial ( while fixing @xmath48 ) in most cases and the slope of @xmath214 ( vs. @xmath59 ) can be @xmath229 , @xmath230 , @xmath209 and @xmath207 .
interestingly , when @xmath231 , the slope of @xmath214 becomes 0 , which means that additional bss are wasteful as long as @xmath59 is within that interval . on the other hand ,
the slope becomes @xmath207 in and in @xmath232 if @xmath233 , in which additional bss are the most effective .
@xmath210 suppose that zf operation is used with a full association and no pilot reuse .
then , the scaling exponents are respectively given by @xmath234 where @xmath235 . please see appendix c. theorem 3 is also illustrated in figs . 2 and 3 according to @xmath74 and @xmath59 , respectively .
although the snr exponent is identical to that in mrt operation , the interference is reduced as the csi accuracy improves so that the gap between @xmath236 and @xmath237 , @xmath238 , varies as @xmath74 increases . again , from the definition of @xmath238 , the interference in mrt operation at a randomly selected user @xmath55 for @xmath220 can be represented as @xmath239 the csi of the users whose access distance is @xmath240 is accurately estimated at the nearest bs , while the csi of the users whose access distance is @xmath241 is poor .
so , when @xmath242 , i.e. , or , the csis of all users are poorly estimated at all bss because the access distance is @xmath243 .
thus , the interference cancellation is not effective so that the interference in zf operation is asymptotically the same to that in mrt operation .
however , when @xmath244 , i.e. , or , accurate csi is available so that the cancellation operation becomes effective . in , the dominant interference comes from the bs apart by @xmath245 so that the received power of the dominant interference is @xmath246 .
the dominant interferers are located in the doughnut of radii of @xmath193 and @xmath247 with an arbitrarily small @xmath28 so that the number of interferers is @xmath248 . in
, the number of dominant interferers becomes @xmath249 , while the received power of them is still @xmath246 which can be explained as follows .
once a bs acquires sufficiently accurate csi of a user , the interference caused from the user becomes proportional to the channel estimation error , i.e. , inversely proportional to the ul transmit power .
thus , in and , the slopes of the reduction in zf operation are @xmath250 and @xmath230 with respect to @xmath74 , respectively , as shown in fig .
, every bs can acquire all user s csi with a sufficiently good accuracy so that the slope becomes @xmath230 . in , however ,
bss far from each user ( outside the circle with radius of @xmath251 ) can not obtain its csi accurately so that zf operation does not reduce the interference from those bss .
note that @xmath237 does not improve as @xmath74 increases within @xmath252 $ ] in or if @xmath220 because no actual interference reduction happens during that interval due to the low user density .
[ asymptotical optimality of zf operation ] in order for zf operation to behave as if operation asymptotically ( i.e. , @xmath253 ) , the dl transmit power should be limited as follows .
if @xmath254 , @xmath255 or if @xmath256 @xmath257 which gives us the following insights .
* unlike mrt operation , the if condition of zf operation depends on the ul transmit power ( @xmath74 ) as well as @xmath59 , @xmath61 , and @xmath182 , but is still independent to the number of bs antennas ( @xmath60 ) . *
@xmath258 in and while @xmath259 in or , which implies that multi - cell cooperation using zf operation is useless without a sufficient csi accuracy .
as the ul transmit power increases , zf operation begins to further reduce the interference caused from other bss users as shown in fig .
2 . [ sir behavior of zf operation ] as can be seen again from fig .
3 , additional bss ( while fixing @xmath48 ) are always beneficial and although the slope of @xmath237 vs. @xmath59 is identical to that of @xmath214 vs. @xmath59 in or , it remains @xmath209 in or , unlike mrt operation . @xmath210 suppose that operation @xmath78 is used with a partial association and no pilot reuse ( @xmath197 ) , in which each user is associated with @xmath260 nearest bss . then , the gap @xmath261 in theorems 2 and 3 changes to @xmath262 where @xmath263 when a bs associates with only a part of users , the cpu does not obtain the csis of the non - associated users from the bs , whose effect is identical to the case where the ul transmit power of user @xmath117 becomes @xmath264 at far bss @xmath265 , which can directly prove this theorem .
[ effect of the limited front / backhaul capacity ] theorem 4 informs the relation between performance degradation caused from the erroneous csi due to the limited network total power and that caused from the partial association due to the limited front / backhaul capacity .
since @xmath217 is independent to @xmath74 , any partial association does not affect mrt operation , which implies that mrt operation is asymptotically same to a single - cell operation ( no cooperation ) . however , since @xmath266 does depend on @xmath74 , the limited front / backhaul capacity degrades the asymptotic performance of zf operation . in @xmath267 or @xmath232 , @xmath268
so that any association range with a positive exponent is wasteful , i.e. , cooperation must be confined among finite number of nearby bss . however , in @xmath269 or @xmath270 , larger association range with exponent up to @xmath271 improves the asymptotic performance , which requires the total front / backhaul capacity up to @xmath272 complex values .
@xmath210 suppose that @xmath273 orthonormal pilot sequences are available in a fully associated network .
then , the scaling exponents are as in theorems 2 and 3 by replacing @xmath274 and @xmath275 with @xmath276 and @xmath277 , respectively , where @xmath278 with @xmath279 |c||c|c|c|c| & interference free&maximum ratio transmission&zero forcing&operating + & @xmath280&@xmath281&@xmath282®ime + & & + & @xmath283 & & + & & + & @xmath284 & & + & & + & @xmath285 & & + & & & + & & & + & & & + & & & + & & @xmath286&@xmath287 & + & & @xmath288&@xmath289 & + & & @xmath290&@xmath291 & + & & @xmath292&@xmath293 & + first , consider how the array gain @xmath274 changes according to the power leakage due to the pilot reuse .
when @xmath294 , no pilot reuse among nearby users ( within a typical nearby bs range ) so that the transmit power leakage due to the pilot contamination effect is negligible , i.e. , @xmath295 .
it means that a network requires @xmath1 orthonormal pilot sequences to achieve the maximum snr scaling exponent .
when @xmath296 , @xmath276 is upper - bounded by @xmath297 due to the non - negligible power leakage ( in the same order ) to the @xmath298 users using the common pilot . now , consider how the gap @xmath261 changes .
let @xmath299 and @xmath300 be the interference at a randomly selected user @xmath55 caused by the users using the common pilot and caused by the other users , respectively .
then , as similarly discussed in ( 20 ) , @xmath301 now , consider the @xmath302 interferers using the same pilot .
since they share the same array gain to the desired user , similarly as in ( 31 ) , we obtain @xmath303 then , the final step is to find the maximum between @xmath217 and @xmath304 .
thus , @xmath305 can be represented as @xmath306 for zf operation , @xmath307 and @xmath299 is same as in ( 32 ) , which completes the proof .
[ effect of the limited pilot resource ] theorem 5 informs the performance degradation due to the limited pilot resource . although the limited pilot or limited backhaul capacity affects the interference ( i.e. , sir ) , the pilot contamination due to the limited pilot resource affects both on the snr and sir , where the snr loss ( array gain reduction ) is caused by the transmit power leakage to the contaminated users while the sir loss comes from the fact that no array gain is available over the contaminated users ( i.e. , the same array gain is available over the background noise ) . @xmath210
for a practical network , it would be of the most interest how the target qos ( in terms of the sinr ) and the number of users satisfying the target qos can simultaneously grow as the network size increases . also , in the future cellular system called the 5th generation , among the most important key performance indicators are the guaranteed ( edge - user ) throughput and the corresponding connectivity by which the tradeoff between them is of the most interest .
the scaling exponent of the number of supportable users for operation @xmath78 , @xmath308 , is defined as the growth rate of the maximum number of users while guaranteeing the qos , which is given by @xmath309 where @xmath310 is the pre - determined qos requirement on the sinr exponent for the network .
the case @xmath311 means that there are a lot of users and each user in a network requires a fixed data rate as the network size increases .
this case can be considered as an internet - of - things ( iot ) scenario in which there are a lot of devices requiring small data volumes .
however , the case @xmath312 means that there are a lot of users and all users in a network require much increased data rate as the network size increases , which can reflect future applications requiring high data volumes such as ultra - high definite ( uhd ) video streaming .
obviously , higher qos requirement ( i.e. , higher @xmath310 ) results in lower number of supportable users ( i.e. , lower @xmath308 ) and vice versa .
however , the tradeoff between them as the network size increases is non - trivial and of interest , especially with the three practical constraints .
the following theorem gives the answer on this question under the total transmit power constraint .
suppose that a fully associated network ( @xmath196 ) without pilot reuse ( @xmath197 ) is constrained by the network total transmit power as @xmath313 .
then , the scaling exponent of the number of supportable users is @xmath314 where @xmath315 is given in table ii .
see appendix e for the sketch of the proof .
the contour plots of @xmath308 on @xmath316 , @xmath317 , are illustrated in fig .
4 , in which higher slope of the contour implies better power efficiency ( @xmath310 over @xmath318 ) . note that theorem 6 can be straightforwardly extended to the case of a partially associated network ( @xmath319 ) , i.e. , with the limited front / backhaul capacity constraint and/or with a pilot reuse ( @xmath320 ) , i.e. , the limited pilot resource constraint .
however , it is too complex to represent it as in table ii so that its numerical results will be shown in this paper .
( operating regimes ) regions @xmath321 , @xmath322 , and @xmath323 are snr - limited operating regimes , ( i.e. , @xmath324 ) so that the same number of users can be supported regardless of operations . with respect to the ul transmit power , regions @xmath321 , @xmath322 , and @xmath325 are in , , and or , respectively , and their slopes in the contour plot are @xmath230 , @xmath326 , and @xmath230 , respectively .
this is because @xmath166 increases proportionally to @xmath327 in , while it increases proportionally to @xmath328 in the other cases .
regions @xmath329 , @xmath330 , @xmath331 , and @xmath332 are sir - limited operating regimes , ( i.e. , @xmath333 ) .
note that if operation is always snr - limited due to @xmath334 so that @xmath335 , @xmath336 , @xmath337 , and @xmath332 are not defined .
with respect to the ul transmit power , regions @xmath329 , @xmath330 , @xmath331 , and @xmath332 are in @xmath232 , @xmath267 , @xmath269 , and @xmath270 .
since a cooperative operation is meaningless in @xmath232 and @xmath267 ( in which the csi of a randomly selected user is inaccurate even in the nearest bs ) , @xmath338 in @xmath339 and @xmath340 . on the other hand ,
since zf operation can cancel more interference in @xmath269 and @xmath270 ( in which the csi of a randomly selected user becomes accurate in some bss ) , @xmath341 can be achieved .
interestingly , the slopes in the contour plot of mrt operation are @xmath229 both in @xmath342 and @xmath343 , which implies that even if higher power is available in the network using mrt operation , the tradeoff between the qos ( @xmath310 ) and the number of users ( @xmath344 ) is not improved .
however , zf operation shows positive slopes ( specifically , @xmath345 and @xmath230 in @xmath346 and @xmath347 , respectively ) so that higher power is always beneficial for improving the tradeoff . @xmath210
+ + ( effect of the partial association and/or the pilot reuse ) the effect of the partial association on the number of supportable users is shown in fig .
5(a ) when zf operation is employed .
compared to fig 4(c ) , it is shown that the contour plots in regions @xmath348 , @xmath349 , @xmath350 , @xmath351 and @xmath352 are unchanged , but those in regions @xmath353 and @xmath354 are considerably degraded .
it is shown that a partial association poses an upper - limit on the target qos scaling exponent @xmath310 even if additional transmit power is consumed , which is caused by the inevitable interference from non - associated bss .
so , the merits of zf operation are reduced .
in fact , if @xmath355 , zf operation becomes mrt operation . however , if @xmath63 is strictly larger than 0 , zf operation can still provide better tradeoff between the target qos scaling exponent ( @xmath310 ) and the growth rate of the number of supportable users ( @xmath356 ) compared to mrt operation in regions @xmath353 and @xmath354 .
the effect of the pilot reuse on the number of supportable users is shown in fig .
since the scaling results does not change when @xmath357 , the contour lines marked as 0 , 0.2 , and 0.4 are not changed .
however , the contour lines marked as 0.6 , 0.8 , and 1 are severely degraded due to the pilot reuse .
similarly as in the case of the partial association , an upper - limit is posed on the target qos scaling @xmath310 even if additional transmit power is consumed , which comes from the fact that the csi accuracy can not be improved due to the interference caused by the users sharing the same pilots .
the effect of both the partial association and the pilot reuse is shown in fig .
evidently , both practical limitations results in lower target qos scaling .
the contour line marked as 1 is the same as in the case with the partial association only , i.e. , the effect of partial association is dominant so that higher front / backhaul capacity can improve the network performance , while the other contour lines are the same as in the case with the pilot reuse only , i.e. , the effect of pilot reuse is dominant so that more orthogonal pilot resource should be provided for better network performance . + + in fig . 6 ,
the contour plots for the required @xmath318 on ( @xmath310,@xmath358 ) is illustrated , which shows the tradeoff between the target qos scaling exponent ( @xmath310 ) and the growth rate of the number of supportable users ( @xmath359 ) . by comparing fig .
6(a ) with fig .
6(b ) , it is shown that the pilot reuse , @xmath360 , does not change the tradeoff if @xmath358 is less than @xmath65 . however , the pilot reuse degrades the tradeoff if @xmath358 is larger than @xmath65 so that a non - achievable region is created even if sufficiently high transmit power is available .
thus , in order to improve the tradeoff under the pilot resource constraint , an advanced technique is necessary to avoid or suppress the interference caused from the pilot reuse . also , by comparing fig .
6(a ) with fig .
6(c ) , it is shown that the partial association also degrades the performance and it creates a non - achievable region for high @xmath310 due to the interference mainly caused from non - associated bss .
thus , a network designer needs to carefully design the association range and the front / backhaul capacity level . @xmath210
this paper presented a comprehensive and rigorous asymptotic analysis on the performance of the large - scale cloud radio access network ( ls - cran ) under the three practical constraints , 1 ) limited transmit power , 2 ) limited front / backhaul capacity , and 3 ) limited pilot resource . as a main performance measure ,
the scaling exponent of the signal - to - interference - plus - noise ratio ( sinr ) was defined and derived as a function of key network parameters , 1 ) the number of bs , @xmath361 , 2 ) the number of bs antennas @xmath362 , 3 ) the number of single - antenna users @xmath363 , 4 ) the uplink ( ul ) transmit power , @xmath364 , 5 ) the dl transmit power , @xmath365 , and the distance - dependent pathloss exponent @xmath66 , when interference - free ( if ) , maximum ratio transmission ( mrt ) , or zero - forcing ( zf ) operation is applied .
then , we show that when mrt or zf operation becomes interference - free , i.e. , order - optimal with the three practical constraints . by considering limited transmit power only
, mrt operation is shown to become interference - free only when the dl transmit power is less then a threshold , but the threshold is too low to be meaningful .
also , zf operation is shown to become interference - free at meaningful dl transmit power so that as higher ul transmit power is provided , more users can be associated and the data rate per user can be increased simultaneously while keeping the order - optimality .
furthermore , when the other two practical constraints are considered together , it is shown that the total front / backhaul overhead of @xmath0 and @xmath1 pilot resources are required to keep the order - optimality .
last , in order to characterize the network - wise performance of the ls - cran , the growth rate of the number of supportable users , @xmath308 , is defined when a requirement on the target qos scaling is given as @xmath310 .
then , the tradeoff between ( @xmath366 ) is derived under the three practical constraints .
the results quantify the achievable tradeoff between them at a given level of total transmit power , which shows that the target qos and the number of users satisfying the qos can grow simultaneously with a nice tradeoff as the network size increases .
it is also shown and quantified that the other two practical constraints , limited front / backhaul capacity and limited pilot resources , may pose an upper - bound on the target qos so that non - achievable tradeoff region is created , which would be a key ingredient providing a guideline for next - generation ls - crans .
here , we provide key preliminary results to prove theorems in this paper . first , we state the method to derive the scaling exponent of a random measure given by a sum of infinitely many i.i.d . random variables . * _ lemma 1_*. let @xmath367 be a sequence of i.i.d . real - valued random variables with common distribution function of @xmath368 and @xmath369 be a non - negative integrable function .
then , as @xmath370 , we have @xmath371 where @xmath372 is an inverse ( or quantile ) function of @xmath368 and @xmath373 is the probability density function .
suppose that @xmath374 are i.i.d .
uniform random variables on @xmath375 $ ] .
then , for any positive small @xmath28 , the following holds : @xmath376 which implies that all of @xmath377 are included in the interval @xmath378 $ ] , but not included in the interval @xmath379 $ ] in probability . by dividing the interval
@xmath378 $ ] into @xmath380 intervals , @xmath381,{{\cal s}_2 } = ( { n^{{u_1}}},{n^{{u_2 } } } ] , \cdots , { { \cal s } _ { \lceil g(n ) \rceil } } = ( { n^{{u _ { \lceil g(n ) \rceil - 1}}}},1]$ ] , where @xmath382 for @xmath383 , @xmath384 denotes the ceiling operation , and @xmath31 is a positive function of @xmath385 .
let @xmath386 for @xmath387 .
then , for @xmath388 , @xmath389 since @xmath390 are included in the interval @xmath378 $ ] in probability , we have @xmath391 where @xmath31 is assumed to be properly chosen so that the second equality holds as a riemann summation .
then , it is seen that @xmath31 should increase logarithmically , i.e. , @xmath392 .
now , assume that @xmath369 is a monotonically decreasing non - negative integrable function .
first , we consider the upper - bound : @xmath393 where the last equality comes from the riemann summation .
similarly , we obtain the lower - bound : @xmath394 so that the gap between the lower - bound and the upper - bound is @xmath395 due to @xmath392 .
since the gap is bounded by a constant , the upper - bound and the lower - bound is asymptotically tight .
additionally , @xmath369 is set to a monotonically decreasing ( or increasing ) function so that such a gap is maximized because the gap is still @xmath396 for any non - negative integrable function @xmath369 . finally , by using the continuous mapping theorem @xcite and changing the variable @xmath397
, we obtain ( [ eq_18 ] ) .
recall that our order notation ignores a logarithm term , but the proof of lemma 1 does not ignore this term so that the lemma 1 still holds when we use the order notation in @xcite .
this result is closely related to the ( weak ) law of large number .
obviously , @xmath398 converges to @xmath399 $ ] as @xmath26 if the function @xmath400 is independent to @xmath385 .
if @xmath399 $ ] is finite and non - zero , @xmath401 , which is consistent with the weak law of large number . in the case where @xmath399 $ ] increases unboundedly or approaches zero as @xmath26
, lemma 1 gives the way to quantify its asymptotic behavior via a simple integral form .
usually , it is too hard to derive the integral directly so a simpler way is necessary . by using lemma 1 and our order notation
, we can easily derive the scaling exponent by solving an optimization problem .
* _ corollary 1_*. let @xmath367 be a sequence of i.i.d . real - valued random variables with common distribution function of @xmath368 and @xmath369 be a non - negative integrable function with @xmath402 .
then , as @xmath26 , we have @xmath403 } \left| { { { \cal x}_t } } \right|h\left ( { { n^t } } \right ) } \right ) , \end{split}\ ] ] where @xmath404\}$ ] for two constants @xmath405 , @xmath406 @xmath407 using the similar argument in the proof of lemma 1 , we can easily show that if @xmath402 for @xmath408 , @xmath409 } \sum\limits_{{x_k } \in { { \cal x}_t } } h ( { x_k } ) } \right).\ ] ] since all elements in @xmath410 scales like @xmath411 , @xmath412 can be replaced by @xmath413 , which concludes the proof .
note that , as @xmath26 , all of @xmath414 are included in at least one set of @xmath415 for @xmath416 in the probabilistic sense . for example , if @xmath412 are a sequence of i.i.d .
uniform random variables on @xmath375 $ ] , we can take @xmath417 and @xmath418 .
this corollary informs that the scaling exponent of @xmath419 is only determined by the supremum of the partial sums , @xmath420 , for all possible @xmath30 under some mild condition on @xmath402 for @xmath408 .
since we use the distance - based decay pathloss model , @xmath421 , with @xmath67 , most cases hold this condition so that the scaling exponents obtained in this paper can be derived only by constructing the sets @xmath410 and checking which set results in highest scaling exponent .
the merits of lemma 1 and corollary 1 are shown in the next examples .
_ example 1 .
_ the aggregated interference at the origin of a circular - shaped network of radius @xmath422 can be written as @xmath423 where @xmath424 is a bound and @xmath425 follows a binomial point process on the circle of radius @xmath422 , i.e. , @xmath426 for @xmath427 .
if @xmath428 , the pathloss model is bounded and @xmath429 $ ] converges to a finite constant so that @xmath430 .
however , if @xmath431 , the pathloss model is unbounded at the origin and @xmath432 $ ] diverges for @xmath433 .
since @xmath434 , the function @xmath435 if @xmath431 so corollary 1 can be applied .
after some manipulations , @xmath436 is obtained for @xmath437 .
then , applying corollary 1 , we have @xmath438 } { n^ { - \alpha t } } } \right ) = \theta \left ( { { n^{\frac{\alpha } { 2 } } } } \right).\ ] ] _ example 2 . _ in @xcite ,
the upper - bound of the capacity of a wireless ad - hoc network on @xmath439\times[0 , \sqrt{n}]$ ] is shown as @xmath440 where @xmath441 for an exponential - decaying pathloss model with a positive @xmath442 or @xmath443 for a power - decaying pathloss model and @xmath444 are i.i.d .
uniform random variables on @xmath445 $ ] , i.e. , @xmath446 for @xmath447 $ ] . from lemma 1 , we obtain @xmath448 where @xmath449 .
note that the last equality in ( [ eq_48_a ] ) comes from that fact @xmath450 where @xmath451 .
for the exponential - decaying pathloss model , @xmath452 by taking @xmath453 as @xmath454 , which is consistent with theorem 2.10 in @xcite and for the power - decaying pathloss model , @xmath455 by taking @xmath456 , as @xmath454 , which is consistent with theorem 2.5 in @xcite . *
_ lemma 2_*. let @xmath457 and @xmath458 be two independent homogeneous ppps over a finite region @xmath459 with densities @xmath460 and @xmath461 for @xmath462 , respectively .
assume that an arbitrarily chosen point @xmath463 is given .
then , the following holds for @xmath464 , \1 ) for @xmath465 and @xmath466 , as @xmath467 , @xmath468 \2 ) for @xmath469 , as @xmath470 , @xmath471 where @xmath24 is given by @xmath472 for @xmath473 , or @xmath474 for @xmath475 , and @xmath476 where @xmath30 is given by @xmath477 for @xmath478 and @xmath479 .
\1 ) to prove the first part of lemma 2 , we construct the lower - bound and upper - bound as follows .
@xmath480 here , the lower - bound is obviously @xmath481 because the minimum distance is @xmath482 . to complete the proof of the first part , it is sufficient to prove @xmath483 . let @xmath484 . for a randomly selected @xmath485 ,
its contact distribution function is given by @xmath486 where @xmath487 and @xmath488 denotes the lebesgue measure ( i.e. area measure ) . since @xmath489 is finite and independent to @xmath490 , @xmath491 and thus @xmath492 , where @xmath493 is independent to @xmath490 .
from lemma 1 , we have @xmath494 where the last equality comes from @xmath67 and the fact @xmath495 completes the proof of the first part .
\2 ) define two sets @xmath496 and @xmath497 for a given point @xmath498 . since @xmath457 and @xmath458 are ppps with density @xmath460 and @xmath17 , respectively , @xmath499 and @xmath500 .
note that @xmath501 if and only if @xmath502 and @xmath503 if and only if @xmath504 .
assume that @xmath505 is given with @xmath506 and @xmath507 .
for any @xmath508 , @xmath509 so that @xmath510
. then , we have @xmath511 with the constraints of @xmath512 and @xmath513 .
similarly , we have @xmath514 with the constraints of @xmath515 and @xmath513 .
then , the second part is easily given by obtaining the supremums of ( [ eq_51 ] ) and ( [ eq_52 ] ) under the above constraints .
since the interference is removed by the genie without any cost , @xmath516 .
inserting @xmath517 into ( [ received_signal_from_bs_l ] ) , @xmath518 and the square of which is given by @xmath519 where the first asymptotic equality comes from the fact that the cross term is asymptotically ignorable and the last equality comes from inserting ( [ estimated_csi ] ) .
define @xmath520 as the set of bss sufficiently near user @xmath55 , given by @xmath521 since @xmath522 , @xmath523 , which means that @xmath520 is a non - empty set if and only if @xmath524 and @xmath525 if @xmath526 .
then , we have @xmath527 _ case _ ( @xmath526 ) : in this case , @xmath528 and @xmath529 . then , we have @xmath530 where the second asymptotic equality comes from lemma 2 and the last asymptotic equality comes from the fact that the first term is always dominant for @xmath526 .
_ cases and _ ( @xmath531 ) : in this case , @xmath532 and @xmath533 , we have @xmath534 where the second asymptotic equality comes from lemma 2 and the third asymptotic equality comes from the fact that the first term is only dominant if @xmath535 . _ case _ ( @xmath536 ) : the scaling is simply derived by combining the results in ( [ eq_38 ] ) and ( [ eq_39 ] ) so that it is given by @xmath537 define an auxiliary variable @xmath538 such that @xmath539 for all @xmath55 .
the final step is to find the scaling exponent of the dl transmit power for user @xmath55 , which is given by @xmath540 by using @xmath541 for all @xmath55 , and the results ( [ eq_38 ] ) , ( [ eq_39 ] ) , and ( [ eq_40 ] ) , we obtain ( [ if_snr_scaling ] ) .
inserting @xmath542 into ( [ received_signal_from_bs_l ] ) , @xmath543 , and the square of which is given by @xmath544 where the first asymptotic equality comes from the fact that the cross term is asymptotically ignorable and the third equality is obtained by inserting ( [ estimated_csi ] ) .
note that @xmath545 so that the snr scaling exponent of mrt operation is the same as that of if operation .
the interference power , denoted as @xmath546 , can be written as @xmath547 where @xmath548 is defined in ( [ csi_accuracy_set ] ) .
define an auxiliary variable @xmath538 such that @xmath549 for all @xmath55 and consider the four regimes , , , and . _ case _ ( @xmath526 ) : in this case , @xmath550 and @xmath551 for all @xmath117 .
then , we have @xmath552 where the second asymptotic equality comes from lemma 2 and the third asymptotic equality comes from the fact that the first term is always dominant due to @xmath553
. _ cases and _ ( @xmath531 ) : in this case , @xmath554 and @xmath555 .
then , we have @xmath556 where the second equality comes from lemma 2 and the third asymptotic equality comes from the fact that the first term is always dominant due to @xmath242 .
_ case _ ( @xmath536 ) : the scaling exponent can be simply derived as @xmath557 because the result in ( [ eq_48 ] ) at @xmath558 is the same to the result in ( [ eq_47 ] ) . thus , by combining ( [ eq_47])-([eq_49 ] ) , the scaling exponent of the interference power is given by @xmath559 and combining it with ( [ if_snr_scaling ] ) and ( [ eq_42 ] ) results in ( [ mrt_sir_scaling ] ) , which completes the proof .
inserting @xmath560 into ( [ received_signal_from_bs_l ] ) , @xmath561_{k,:}}{{\bf{f}}^{zf}}{{\bf{e}}_j}$ ] , where @xmath562 denotes the @xmath117th column of the identity matrix @xmath563 , and the square of which is given by @xmath564}_{k,:}}{{\bf{f}}^\mathsf{zf}}{{\bf{e}}_j } } \right\ } + { \rm{tr}}\left ( { { { \left ( { { { \bf{f}}^\mathsf{zf } } } \right)}^h}{{\bf{r}}_k}{{\bf{f}}^{zf}}{{\bf{e}}_j}{\bf{e}}_j^h } \right ) \\ & \asymp \delta ( k - j ) + { \rm{tr}}\left ( { { { \left ( { { { \bf{f}}^\mathsf{zf } } } \right)}^h}{{\bf{r}}_k}{{\bf{f}}^\mathsf{zf}}{{\bf{e}}_j}{\bf{e}}_j^h } \right ) , \end{split}\ ] ] where @xmath565_{k , : } ^h { { [ { \widetilde { \bf{g } } } ] } _ { k , : } } $ ] and the last asymptotic equality comes from the fact that @xmath566 . also , the dl transmit power consumed for user @xmath55 is given by @xmath567 note that we have @xmath568_{i , j } } & = \sum\limits_{{x_l } \in { \cal x } } { \sqrt { { \beta _ { li}}{\beta _ { lj } } } { \bf{\widehat { h}}}_{li}^h{{{\bf{\widehat { h}}}}_{lj}}}\\ & = \left\ { { \begin{array}{*{20}{l } } { m\sum\limits_{{x_l } \in { \cal x } } { \frac{{p_j^{{\rm{ul}}}\beta _ { lj}^2}}{{p_j^{{\rm{ul}}}{\beta _ { lj } } + 1 } } } , } & { { \text{if } } i = j,}\\ { \sqrt m \sum\limits_{{x_l } \in { \cal x } } { \sqrt { \frac{{p_i^{{\rm{ul}}}\beta _ { li}^2}}{{p_i^{{\rm{ul}}}{\beta _ { li } } + 1}}\frac{{p_j^{{\rm{ul}}}\beta _ { lj}^2}}{{p_j^{{\rm{ul}}}{\beta _ { lj } } + 1 } } } } , } & { { \text{if } } i \ne j , } \end{array } } \right .
\end{split}\ ] ] and consider the four regimes . _ case _ ( @xmath526 ) : in this case , @xmath569 so that @xmath570 and @xmath571_{i , j } } \asymp \left\ { { \begin{array}{*{20}{c } } { m\sum\limits_{{x_l } \in { \cal x } } { { \beta _ { lj } } } , } & { { \text{if } } i = j,}\\ { \sqrt m \sum\limits_{{x_l } \in { \cal x } } { \sqrt { { \beta _ { li}}{\beta _ { lj } } } } , } & { { \text{if } } i \ne j , } \end{array } } \right . \end{split}\ ] ] inserting ( [ zf_eh2 ] ) into ( [ zf_psi ] ) , we have @xmath572 and @xmath573 where the last asymptotic equality comes from lemma 2 and @xmath526
. defining an auxiliary variable @xmath538 such that @xmath574 , the dl transmit power can be written as @xmath575 for a reasonable dl transmit power @xmath576 , @xmath577 and thus @xmath578 _ case _ ( @xmath579 ) : define @xmath580 . since @xmath581 , where @xmath582 , if @xmath583 , or @xmath584 , if @xmath585 , and @xmath586_{i , j}}\asymp \left\ { { \begin{array}{*{20}{l } } { m\left ( { \sum\limits_{{x_l } \in { { \cal l}_j } } { { \beta _ { lj } } } + { { p_j^{{\rm{ul}}}}}\sum\limits_{{x_l } \in { { \cal x}_j}\backslash { { \cal l}_j } } { \beta _ { lj}^2 } } \right),}&{{\text{if } } i = j,}\\ { \sqrt m \left ( \begin{array}{l } \sum\limits_{{x_l } \in { { \cal l}_i}\bigcap { { { \cal l}_j } } } { \sqrt { { \beta _ { li}}{\beta _ { lj } } } } + \sqrt { { { p_j^{{\rm{ul } } } } } } \sum\limits_{{x_l } \in { { \cal l}_i}\backslash { { \cal l}_j } } { \sqrt { { \beta _ { li}}\beta _ { lj}^2 } } \\ + \sqrt { { { p_i^{{\rm{ul } } } } } } \sum\limits_{{x_l } \in { { \cal l}_j}\backslash { { \cal l}_i } } { \sqrt { \beta _ { li}^2{\beta _ { lj } } } } + { { \sqrt { p_i^{{\rm{ul}}}p_j^{{\rm{ul}}}}}}\sum\limits_{{x_l } \in { { \cal x}_i } \cap { { \cal x}_j}\backslash { { \cal l}_i } \cup { { \cal l}_j } } { \sqrt { \beta _ { li}^2\beta _ { lj}^2 } } \end{array } \right),}&{{\text{if } } i \ne j. } \end{array } } \right . \end{split}\ ] ] then , @xmath587 is given as in ( [ eq_90 ] ) , where the last asymptotic equality comes from the fact that @xmath588 @xmath589 here , @xmath590 , @xmath591 , @xmath592 , and @xmath593 can be further derived from lemma 2 as follows .
suppose that @xmath594 and @xmath595 . in this case ,
@xmath596 for all @xmath597 in probability so that , by using lemma 2 , we obtain @xmath598 since @xmath599 and @xmath600 , @xmath592 becomes always dominant and thus @xmath601 now , suppose that @xmath594 and @xmath602 . in this case , @xmath603 if @xmath604 and @xmath605 if @xmath606 for all @xmath607 in probability so that , by using lemma 2 , we obtain @xmath608 since @xmath609 , @xmath590 or @xmath592 becomes dominant and thus @xmath610 when @xmath611 , the similar derivation can be done and we obtain @xmath610 also , @xmath612 where the second and last asymptotic equalities come from lemma 2 and the fact @xmath192 , respectively . since the dl transmit power can be written as @xmath613 so that @xmath577 for a reasonable dl transmit power , @xmath614 and @xmath615 _ cases and _ ( @xmath616 ) : in these cases , @xmath617 , and @xmath618_{i , j } } \asymp \left\ { { \begin{array}{*{20}{l } } { { { mp_j^{{\rm{ul}}}}}\sum\limits_{{x_l } \in { { \cal x}_j } } { \beta _ { lj}^2 } , } & { { \text{if } } i = j,}\\ { { { \sqrt { mp_i^{{\rm{ul}}}p_j^{{\rm{ul } } } } } } \sum\limits_{{x_l } \in { { \cal x}_i}\bigcap { { { \cal x}_j } } } { \sqrt { \beta _ { li}^2\beta _ { lj}^2 } } , } & { { \text{if } }
i \ne j , } \end{array } } \right . \end{split}\ ] ] inserting ( [ ml_2 ] ) into ( [ zf_psi ] ) , we have @xmath619 which is obtained in a similar way as before as @xmath620 and @xmath621 since the dl transmit power can be written as @xmath622 so that @xmath623 for a reasonable dl transmit power , we can obtain @xmath624 and @xmath625 which concludes the proof .
since the total transmit power is given by @xmath626 by definition , in order to guarantee @xmath627 , @xmath628 and @xmath629 at the power constraints of @xmath630 and @xmath631 .
formally , in order to find the supportable user scaling exponent , we should solve the following optimization problem ( p1 ) : @xmath632 one important and straightforward observation is that more ul or dl transmit power results in higher snr and sir scaling exponents .
so , the maximum of @xmath61 is obtained when @xmath633 inserting ( [ power_condition ] ) into the snr and sir scaling exponents of each operation , we can readily derive the conditions @xmath634 and @xmath635 , and then the supportable user scaling exponent is determined within the intersection of the two conditions . since the proof for the results for one region is similar to those for other regions , the proof for region @xmath351
is shown only for brevity . from theorems 1 and 3 , the snr scaling exponent of zf operation is identical that of if operation , and by inserting ( [ power_condition ] ) into ( [ zf_sir_scaling ] ) , the conditions for the sir scaling exponents are given by @xmath636 then , it is sufficient to show that the above conditions hold when inserting @xmath637 into ( [ zf_condition1 ] ) and ( [ zf_condition2 ] ) . for the case
@xmath638 , inserting @xmath639 into ( [ zf_condition1 ] ) and ( [ zf_condition2 ] ) yields @xmath640 where the first equality comes from the fact that @xmath641 and the second equality comes from the fact that @xmath642 .
since it is easily seen that the conditions are violated for larger @xmath637 , the proof for region @xmath643 is completed .
1 j. g. andrews et al .
, `` what will 5 g be ?
_ ieee j. sel .
areas commun .
32 , pp . 10651082 , june 2014 . n. bhushan , j. li , d. malladi , r. gilmore , d. brenner , a. damnjanovic , r. t. sukhavasi , c. patel , and s. geirhofer , `` network densification : the dominant theme for wireless evolution into 5 g , '' _ ieee commun . mag .
2 , pp . 8289 , feb . 2014 .
f. rusek , d. persson , b. k. lau , e. g. larsson , t. l. marzetta , o. edfors , and f. tufvesson , `` scaling up mimo : opportunities and challenges with very large arrays , '' _ ieee signal process . mag .
_ , vol . 30 , no .
1 , pp . 4046 , jan .
a. gotsis , s. stefanatos , and a. alexiou , `` ultra dense networks : the new wireless frontier for enabling 5 g access , '' _ ieee veh .
technol . mag .
2 , pp . 7178 , june 2016 . n. q. ngo , e. g. larsson , and t. l. marzetta , `` energy and spectral efficiency of very large multiuser mimo systems , '' _ ieee trans .
4 , pp . 14361449 , apr . 2012 .
i. hwang , b. song , and s. soliman , `` a holistic view on hyper - dense heterogeneous and small cell networks , '' _ ieee commun .
51 , pp . 2027 , june 2013 . j. hoydis , m. kobayashi , and m. debbah , `` green small - cell networks , '' _ ieee veh .
technol . mag .
6 , pp . 3743 , mar .
h. huang , m. trivellato , a. hottinen , m. sha , p. smith , and r. valenzuela , `` increasing dl cellular throughput with limited network mimo coordination , '' _ ieee trans .
wireless commun .
8 , pp . 29832989 , june 2009 .
r. irmer , h. droste , p. marsch , m. grieger , g. fettweis , s. brueck , h .-
mayer , l. thiele , and v. jungnickel , `` coordinated multipoint : concepts , performance , and field trial results , '' _ ieee commun . mag .
2 , pp . 102111 , feb . 2011 .
a. checko et al . , `` cloud ran for mobile networks a technology overview , '' _ ieee commun .
17 , no . 1 , 1 quarter 2015 , pp .
405426 g. caire and s. shamai , `` on the achievable throughput of a multiantenna gaussian broadcast channel , '' _ ieee trans .
inf . theory _ , vol .
16911706 , july 2003 .
i. chih - lin et al . , `` toward green and soft : a 5 g perspective , '' _ ieee commun . mag .
2 , pp . 6673 , feb . 2014 .
e. bjrnson , j. hoydis , m. kountouris , and m. debbah , `` massive mimo systems with non - ideal hardware : energy efficiency , estimation , and capacity limits , '' _ ieee trans .
inf . theory _ ,
71127139 , nov .
t. biermann , l. scalia , c. choi , w. kellerer and h. karl , `` how front / backhaul networks influence the feasibility of coordinated multipoint in cellular networks , '' _ ieee commun . mag .
168176 , aug .
2013 .
q. h. spencer , a. l. swindlehurst , and m. haardt , `` zero - forcing methods for dl spatial multiplexing in multiuser mimo channels , '' _ ieee trans .
signal process .
2 , pp . 461471 , feb .
2004 .
d. a. basnayaka , p. j. smith , and p. a. martin , `` performance analysis of dual - user macrodiversity mimo systems with linear receivers in flat rayleigh fading , '' _ ieee trans
. wireless commun .
43944404 , dec . 2012 .
d. a. basnayaka , p. j. smith , and p. a. martin , `` performance analysis of macrodiversity mimo systems with mmse and zf receivers in flat rayleigh fading , '' _ ieee trans .
wireless commun .
5 , may 2013 .
d. b. cheikh , j .-
m . kelif , m. coupechoux , and p. godlewski , `` analytical joint processing multi - point cooperation performance in rayleigh fading , '' _ ieee wireless commun .
_ , vol . 1 , no
. 4 , aug . 2012 .
r. tanbougi , s. singh , j. g. andrew , and f. k. jondral `` a tractable model for noncoherent joint - transmission base station cooperation , '' _ ieee trans .
wireless commun .
_ , vol . 13 , no .
9 , pp . 49594973 , sep . 2014
. g. nigam , p. minero , and m. haenggi , `` coordinated multipoint joint transmission in heterogeneous networks , '' _ ieee trans .
41344146 , nov .
a. giovanidis and f. baccelli , `` a stochastic geometry framework for analyzing pairwise - cooperative cellular networks , '' _ ieee trans .
wireless commun .
2 , pp . 794808 , feb . 2015 .
k. huang , and j. g. andrews , `` an analytical framework for multicell cooperation via stochastic geometry and large deviations , '' _ ieee trans .
inf . theory _ ,
4 , pp . 25012516 , apr . 2013 .
n. lee , f. baccelli , and r.w .
heath , `` spectral efficiency scaling laws in dense random wireless networks with multiple receive antennas , '' _ ieee trans . inf .
theory _ , vol .
13441359 , mar .
p. marsch and g. fettweis , `` ul comp under a constrained backhaul and imperfect channel knowledge , '' _ ieee trans .
wireless commun_. vol .
10 , no . 6 , pp . 17301742 , june 2011 . q. wang , and y. jing , `` performance analysis and scaling law of mrc / mrt relaying with csi error in massive mimo systems , '' _ submitted to ieee trans .
wireless commun .
_ , june 2016 .
[ online ] .
available : http://arxiv.org/pdf/1606.07480v1.pdf . c. kong , c. zhong , m. matthaiou , e. bjrnson , and z. zhang , `` multipair two - way half - duplex relaying with massive arrays and imperfect csi , '' _ submitted to ieee trans .
inf . theory _
, july 2016 , [ online ] .
available : http://arxiv.org/pdf/1607.01598v1.pdf .
p. xia , h .- s .
jo , and j. andrews , `` fundamentals of inter - cell overhead signaling in heterogeneous cellular networks , '' _ ieee j. sel .
topics sig .
_ , vol . 6 , no .
3 , pp . 257269 , june 2012 . o. simeone , o. somekh , h. v. poor and s. shamai , `` local base station cooperation via finite - capacity links for the uplink of linear cellular networks , '' _ ieee trans inf .
1 , pp . 190204 , jan .
o. simeone , o. somekh , h. v. poor , and s. shamai , `` downlink multicell processing with limited - backhaul capacity , '' _ eurasip j. advances in signal process .
2009 , feb . 2009 .
m. liu , y. teng , and m. song , `` performance analysis of coordinated multipoint joint transmission in ultra - dense networks with limited backhaul capacity , '' _ ieee j. sel .
areas commun .
10 , pp . 21112113 , oct . | large - scale cloud radio access network ( ls - cran ) is a highly promising next - generation cellular network architecture whereby lots of base stations ( bss ) equipped with a massive antenna array are connected to a cloud - computing based central processor unit via digital front / backhaul links .
this paper studies an asymptotic behavior of downlink ( dl ) performance of a ls - cran with three practical constraints : 1 ) limited transmit power , 2 ) limited front / backhaul capacity , and 3 ) limited pilot resource . as an asymptotic performance measure , the scaling exponent of the signal - to - interference - plus - noise - ratio ( sinr )
is derived for interference - free ( if ) , maximum - ratio transmission ( mrt ) , and zero - forcing ( zf ) operations .
our asymptotic analysis reveals four fundamental operating regimes and the performances of both mrt and zf operations are fundamentally limited by the ul transmit power for estimating user s channel state information , not the dl transmit power .
we obtain the conditions that mrt or zf operation becomes interference - free , i.e. , order - optimal with three practical constraints . specifically , as higher ul transmit power is provided , more users can be associated and the data rate per user can be increased simultaneously while keeping the order - optimality as long as the total front / backhaul overhead is @xmath0 and @xmath1 pilot resources are available .
it is also shown that how the target quality - of - service ( qos ) in terms of sinr and the number of users satisfying the target qos can simultaneously grow as the network size increases and the way how the network size increases under the practical constraints , which can provide meaningful insights for future cellular systems . |
understanding physical aging in glasses is essential at a technological level , for controlling the material properties and their temporal evolution , as well as at a fundamental level , since glasses are often considered as archetypes for systems far from thermodynamical equilibrium @xcite .
however , and despite the strong efforts done in the field , a complete picture of the dynamics in the glassy state is still missing .
the main difficulty arises from the complexity of the glassy properties .
when a supercooled liquid is cooled below its glass transition temperature , @xmath3 , the system falls out of equilibrium in a metastable state which depends on the previous thermal history , and which evolves , with waiting time , through different states toward the corresponding supercooled equilibrium liquid phase .
the existence of a multitude of different metastable glassy states which can be explored through different annealing procedures or cooling rates , makes difficult the development of a universal theory for glasses or even a plain comparison between different experimental results .
+ physical aging in polymeric , metallic or more conventional inorganic glasses , is commonly studied by looking at the temporal evolution of a given observable , keeping fixed other external parameters , such as the temperature and the pressure of the system .
macroscopic quantities like viscosity , refractive index , volume or elastic constants , are usually the subject of this kind of studies @xcite . in these works ,
the investigated quantities evolve with waiting time , @xmath4 , in a way which depends on the previous thermal history and which can be accurately described by stretched exponential functions @xcite .
similar results have been reported also in studies of aging effect on the intensity of the high frequency tail of the dielectric spectra @xcite .
all these works provide information on the equilibration toward the supercooled liquid phase , but they do not give any microscopic details on the aging dynamics in the glassy state .
this information can instead be obtained by the investigation , on a microscopic scale , of the temporal evolution of the structural relaxation time , @xmath5 , which represents the time necessary for the system to rearrange its structure toward a more stable configuration .
its value can be measured , for instance , by monitoring the long time decay of the intermediate scattering function , @xmath6 , on a length scale @xmath7 , with the wave - vector @xmath8 corresponding to the mean inter particle distance .
the intermediate scattering function is related to the density - density correlation function and contains all the information on the relaxation dynamics in the system on a length scale determined by @xmath8 . while the dynamics in the supercooled liquid phase has been widely investigated in the past , the difficulty of probing the particle - level dynamics in realistic molecular glasses both with experiments and simulations has instead strongly delayed progress in the field .
in fact only very few studies exist of the behavior of @xmath5 below @xmath3 @xcite .
numerical simulations on a lennard jones glass show that for small quenching from an equilibrium configuration , @xmath5 slows down linearly or sub linearly with waiting time @xcite in agreement with previous results on polymeric glasses @xcite .
in addition , the long time decay of the correlation functions can be rescaled into a single master curve , leading to the validity of a time - waiting time superposition principle in the glassy state @xcite .
+ recently , the aging of a mg@xmath9cu@xmath10y@xmath11 metallic glass has been studied in both the supercooled liquid phase and the glassy state by looking directly at the evolution of the dynamics at the interatomic length scale @xcite .
the glass transition has been found to be accompanied by i ) a crossover of the intermediate scattering function from the well characterized stretched exponential shape - thus described by a stretching exponent @xmath12 - in the supercooled liquid phase , to an unusual compressed exponential form - defined by a parameter @xmath13 - in the glassy state and ii ) to the presence of two very distinct aging regimes below @xmath3 . in particular , a fast , exponential growth of @xmath5 has been reported for short waiting times , followed by an extremely slow aging regime at larger @xmath4 . while the latter slow regime has been already reported both in numerical simulations @xcite and in experiments on polymeric glasses @xcite , the compressed shape of the correlation functions and the fast aging regime had not been reported before in structural glasses .
two similar aging regimes have been recently observed in a numerical study of a concentrated system of nearly hard spheres @xcite . in this case , however , the correlation functions display a more stretched exponential behavior , and no signature of a compressed shape was reported @xcite .
+ compressed correlation functions and different aging regimes have instead been reported for out of equilibrium soft materials , where the dynamics is controlled more by ballistic - like rather than by diffusive motions @xcite .
the previous work on the metallic glass thus suggests a common nature of the dynamics in metallic glasses and soft materials .
a very similar faster than exponential decay and strong aging behavior for the intermediate scattering function has been observed also in a zr - based metallic glass by leitner and coworkers @xcite . in this case , however , this unusual dynamics has been associated with the onset of a crystallization process .
there are consequently several questions to clarify : i ) are the results found in mg@xmath9cu@xmath10y@xmath11 really a universal feature of metallic glasses or do they depend on the microscopic details of the system ?
ii ) is it possible to find a common description of physical aging for different glasses ?
iii ) what is the effect played by the wave - vector , @xmath8 , on the dynamics ? + here , we answer to these questions by presenting a detailed investigation on physical aging in a zr@xmath0ni@xmath1 metallic glass at the atomic length scale . we find that in this case the dynamics in the glassy state is also characterized by compressed correlation functions and different aging regimes , supporting the idea of a universal mechanism for physical aging in metallic glasses .
this hypothesis is moreover strengthened by the possibility of describing the aging dynamics using the same phenomenological exponential growth and rescaling the results obtained for different systems into a single master curve .
the zr@xmath0ni@xmath1 samples ( @xmath14 k @xcite ) were produced as thin ribbons by melt - spinning at the polytechnic university of catalonia .
the zr and ni pure metals were pre - alloyed by arc - melting under ti - gettered ar atmosphere .
the melt was then fast - quenched with a cooling rate of @xmath15 k / s by injecting it on a cu wheel spinning with a 40 m/s perimeter velocity .
the resulting metallic ribbons have a thickness of 22@xmath164 @xmath17 m , close to the optimal value required to maximize the scattered intensity in the experiments .
the ribbons were then held in a resistively heated furnace covering a temperature range between 270 k and 1000 k with a thermal stability of less than 1 k. the temperature of the sample was held at @xmath18 k , reached by heating from room temperature with a fixed rate of @xmath19 k / min . +
x - ray photon correlation spectroscopy ( xpcs ) measurements were carried out at the beamline id10 at the european synchrotron radiation facility in grenoble , france .
an incident x - ray beam of 8.0 kev ( @xmath20 ) was selected by using a single bounce si(1,1,1 ) crystal monochromator with a bandwidth of @xmath21 .
hard x - rays originating from higher order monochromator reflections were suppressed by a si mirror placed downstream of the monochromator .
a partially coherent beam with @xmath22 photons / s was then obtained by rollerblade slits opened to @xmath23 @xmath17 m , placed @xmath240.18 m upstream of the sample . in order to enhance the scattered intensity ,
speckle patterns were collected in transmission geometry by two ikonm charge - coupled devices from andor technology ( @xmath25 pixels , @xmath26 @xmath27 pixel size each one ) installed in the vertical scattering plane @xmath2470 cm downstream of the sample .
all pixels of the two ccd detectors were considered to belong to the same @xmath8 with a resolution of @xmath28 @xmath29 .
the detectors were mounted on a diffractometer and could rotate around the center of the sample position , covering a @xmath8 range up to several @xmath29 .
xpcs measurements were performed for different wave - vectors at and around the first diffraction peak of the static structure factor which corresponds to @xmath30 @xmath29 , thereby probing directly the temporal relaxation of density fluctuations at the inter - particle distance @xmath31 .
time series of up to @xmath245000 images were taken with 7 s exposure time per frame and were analyzed following the procedure described in ref .
no beam damage was observed and the amorphous structure of the sample was checked during the whole experiment by measuring the corresponding static structure factor of the system .
as previously discussed , physical aging implies a relaxation dynamics which is not the same at all times , but strongly depends on the sample age or waiting time .
the temporal dependence of the dynamics can be entirely captured by xpcs through the determination of the two - time correlation function @xmath32 ( ttcf ) which represents the instantaneous correlation of the intensity @xmath33 at two times @xmath34 and @xmath35 , being : @xmath36 here , the average is performed on all the pixels of the ccd , that correspond to the same wave - vector @xmath8 @xcite .
+ c ni@xmath1 at @xmath37=373 k. data acquisition started at @xmath38=1400 s ( top panel ) and @xmath38=49000 s ( bottom panel ) from temperature equilibration .
the age of the sample at any point in the figure can be taken as @xmath39 s. the huge slow down of the dynamics at the different sample ages is well illustrated by the clear broadening of the intensity around the diagonal from the left bottom corner to the top right corner in the bottom panel with respect to the case corresponding to a much shorter @xmath4 ( top panel).,width=9 ] + ni@xmath1 at @xmath37=373 k. data acquisition started at @xmath38=1400 s ( top panel ) and @xmath38=49000 s ( bottom panel ) from temperature equilibration .
the age of the sample at any point in the figure can be taken as @xmath39 s. the huge slow down of the dynamics at the different sample ages is well illustrated by the clear broadening of the intensity around the diagonal from the left bottom corner to the top right corner in the bottom panel with respect to the case corresponding to a much shorter @xmath4 ( top panel).,width=9 ] figure [ fig1 ] shows the ttcf measured in zr@xmath0ni@xmath1 in the glassy state at two different sample age times from temperature equilibration .
the colors represent the values of @xmath40 .
the diagonal from bottom left to top right corresponds to the time line of the experiment and it thus indicates the sample age .
the width of this diagonal contour is instead proportional to the relaxation time @xmath5 .
the steady slowing down of the dynamics , and thus the increase in the relaxation time due to physical aging , is illustrated by the broadening of the diagonal contour on increasing the sample age .
this effect is clearly visible in the top panel of figure [ fig1 ] where we report data taken between @xmath41 and @xmath42 s after temperature equilibration . after almost 13 hours of annealing ( lower panel of figure [ fig1 ] ) ,
the dynamics was so slow that the intensity along the diagonal covered almost the whole image , despite the time interval was almost the double with respect to the one reported in the top panel .
after this very long annealing the width of the diagonal contour seems to be constant with time , as it would be in the case of stationary equilibrium dynamics . as discussed later , here the effect of physical aging
is probably so slow that the dynamics appears as stationary at least on our experimental time scale .
+ although the dynamics is not stationary , the aging is sufficiently slow that it is possible to find time intervals over which the @xmath32 can be averaged and thus to get the standard intensity autocorrelation function @xmath43 where @xmath44 is the temporal average over all the possible starting times @xmath34 .
the intensity autocorrelation function is related to the intermediate scattering function through the siegert relation @xmath45 , with @xmath46 a setup - dependent parameter @xcite , and its determination allows capturing snapshots of the dynamics at different waiting times . the siegert relation requires the temporal average to be equal to the ensemble average .
this condition is fulfilled for systems at thermodynamic equilibrium .
the extension to non - ergodic systems , such as glasses , is possible thanks to the use of an area detector , like the cdd used during our experiments . in this case in fact
, the intensity distribution is sampled correctly by the average over all the pixels , independently of the nature of the dynamics @xcite .
+ the time interval for temporal averaging was chosen small enough that the shape of @xmath47 was not noticeably affected by the evolution of the dynamics , while keeping a good signal to noise ratio . for each set of images ,
the sample age was defined as @xmath48 , where @xmath38 is the delay of the starting point for the measurement from temperature equilibration , and @xmath49 and @xmath50 are the final and starting times of the interval used for the average . + the @xmath51 relaxation process in glass - formers can be well described by using the empirical kohlrausch williams watt ( kww ) model function @xcite @xmath52\ ] ] in this expression , @xmath5 is the characteristic relaxation time , @xmath2 is the stretching parameter which quantifies the deviation from a simple exponential behavior , and @xmath53 is the nonergodicity factor .
it represents the height of the intermediate - time plateau before the final decay associated to the @xmath51-relaxation and it slightly depends on temperature in the glassy state . +
following the siegert relation , the intensity correlation functions measured with xpcs can then be described by @xmath54 \label{kww}\ ] ] where @xmath55 , being @xmath46 the parameter entering the siegert relation .
+ a selection of normalized intensity autocorrelation functions obtained at @xmath18 k for a fixed @xmath56 and different waiting times is reported in figure [ fig2 ] .
the data are shown together with the best - fit line shapes obtained by using equation .
k for different waiting times from temperature equilibration . on increasing the waiting time , the decay clearly shifts toward longer time scales due to physical aging .
the data are reported together with the resulting fit obtained using equation .,width=9 ] the intensity correlation functions decorrelate completely on the experimental time scale , implying that the system is able to rearrange its structure on the probed spatial length scale of few .
the effect of aging is illustrated by the shift of the decay toward longer time scales upon increasing the sample age , in agreement with the analysis of the two times correlation functions reported in figure [ fig1 ] . as shown in figure [ fig2 ] , aging is so important that the decay time increases of almost a factor 10 in less than 13 hours .
this fast aging is likely to be the reason of the well known rapid embrittlement of metallic glasses @xcite .
+ interestingly , the correlation functions display a compressed , thus faster than exponential behavior , which can be well described by equation with a shape parameter @xmath57 , independent of the sample age .
such a high value is in agreement with the results reported for other metallic glasses @xcite , while it can not be explained within the current theories for the dynamics in the glassy state @xcite .
based on the observations of the dynamics in the liquid above the glass transition temperature , these theories predict a stretched , thus with @xmath58 , decay of the density fluctuations even in the glassy state .
the idea is that the dynamic heterogeneities present in the liquid phase partially freeze below @xmath3 , leading to a narrower distribution of relaxation times , and thus to an increase of the shape exponent , up to the limiting case of @xmath59 .
this explanation well describes the correlation functions calculated in numerical simulation @xcite and the increase of @xmath2 measured in some polymeric glasses @xcite , but can not take into account the behavior reported in figure [ fig2 ] .
+ from the comparison with other systems , it seems then clear that metallic glasses behave in a different universal way , independently of their atomic structure . in the case of mg@xmath9cu@xmath10y@xmath11 ,
the unusual faster than exponential behavior has been attributed to the strain field generated by a random distribution of slowly - evolving sources of internal stresses stored in the system during the quench , and which can then be released upon annealing the system close to @xmath3 @xcite .
this interpretation suggests the existence of a different kind of dynamics , rather than pure diffusion , in metallic glasses , which display many similarities with that of out - of - equilibrium soft materials , such as concentrated colloidal suspensions and nanoparticle probes in a glass former matrix @xcite .
+ in the case of zr@xmath0ni@xmath1 the shape parameter is even larger than the one reported for mg@xmath9cu@xmath10y@xmath11 ( @xmath60 for a similar thermal path ) .
this may suggest the presence of larger atomic mobility or more internal stresses in zr@xmath0ni@xmath1 , a conclusion which however requires further investigations . while the value @xmath60 has been often observed in numerous experimental studies of soft materials and
corresponds to the early - time regime proposed in the model of bouchaud and pitard @xcite , values of @xmath2 as large as those here reported for zr@xmath0ni@xmath1 have been observed only in another zr - based metallic glass @xcite and in few other soft systems , and their interpretation is still far from being reached @xcite .
+ = 373 k for different wave - vectors around the first sharp diffraction peak ( @xmath30 @xmath29 ) .
the inset * b ) * reports a zoom of the data taken at large waiting times .
* c ) * corresponding shape parameter .
* d ) * selection of the correlation curves corresponding to the data reported in * a ) * as a function of the time rescaled by the structural relaxation time obtained from the analysis of the curves with the kww model function of eq . .
the inset * e ) * indicates the position of the different investigated @xmath8s.,width=9 ] the waiting time dependence of the structural relaxation times obtained from the analysis of the @xmath61 is shown in figure [ fig4 ] a ) ( squares ) .
data corresponding to different @xmath8 values around the main peak of the diffraction pattern are also reported .
as explained in the experimental details , the low scattered signal allowed us to measure just one @xmath8 vector for each detector setting .
consequently the different @xmath8s are taken at different @xmath4 from temperature equilibration .
at all the investigated @xmath8s , @xmath5 grows rapidly with the sample age , in agreement with the results found in the mg@xmath9cu@xmath10y@xmath11 glass @xcite .
interestingly , the dynamics is characterized by structural rearrangements which require @xmath62 s thus , relatively short times with respect to the common idea of an almost frozen dynamics in the deep glassy state .
these values are however in perfect agreement with those reported for both the mg - based and zr - based metallic glasses @xcite and suggest the presence of a peculiar dynamical behavior at the atomic length scale .
hence , it would be interesting to investigate the nature of the atomic dynamics in the glassy state through the @xmath8 dependence of the structural relaxation time .
unfortunately , this information would require at each investigated @xmath8 a good knowledge of the aging in a wide @xmath4 range , which can be gained only by measuring the dynamics simultaneously for different wave - vectors . a higher incident flux , as the one available at free electron laser sources ,
will make this kind of studies possible .
+ interestingly , the shape parameter does not display any dependence on @xmath8 or @xmath4 , and remains constant with @xmath63 for all the investigated @xmath8s ( figure [ fig4 ] c ) ) .
this means that it is possible to rescale all the data sets to a single master curve and define a time - waiting time - wave vector superposition principle in the glass .
this scaling is reported in [ fig4 ] d ) where a selection of intensity auto correlation functions corresponding to the data reported in panel a ) are plotted as a function of @xmath64 . in the case of mg@xmath9cu@xmath10y@xmath11 , @xmath2 was found to be temperature and @xmath4 independent in the fast aging regime @xcite .
our results indicate that this parameter is also not affected by the choice of the wave - vector , at least for values of @xmath8 close to that of the maximum of the static structure factor ( see figure [ fig4 ] e ) ) . +
as shown in figure [ fig4 ] a ) and b ) , for @xmath56 and large @xmath4 , the fast aging regime seems to change to a more stationary or very slow one , as also suggested by the analysis of the two times correlation function reported in the bottom panel of figure [ fig1 ] .
the presence of a complex hierarchy of aging regimes has been observed also in mg@xmath9cu@xmath10y@xmath11 but for temperatures very close to the glass transition one ( @xmath65 ) .
the results here reported for @xmath66 show that this second aging regime can be found also in the deep glassy state , after a sufficiently long annealing time . in the case of mg@xmath9cu@xmath10y@xmath11 ,
the second aging regime was accompanied by a decrease in the shape parameter .
as shown in figure [ fig4 ] c ) , in the present case the lack of a full complete decay of @xmath67 for the highest @xmath4 leads to a larger error in the determination of @xmath2 , which can be determined only for @xmath68 s and remains however similar to the value obtained at lower @xmath4 .
+ from the energetic point of view , the abrupt crossover to an almost stationary dynamics suggests the existence of a deep energy minimum in the potential energy landscape , where the system remains trapped for a very long time before eventually jumping into a different configuration @xcite .
a similar plateau has been recently observed in the enthalpy recovery of glassy polymers and has been interpreted as a thermodynamically stable state @xcite .
this explanation can not be extended to the case of metallic glasses .
in fact , the comparison between xpcs and low temperature equilibrium viscosity data for mg@xmath9cu@xmath10y@xmath11 suggests instead a gradual approach toward equilibrium @xcite , in agreement with the sub linear aging regime found in other polymeric glasses and colloids @xcite . + following the procedure reported in ref .
@xcite , we can describe the short @xmath4 behavior of @xmath69 using the phenomenological expression @xmath70 where @xmath71 is the value of @xmath5 for @xmath72 at the temperature equilibration time , and @xmath73 is a fitting parameter which describes the rate growth and is independent of temperature @xcite .
a fast exponential growth with @xmath4 is evident also at the other investigated @xmath8s .
however , in that cases the lack of information on the behavior at shorter @xmath4 does not allow for a more quantitative description of the aging dependence for @xmath5 .
+ cu@xmath10y@xmath11 at @xmath65 with different thermal histories and in zr@xmath0ni@xmath1 at @xmath66 .
all the data are normalized for the @xmath74 parameter obtained from the analysis of the fast aging with the empirical model function and are reported as a function of the time rescaled by the @xmath73 parameter .
the data for mg@xmath9cu@xmath10y@xmath11 are taken from ref .
@xcite.,width=9 ] for @xmath56 we find @xmath75 s and @xmath76 s. the latter value is quite close to the one reported in ref .
@xcite for mg@xmath9cu@xmath10y@xmath11 at @xmath65 and it is of the same order of magnitude of that previously reported in the case of colloidal gels and laponite @xcite , thus strengthening the idea of a universal out - of - equilibrium behavior .
the comparison with the mg - based metallic glass suggests that the physical aging is very similar for both systems and that it should be possible to rescale the rapid growth of @xmath77 to a unique curve , independently on the microscopic details of the system .
this scaling is shown in figure [ fig5 ] , where the structural relaxation times of zr@xmath0ni@xmath1 at @xmath66 are compared to those measured in mg@xmath9cu@xmath10y@xmath11 at @xmath65 and for two different thermal paths .
while squares and circles correspond to as - quenched samples heated up to a given temperature , stars refer to a state reached after having partially annealed the glass at higher temperatures without equilibration @xcite .
all the data are rescaled for the corresponding @xmath74 parameter obtained from the analysis of the fast aging using equation and are reported as a function of the time rescaled by the @xmath73 parameter . for short @xmath4 ,
all the data fall into a unique master curve , suggesting a common origin of the fast aging regime in metallic glasses .
differently , the crossover to the subsequent almost stationary dynamics strongly depends on the temperature and thermal path , and is reached only for very long annealing times ( circles ) or after particular thermal treatments ( stars ) .
in conclusion , we have studied the physical aging in a zr@xmath0ni@xmath1 metallic glass , by following the temporal evolution of the structural relaxation process at the interatomic length scale .
we find that the decay of the density fluctuations can be well described by compressed , thus faster than exponential correlation functions , suggesting a peculiar origin for the atomic dynamics @xcite .
this behavior seems to be universal in metallic glasses @xcite , and can not be explained by the well - known theories for glasses dynamics , which predict a stretched exponential shape of the correlation functions , even in the glassy state @xcite .
model systems such as hard spheres and lennard - jones are unable to reproduce these intriguing dynamic effects @xcite . in these model systems , indeed , the dynamics appears to be always characterized by a broad spectrum of relaxations .
conversely , very similar results have been previously reported for out of equilibrium soft materials , suggesting a universal microscopic dynamics for these very different systems @xcite . in these systems ,
the compressed behavior arises from ballistic - like motion due to the presence of internal stresses in the out of equilibrium state @xcite . it would be interesting to check whether this is also the case for metallic glasses
however , a @xmath8 dependence study would require a much stronger signal than the one achievable even at a third generation synchrotron source as the one used for the present study .
notwithstanding , indication of the presence of distinct processes of atomic motion in metallic glasses have been reported for different zr - based metallic glass - formers studied with nuclear magnetic resonance @xcite . in these systems the diffusive , collective motion is found to be accompanied by hopping which becomes dominant in the glassy state , and could then be related to the observed compressed relaxation process in our glass .
+ albeit we could not get information on the @xmath8 dependence of the relaxation time due to the impossibility of measuring simultaneously different @xmath8s , we have investigated the nature of the dynamics at different wave - vectors and sample ages .
we find that the shape of the correlation functions does not show any dependence neither on the sample age nor on the wave - vector , at least for @xmath8s around the maximum of the static structure factor of the system , which is at @xmath30 @xmath29 .
these findings allow then to rescale all the correlation curves on the top of each other and define a time - waiting time - wave - vector superposition principle in the fast aging regime .
+ the value of @xmath2 seems to depend on the details of the system , being larger in zr - based systems than in the previously reported mg@xmath9cu@xmath10y@xmath11 @xcite .
as both these glasses were obtained with the same thermal protocol , we speculate that the observed difference could be related to a larger atomic mobility in the first system and to the consequent presence of larger internal stresses . despite this small difference ,
the comparison between the two metallic glasses suggests the existence of a universal dynamical behavior , also indicated by the scaling of the fast aging regime for the relaxation time to a single curve , independently of the microscopic details of the systems .
additional theoretical and experimental works is clearly required to elucidate this remarkable dynamical behavior of metallic glasses .
+ + we acknowledge f. zontone , h. vitoux , k. lhoste , l. claustre for technical support during the xpcs experiments and e. pineda an p. bruna for providing the samples . | we present x - ray photon correlation spectroscopy measurements of the atomic dynamics in a zr@xmath0ni@xmath1 metallic glass , well below its glass transition temperature .
we find that the decay of the density fluctuations can be well described by compressed , thus faster than exponential , correlation functions which can be modeled by the well - known kohlrausch - williams - watts function with a shape exponent @xmath2 larger than one .
this parameter is furthermore found to be independent of both waiting time and wave - vector , leading to the possibility to rescale all the correlation functions to a single master curve .
the dynamics in the glassy state is additionally characterized by different aging regimes which persist in the deep glassy state .
these features seem to be universal in metallic glasses and suggest a non diffusive nature of the dynamics .
this universality is supported by the possibility of describing the fast increase of the structural relaxation time with waiting time using a unique model function , independently of the microscopic details of the system . \\ |
the reconstruction of an unknown quantum state known as quantum tomography is a fundamental task in quantum information science , where a myriad of new technologies promise to exploit special features of quantum states in order to enhance communication , metrology , and computation .
since the quantum state represents maximal information about a physical system , all physical properties can be calculated from it .
checking for the existence for a highly entangled state , a state which can violate a bell inequality , or even the initial state required in a gate - based quantum computer are thus just some examples of the importance of inferring the quantum state from laboratory data . as experimental methods progress , the complexity of quantum systems that can be well controlled in the laboratory grows . in recent times , for example , various groups have been able to demonstrate quantum control of a high number of qubits @xcite .
to gain an idea of the challenge of state reconstruction , one need only consider that the number of real parameters required to describe the joint state of @xmath0 qubits scales as order @xmath1 . alternatively , the orbital angular momentum of single photons , for example , is a single degree of freedom with a large amount of internal structure : it has recently been characterised via the reconstruction of a 100,000 dimensional statevector @xcite . quite apart from the challenges presented by preparation and measurement of quantum states , tackling the state reconstruction problem in the face of such complexity calls for sophisticated data processing techniques , which are the focus of this paper .
[ [ section ] ] tomography experiments involve subjecting a system , described by some unknown quantum state , to a well - defined measurement procedure and recording the measurement outcome . the central tenets of quantum theory place severe restrictions on one s ability to characterise an unknown quantum system given measurements made on only a single copy .
one assumes , therefore , access to a large but finite number of copies of a system , all prepared in an identical quantum state . as the complexity of the quantum state grows , the number of detector counts for each state parameter necessarily shrinks . particularly in optical systems ,
the prevalence of low detection efficiency exacerbates this problem , leaving the tomographer with a noisy data - set from which to make her inferences . in an idealised situation where all noise ( including statistical ) is absent
, the true state @xmath2 can be found exactly .
here we concentrate on the more realistic situation , and assume only that the measurement procedure is informationally complete ( that is to say , the measurement record contains a nonzero amount of information available about each quantum state parameter ) , and turn our attention to the question of processing this data toward the most likely estimate of the unknown state .
[ [ section-1 ] ] in most cases , quantum tomographers must employ numerical techniques to search for the best estimate possible , given the data that has been collected . in this work ,
we analyse the algorithmic method of projected gradient descent as applied to quantum tomography , and benchmark it against a number of existing methods .
the state - of - the - art with regards to full quantum tomography methods include the diluted iterative algorithm ( dia ) @xcite and convex programming @xcite .
both methods benefit from a theoretical guarantee : that they converge to the maximum likelihood ( ml ) state @xmath3 ( discussed below ) .
however , the dia has been observed to converge slowly @xcite , and convex programming solvers such as sdpt3 and sedumi are known to require computational time that scales poorly with non - sparse matrix dimensionality @xcite .
[ [ section-2 ] ] a non - iterative quantum tomography method was devised by smolin _
_ , who showed that , if the measurement operators are traceless and the noise is of the gaussian type , the constrained maximum likelihood state @xmath3 can be retrieved in a single projection step from the unconstrained maximum likelihood state @xmath4 , which can be found very quickly using linear inversion @xcite . with an implementation of this method s core algorithm using a gpu , guo _ et al .
_ were able to recover a simulated 14-qubit density matrix @xcite .
however , the restrictive above conditions motivate the search for more broadly applicable efficient techniques .
projected gradient descent ( pgd ) methods are emerging as promising candidates for quantum tomography @xcite .
we present three pgd algorithms that converge towards the maximum likelihood quantum state : projected gradient descent with backtracking ( pgdb ) @xcite , fast iterative shrinkage - thresholding algorithm ( fista ) @xcite and projected gradient descent with momentum ( pgdm ) .
we provide evidence that they converge faster than both dia and sdpt3 , and scale more favourably than sdpt3 .
we also find that the pgd methods are very versatile in that one can model a wide variety of types of noise ; in particular , the case of ill - conditioned measurements . gonalves _ et al . _
@xcite and shang _
et al . _
@xcite have both recently discussed pgdb as an efficient technique for quantum tomography : however , the algorithmic variants we introduce here can significantly outperform it . furthermore , shang _
considered pauli measurements , for which the technique from ref .
@xcite is highly efficient .
it is therefore vital to study the performance of projected gradient descent outside of this realm to establish its true usefulness .
we here confirm that pgd methods continue to exhibit excellent properties in scenarios where the technique from ref .
@xcite is not applicable .
[ [ section-3 ] ] the remainder of this paper is structured as follows : after giving an introduction to quantum tomography and the general idea of projected gradient descent , we lay out three pgd algorithms and discuss their performance : namely , convergence profiles and running time . to the best of our knowledge , fista has never previously been applied to quantum tomography and pgdm is a novel algorithm altogether . dia and
sdpt3 , considered as current state - of - the - art algorithms , will serve as benchmarks by which the pgd approach will be judged .
finally , we report on the results of state reconstruction using pseudo - experimental data , i.e. simulations of realistic quantum tomographic experiments with noise .
we consider three figures of merit for quantum tomographic techniques : the convergence time and the fidelity between the recovered state and the actual one @xmath2 . over a broad range of hilbert space dimensions , we run the algorithms multiple times , each time with a randomly generated density matrices with fixed purity @xcite , and
record the running times and fidelities . the born rule , @xmath5 , being the central equation of quantum theory , encodes the probability @xmath6 to obtain a certain measurement outcome given a particular quantum state .
it involves a quantum state @xmath7 , which takes the form of a @xmath8 positive - semidefinite matrix of unit trace , and an hermitian measurement operator @xmath9 .
quantum tomography is essentially the process of finding the density matrix whose calculated outcome probabilities ( for an informationally complete set of @xmath10 operators ) most closely match the experimentally observed data .
[ [ section-4 ] ] the probabilities @xmath6 are of course not directly observable , only the number of clicks @xmath11 recorded in a real measurement device after a finite number of trials . in the absence of noise ,
the probabilities relate to the number of clicks through a multiplicative factor @xmath12 : @xmath13 . in real situations
, there is a discrepancy between @xmath14 and @xmath11 due to i ) technical noise in the measurement device and ii ) statistical uncertainty .
furthermore , if the noise is particularly severe , the matrix reconstructed with naive methods ( such as linear inversion ) will fail to qualify as a physical quantum state : the positivity or unit - trace properties can be violated .
multiple techniques have therefore been developed that allow one to search for an estimate matrix that is guaranteed to be physical .
examples include searching for the cholesky factor @xmath15 ( where @xmath16 is guaranteed positive ) and using a lagrange multiplier ( to ensure unit trace ) @xcite .
however , searching in the factored space can often lead to an ill - conditioned problem and slow convergence @xcite . as we evidence ,
there are advantages to be had by instead allowing the search to temporarily wander into unphysical territory .
[ [ section-5 ] ] the measurement operators , the expectation values @xmath6 and the detector clicks @xmath11 can be stacked into a matrix and two vectors , respectively : @xmath17 where @xmath10 is the total number of projectors . with the above notation
, the expectation value vector reads @xmath18 .
the computation of this vector takes @xmath19 floating - point operations in general , but a lower computational complexity can be achieved when the operators originate from tensor products @xcite .
the accuracy of the maximum likelihood state depends significantly on the condition number ( which is the ratio of maximum singular value to minimum singular value ) of the measurement matrix @xmath20 @xcite .
high condition numbers , which correspond to ill - conditioned measurement matrices , arise in the fields of detector tomography @xcite and superconducting artificial atoms @xcite .
the multiplicative factor @xmath12 can readily be estimated if at least a subset @xmath21 of the measurement matrix @xmath20 forms a povm , in which case the sum of the probabilities belonging to the set @xmath21 , @xmath22 , is independent of the state @xmath7 .
for example , if the measurement operators in @xmath21 form a basis , the sum in question amounts to unity . for an arbitrary povm ,
the best estimate for the multiplication factor is given by @xmath23 where @xmath24 is the number of projectors in the povm .
moreover , the average number of clicks per outcome is @xmath25 , and the total number of clicks for this povm is @xmath26 .
[ [ section-6 ] ] the distance between @xmath6 and @xmath27 is to be considered as a ` cost function ' @xmath28 in the sense of numerical optimisation . in the minimisation of a cost function , pgd algorithms are useful when one seeks a solution in a proper subset of a larger search space .
since physical quantum states , represented as unit - trace positive - semidefinite matrices , exist in a ( convex ) subset of the ( convex ) set of @xmath8 matrices , quantum tomography is indeed a problem of this kind .
projected gradient descent is an iterative procedure with two substeps . starting with a well - chosen physical state , first a step is taken in the downhill direction of the cost function , which has the chance to result in a nonphysical matrix .
second , to bring the estimate back within the constrained , physical space , we project it to the closest point in the solution space ( for example , using a matrix norm ) .
this two step process is then repeated until the cost function converges towards a low enough value .
since we are searching over a convex set , as long as the cost function is a strictly convex function of @xmath7 , there will be a unique solution that minimises it .
[ fig : scheme ] shows the evolution of the density matrix estimate of a qubit through six iterations of pgd .
we have yet to specifically define the figure of merit for _ closeness _ between the estimated probabilities and the outcome frequencies .
maximum likelihood analysis provides a principled way to derive such a figure of merit . when doing statistical estimation , it is necessary to operate within a statistical model or belief system : one approach is bayesian estimation , which works with iteratively updating such beliefs using bayes rule . here , however , the beliefs are encoded in the likelihood function for a multinomial experiment : @xmath29 maximizing this function
i.e. finding the quantum state @xmath7 that makes the observed data the most likely is the most widely applied philosophy for tomography @xcite .
since the maximum - likelihood state @xmath30 is also the minimum of @xmath31 , we can proceed by minimising the second function , and we may ignore any scale or shift by a constant that is independent of @xmath7 .
we therefore define the cost function @xmath32 that we seek to minimise .
when the number of trials is large , this is well approximated by the poisson - approximated gaussian likelihood function @xmath33 with @xmath34 .
assuming poisson - distributed data , the variance for outcome @xmath35 is the number of clicks @xmath11 .
hence @xmath36 corresponds to the ratio of the error to the expected error on outcome @xmath35 .
the true density matrix gives an expected negative log - likelihood per outcome @xmath37 of unity because of the noise on the outcomes @xmath11 .
a value greater than unity indicates a poor density matrix estimate or an incomplete noise model , whereas a value smaller than unity is a sign of overfitting to noise .
in general , the maximum likelihood density matrix overfits the data slightly @xcite , but one can not achieve a better estimate in the absence of prior knowledge .
we are now ready to detail the algorithms for reconstructing the density matrix . in all of the following algorithms ,
the completely mixed state @xmath38 will serve as the starting point .
our selection of four iterative algorithms are then defined by a recursion relation relating the density matrix at the next iteration to the matrix at the current iteration .
the process of any pgd algorithm involves steps in the general gradient direction , interspersed with leaps back into the constrained set @xcite .
the simplest such algorithm can be written in a single line @xcite : @xmath39 , \label{baba}\ ] ] where @xmath40 is a step size and @xmath41 $ ] is a projection onto the set of unit - trace positive matrices , seeking the ` closest ' unit - trace positive matrix to its argument .
various approaches can be used to to establish what ` closest ' means ( including operator norms , see supplementary information ) .
we adopt the _ simplex projection _ @xmath42\rightarrow\mathcal s[\cdot]$ ] , which essentially consists of transforming the eigenvalues of the density matrix such that the sum is unity trace and they are all positive @xcite .
if the multiplicative factor @xmath12 is known or computed in advance , the version described in detail in ref .
@xcite applies , otherwise the projection must instead be performed over the space of positive matrices , preserving the trace of the argument @xcite .
[ [ section-7 ] ] we now proceed to discuss three pgd algorithms which are all are extensions of eq .
( [ baba ] ) .
here we augment the basic pgd algorithm of eq .
( [ baba ] ) with a technique borrowed the momentum - aided gradient descent method from the field of machine learning @xcite .
this technique stores a running weighted - average @xmath43 of the log - likelihood gradient .
this running average provides a memory of previous descent directions which is used to better estimate the next descent step .
the core of this algorithm reads @xmath44 where @xmath45 codes for the level of ` inertia ' for the descent direction , and @xmath46 is the step size . in general
, these metaparameters depend on the iteration number @xmath47 , but can also be set as constants throughout the algorithm .
we provide full pseudo - code for this and the other algorithms in the methods section , as well as matlab implementations .
fista was first developed in the context of image denoising @xcite , but here we introduce the method for use in quantum state tomography with adapted refinements . in this implementation of pgd , the change in the iterate @xmath48 is not always in the descent direction , i.e. the log - likelihood function can go up , but as we shall see it descends much faster on average than the basic gradient descent algorithm from eq .
( [ baba ] ) .
the core of the algorithm is given by @xmath49 ~,\ ] ] where @xmath40 is a step size .
this pgd method has recently been applied to quantum tomography simulations in ref .
a similar variant has been studied in ref .
@xcite , where the authors report on a hybrid method between the dia and pgd .
the pgdb algorithm , whose main characteristic consists of trying to find the maximum step size that reduces the negative log - likelihood , can be written as @xmath50 ~,\ ] ] where @xmath51 is a parameter to be loosely optimised at each step through backtracking .
each iteration of this algorithm is guaranteed to lower the negative log - likelihood unless a stationary point is reached , in which case the stopping criterion is satisfied .
the diluted iterative algorithm ( dia ) is based on the gradient of the log - likelihood function .
the algorithm can be simply stated with the following two iterative equations @xcite @xmath52 h^{-1/2 } ~ , \\
\rho_k & = \frac{(i+\epsilon { r}_{k } ) \rho_{k-1}(i+\epsilon { r}_{k } ) } { \text{tr}[(i+\epsilon { r}_{k } ) \rho_{k-1}(i+\epsilon { r}_{k } ) ] } ~ , \end{aligned}\ ] ] where @xmath53 .
the variable @xmath54 is optimised at every iteration , such that it minimises the log - likelihood function , and can be implemented in various ways @xcite .
the matrix @xmath55 reduces to the identity ( up to a constant ) when all the measurement operators form a povm .
the dia leaves the density matrix estimate @xmath48 positive at every iteration .
[ [ section-8 ] ] it has been observed that the dia converges quickly for the first few iterations and converges very slowly later @xcite . in the results section ,
we corroborate these observations . as we emphasised above , quantum tomography is often equivalent to minimising a convex function over a convex set . in the field of numerical optimisation , a problem
is considered effectively solved if it can be cast into this form , partly because of the powerful and efficient algorithms and software packages that are available , and also because of the guarantee of global optimality for the solutions that they find .
such a software package therefore makes a natural benchmark for quantum tomography algorithms , with the the understanding that ( because of their general purpose nature ) they are not optimised for full tomography and unlikely to be as fast as other methods .
we use the cvx software environment ; and the sdpt3 solver , which is an example of an infeasible path - following algorithm .
we perform quantum tomography simulations on multi - qubit systems with all of the techniques mentioned in the previous section . when using canonical pauli measurements , all of the techniques are found to work well in that they all recover @xmath3 .
since the simulations consistently reach high likelihoods in practice , we concentrate on the total computation time .
the exit criterion for all techniques except for sdpt3 whose code we do not change is such that when the average gradients of the last 20 iterations is sufficiently small , the optimisation procedures are terminated .
examples of convergence curves for 6-qubit systems using pauli measurements are shown in fig .
[ fig : cost ] .
the details of the implementations and simulated data are provided in the methods section . as already remarked in refs .
@xcite , the dia displays fast convergence in the first few iterations but requires many more to finally satisfy the exit criterion .
sdpt3 converges in between only 10 and 15 iterations , but each iteration has a computational complexity of @xmath56 , rendering it slow in high dimensions .
we put the tomographic techniques to the test using ill - conditioned measurement matrices @xcite , see fig .
[ fig : cost ] b ) .
if the measurement operators are limited to a restricted region of the hilbert space , the condition number of the measurement matrix is greater than unity , and the error on the final density matrix estimate will necessarily increase @xcite . here
, the measurement matrix is built using the bases @xmath57;[1~0];[\cos(\beta/2),\sin(\beta/2)];[\sin(\beta/2),\cos(\beta/2 ) ] ; \\ & [ \cos(\beta/2),\bold{i}\sin(\beta/2)];[\sin(\beta/2),\bold{i}\cos(\beta/2 ) ] \end{aligned}\ ] ] with @xmath58 for regular canonical pauli operators and @xmath59 for the ill - conditioned case ; see the bloch spheres on fig .
[ fig : cost ] for an illustration of these vectors .
gonalves _ et al .
_ provide a proof of the monotonicity of @xmath60 for pgdb , that is to say that the cost function never increases in this algorithm .
by contrast , pgdm and fista are both algorithms for which the cost function may increase from one iteration to the next , but interestingly , this tends to speed up their performance relative to pgdb with regards to the ill - conditioned measurement matrices .
we show the running time of each algorithm as a function of hilbert space dimensionality ( number of qubits ) in fig .
[ fig : dim ] .
the speed - up of pgdm over pgdb for high - dimensions is present in both panels of fig .
[ fig : dim ] , but particularly pronounced for the quantum state reconstruction task based on the ill - conditioned measurement matrices , where pgdm is about ten times faster than pgdb on average for the eight and seven - qubit cases .
the running times of pgdm and fista algorithms are more resilient to a condition number change than pgdb , due to the fact that the number of pgdm and fista iterations required to reach @xmath3 grows very little as a function of the measurement matrix condition number .
[ fig : mubness ] a ) illustrates this dependence .
the semidefinite programming technique does not depend on the condition number : in our simulations , sdpt3 always took about 15 iterations to reach @xmath3 .
there exists a monotonic relationship between ill - posedness and the accuracy of the recovered state : for a fixed number of events per measurement , the more ill - posed the measurement matrix is , the lower the fidelity between the recovered state and the true one .
this relationship is illustrated in fig .
[ fig : mubness]b , where the vertical axis corresponds to one minus the fidelity @xmath61 between the recovered density matrix and the true state . the extreme cases , i.e. @xmath62 and @xmath63 , correspond to mutually unbiased bases and a single basis measurement , respectively .
we show a zoomed subset of this data around the intermediate angle , i.e. @xmath64 , that gives statistical evidence that the projected gradient descent techniques consistently converge to @xmath3 .
[ [ section-9 ] ] a key advantage of the pgd techniques is their versatility .
they successfully and quickly converge to the maximum likelihood state in a wide range of cases , regardless of the desired accuracy and whether the measurement matrix is well- or ill - conditioned .
pgdm and fista our new pgd algorithms are shown to be especially well suited for ill - conditioned problems .
quantum tomography includes three main subfields : detector , process and state tomography .
state tomography algorithms are straightforwardly transferable to process tomography , but applying the same algorithms to detector tomography with success is not trivial .
the problem of detector tomography lies in characterising an unknown detector povm from an informationally complete set of known states .
if the tomographer probes an optical detector with coherent states , the problem of detector tomography is ill - conditioned , and like the density matrix , the povm elements must be positive - semidefinite @xcite .
currently , the state - of - the - art algorithms to solve this problem are semidefinite program solvers such as sedumi @xcite . because pgdm and fista perform well in the case of ill - conditioned measurement matrices , these algorithms hold great promise for optical detector tomography .
one avenue for future work is the application of these two algorithms to the characterisation of detector povms in high dimensions . in summary ,
the pgd techniques have proven their worth in that they all converge towards the maximum likelihood density matrix reliably .
the different pgd techniques complement each other such that pgdb is fastest in low dimensions and , according to our simulations , pgdm is fastest beyond five - qubit systems .
further , we find that the pgd techniques reach @xmath3 significantly faster than the dia and sdpt3 in the vast majority of scenarios , thus surpassing the state - of - the - art techniques with regards to assumption - free quantum state tomography . to reduce the memory requirements , we chose to use rank-1 measurement operators in the simulation . instead of matrix operators , the measurements
take the form of @xmath65-dimensional vectors .
given rank-1 projectors @xmath66 , the born rule is written @xmath67 and the measurement matrix takes the following form @xmath68 in this compact notation , the measurement matrix is @xmath69-dimensional , thus requiring @xmath65 times less ram memory compared to the full - rank case .
the gradient of the gaussian negative log - likelihood function is compactly written as @xmath70 where the elements of the @xmath71 matrix are defined : @xmath72 .
the computation of the above gradient takes @xmath19 floating - point operations .
this is the gradient that we use in the pgd algorithms for the simulations .
for all simulations in the main text , the average number of events per outcome @xmath73 is set to @xmath74 .
density matrices were chosen randomly in the haar sense but always a with purity of 0.5 .
all simulations were performed on a single thread on an intel xeon haswell processor .
e.b . acknowledges the financial support of the fqrnt , grant # 176729 .
j.l . acknowledges the financial support of the engineering and physical sciences research council ( epsrc , uk , grants ep / m006514/1 and ep / m01326x/1 ) .
g.c.k . was supported by the royal commission for the exhibition of 1851 .
e.m.g . acknowledges support from the royal society of edinburgh and the scottish government . | accurate quantum tomography is a vital tool in both fundamental and applied quantum science .
it is a task that involves processing a noisy measurement record in order to construct a reliable estimate of an unknown quantum state , and is central to quantum computing , metrology , and communication . to date , many different approaches to quantum state estimation have been developed , yet no one method fits all applications , and all fail relatively quickly as the dimensionality of the unknown state grows . in this work , we suggest that projected gradient descent is a method that can evade some of these shortcomings .
we present three novel tomography algorithms that use projected gradient descent and compare their performance with state - of - the - art alternatives , i.e. the diluted iterative algorithm and convex programming .
our results find in favour of the general class of projected gradient descent methods due to their speed , applicability to large states , and the range of conditions in which they perform . |
the standard model of cosmology accounts for the phase of accelerated expansion that the universe is currently undergoing by introducing a cosmological constant term in the gravitational action .
according to observations , the value of this cosmological constant should be @xmath0 ev@xmath1 , which is very far away from the natural scale of gravitation set by the planck mass @xmath2 gev .
this fact poses the so - called naturalness problem associated to the cosmological constant .
however , this could be signaling that the cosmological constant causing the accelerated expansion is not a true constant , but an effective description arising from some underlying physical theory . on the other hand ,
the accelerated expansion could also be caused by some other physical mechanism like a cosmological scalar field or a modification of the gravitational interaction @xcite .
nonetheless , in most of these models , the original naturalness problem still remains because new dimensional scales are needed in order to produce the correct acceleration .
furthermore , they usually have some other problems related to the presence of instabilities or local gravity constrains . in this work
we shall consider vector - tensor theories of gravity and show how vector fields can help to alleviate the naturalness problem .
indeed , we shall show that the very familiar electromagnetic field can play the role of dark energy without any of the aforementioned problems .
we shall consider the most general action for a vector - tensor theory leading to second order linear equations of motion for the vector field : @xmath3=\int d^4x\sqrt{-g}\left[-\frac{1}{16\pi g}r+\omega ra_\mu a^\mu+\sigma r_{\mu\nu}a^\mu a^\nu + \lambda(\nabla_\mu a^\mu)^2+\epsilon f_{\mu \nu}f^{\mu\nu}\right]\label{action}\end{aligned}\ ] ] with @xmath4 , @xmath5 , @xmath6 and @xmath7 dimensionless parameters .
the interest of this theory is that @xmath8 is the only dimensional constant and we do not have potential terms for the vector field so that no new scales are introduced in the action . a general study of the homogeneous cosmological evolution for this action is given in @xcite where a classification attending to the behavior of the vector field throughout the different epochs of the universe evolution is performed .
also , the case in which the vector field dominates the universe is considered and the corresponding autonomous system is analyzed . for this system , it is shown the presence of both attractors and repellers in which accelerated expansion can take place .
these repellers are interesting for inflaton models because we have an initial state with accelerated expansion that can eventually end as the trajectory in the phase space goes away from the critical point . on the other hand , attractors with accelerated expansion can give rise to dark energy candidates since they provide late - time accelerated expansion . finally ,
when a matter component is introduced in addition to the vector field , we have that , for quite general conditions , there are solutions with a transition from a matter dominated universe to a phase of accelerated expansion with a phantom behavior for the vector field that leads to a future type iii singularity in which the scale factor remains finite but the expansion rate diverges . the small scales behavior of the vector - tensor theory given by ( [ action ] ) is well described by means of its ppn parameters . since gr is in excellent agreement with solar system experiments , an alternative gravitational theory whose ppn parameters are the same as those of gr will agree with small scales viability conditions .
the ppn parameters are obtained by performing perturbations around a minkowski spacetime with a constant background vector field of the form @xmath9 .
when we impose the ppn parameters of the vector - tensor theory to be identical to those of gr we find that there are six models which satisfy such a condition for any value of the background vector field @xcite .
these six models have @xmath10 and can be classified as : gauge non - invariant models : @xmath11 , @xmath12 , @xmath13 ; gauge invariant models : @xmath14 and @xmath15 with @xmath16 . for these particular vector - tensor theories
we study the presence of instabilities both at the classical and the quantum level in the inhomogeneous perturbations .
the classical stability can be achieved by imposing the absence of modes whose propagation speed is imaginary so that we do not have exponentially growing modes .
on the other hand , the quantum stability will be guaranteed by eliminating the modes with negative energy ( ghosts ) in order to have a stable quantum vacuum ( see table 1 ) .
we find that the only two theories which can be made completely free from instabilities are the pure maxwell action and maxwell action supplemented with a gauge fixing term .
15.9cm|x|c|c|c|c| & & & & + & model i & model ii & model iii & gauge invariant models + & & & & @xmath17 + & & & & + classical stability & @xmath18 & @xmath19 & always & @xmath20 $ ] + & & & & + & & & & + gravitational waves & @xmath21 & @xmath22 & unaffected & @xmath23 + & & & & + & & & & + quantum stability & @xmath24 & @xmath25 & @xmath26 and @xmath27 & @xmath28 + & & & & + & & & & + viability condition & @xmath29 & @xmath30 & @xmath31 and @xmath32 & @xmath33 + & & & & + [ tablappn ]
the cosmic vector model having a scaling behavior in the early universe proposed in @xcite is described by the following action : @xmath34.\ ] ] in this model , the fraction of energy density stored in the vector field remains constant throughout the radiation dominated epoch and , in addition , the value of the vector field during such an epoch also remains constant .
this prevents any dependence on the moment at which we set the initial conditions because the subsequent evolution of the universe will be the same irrespective of the initial redshift ( as long as it is well inside the radiation dominated epoch ) .
moreover , the initial fraction of energy density corresponding to the vector field needed to explain the snia measurements turns out to be @xmath35 and the vector field must take a value @xmath36 , which are values that can arise _ naturally _ in the early universe , as quantum fluctuations for instance . that way , this model allows to alleviate the naturalness problem and it does not require any fine - tunings . finally , we have confronted the predictions of the model to snia , cmb and bao measurements .
an interesting property of this model is that it provides the best fit to date to the sn gold dataset with the same number of parameters as the standard @xmath37cdm .
moreover , cmb measurements predicts a closed universe for this model , being the flat case strongly disfavored .
finally , we find some tension with bao observations , which could be due to the dependence of such data on the fiducial model ( @xmath37cdm ) whose evolution is very different from that given by the cosmic vector model .
the second vector dark energy model we are going to discuss is nothing but the very familiar electromagnetic field .
it is well - known @xcite that the canonical quantization of the electromagnetic field needs to impose the subsidiary lorenz condition @xmath38 in order to reduce the corresponding hilbert space so that we end up with the two transverse states for the photon .
however , when we consider the covariant quantization in an expanding universe , the lorenz condition turns out to be difficult to implement .
this is due to the fact that @xmath39 behaves as a free scalar field so that it can be excited by means of gravitational fields . for these reasons ,
we consider the action including a gauge fixing term@xcite : @xmath40\label{emaction}\ ] ] as the fundamental action for the electromagnetic field . in this action
the gauge fixing term is a physical term , rather than being a mathematical trick , so that we break the @xmath41 invariance , although a residual gauge symmetry @xmath42 with @xmath43 remains .
notice that this action is also the starting action in the path integral formalism once we have fixed the element of the orbits of the group .
the corresponding modified maxwell equations deduced from ( [ emaction ] ) are : @xmath44 now , if we take the four divergence of this equation we obtain : @xmath45 where we have used the conservation of the electromagnetic current .
thus , as we said before , @xmath39 behaves as a free scalar field non - minimally coupled to gravity .
moreover , in an expanding universe , a free scalar field gets frozen for super - hubble modes so that the gauge fixing term appearing in ( [ emaction ] ) will give an effective cosmological constant on large scales . on the other hand , we also know that a free scalar field decays as @xmath46 for sub - hubble modes which implies that the gauge fixing term will be negligible at small scales and we effectively recover the usual maxwell equations .
since we are breaking the @xmath41 invariance and we do not impose any extra subsidiary condition , the theory contains one extra degree of freedom so that the field can be decomposed as follows : @xmath47 where @xmath48 with @xmath49 are the two transverse modes of the massless photon , @xmath50 is a new scalar state that represents the mode that would have been eliminated if we had imposed the lorenz condition and , finally , @xmath51 is a purely residual gauge mode , which can be eliminated by means of a residual gauge transformation in the asymptotically free regions , in a completely analogous way to the elimination of the @xmath52 component in the coulomb gauge .
the complete quantization procedure in an inflationary phase with de - sitter expansion can be found in @xcite where the primordial power spectrum defined by @xmath53 for super - hubble modes is computed and given by : @xmath54 with @xmath55 the constant hubble parameter during inflation .
this result implies that the effective cosmological constant given by the gauge fixing term is @xmath56 so that the measured value of the cosmological constant requires @xmath57 ev , which corresponds to an inflationary scale @xmath58 tev .
thus , we see that the cosmological constant scale can be naturally explained in terms of physics at the electroweak scale .
although the homogeneous evolution is the same as in the standard @xmath37cdm , we have a crucial difference because , in this case , we can have inhomogeneous perturbations , unlike in the standard model where the cosmological constant is a true constant . in @xcite
we have considered the effects of the inhomogeneous perturbations on the cmb anisotropies and large scale structures and found that only at the largest scales we can have discriminatory consequences . if the primordial amplitude of the perturbations are not extremely large , the cmb temperature and matter power spectra are perfectly compatible with observations .
we have shown that vector fields are compelling candidates to drive an accelerated phase and they can alleviate or even solve some of the usual naturalness or coincidence problems arising in most of dark energy models .
more interestingly , we have shown that the electromagnetic field can account for the present phase of accelerated expansion of the universe and provides a natural explanation for the observed value of the cosmological constant .
this shows that it is possible to establish the fundamental nature of dark energy without resorting to new physics . :
this work has been supported by ministerio de ciencia e innovacin ( spain ) project numbers fis 2008 - 01323 and fpa 2008 - 00592 , ucm - santander pr34/07 - 15875 , cam / ucm 910309 and mec grant bes- 2006 - 12059 .
1 url # 1#1urlprefix[2][]#2 wetterich c 1988 _ nucl . phys _ * b302 * 668 ; caldwell rr , dave r and steinhardt p j 1998 _ phys .
_ 80 1582 ; armendariz - picon c , damour t. and mukhanov v 1999 _ phys .
lett . _ * b458 * 209 ; koivisto t and mota d f 2007 _ phys .
* d75 * 023518 ; carroll sm , duvvuri v , trodden m and turner m s 2004 _ phys .
rev . _ * d70 * 043528 ; dvali g , gabadadze g and porrati m 2000 _ phys .
* b485 * 208 . beltrn jimnez j and maroto a l 2009 _ jcap _ * 0903 * 016 ; beltrn jimnez j and maroto a l 2009 ( _ preprint _ ) ; beltrn jimnez j and maroto a l 2009 ( _ preprint _ ) ; beltrn jimnez j and maroto a l 2009 ( _ preprint _ ) . | we consider a general class of vector - tensor theories of gravity and show that solutions with accelerated expansion and a future type iii singularity are a common feature in these models .
we also show that there are only six vector - tensor theories with the same small scales behavior as general relativity and , in addition , only two of them can be made completely free from instabilities .
finally , two particular models as candidates for dark energy are proposed : on one hand , a cosmic vector that allows to alleviate the usual naturalness and coincidence problems and , on the other hand , the electromagnetic field is shown to give rise to an effective cosmological constant on large scales whose value can be explained in terms of inflation at the electroweak scale . |
it is a classical result of bott and samelson @xcite that the orbits of a variationally complete group action are taut .
isometric group actions are generalized to foliations via the notion of singular riemannian foliation , see @xcite ; and lytchak and thorbergsson defined in @xcite the analog of variational completeness for such foliations via the notion of foliation without horizontally conjugate points .
the main aim of this paper is to show that tautness also holds in this more general situation .
a basic approach is to generalize the explicit construction of linking cycles done by bott and samelson . pursuing this strategy
, we need and find global structural characteristics of the foliations concerned .
while singular riemannian foliations are well understood locally , their global structure can be arbitrarily complicated at least in the non compact case .
after a short introduction , we restrict our interest to properly embedded leaves from chapter [ s : ex ] forward . at first
, we show the existence of global tubular neighborhoods , see theorem [ theotube ] , which is the basic tool to reveal global structural results in the following .
then , we have a look at exceptional leaves , i.e. leaves of highest dimension with non - trivial holonomy .
this kind of leaves does not occur too often as the set of specificly regular leaves is open , connected and dense in the manifold ( propositions [ propregdense ] and [ propregcon ] ) .
we define the type of a leaf and show that the stratum of exceptional leaves as well as the manifold itself are stratified by the types of leaves . in the third chapter ,
we analyse foliations without horizontally conjugate points ( with properly embedded leaves ) .
locally constructed variations of horizontal geodesics can be extended globally , which leads to the geometric characterization formulated in theorem [ theowhcp ] : a singular riemannian foliation has no horizontally conjugate points if and only if the singular points are exactly the focal points of the leaves . for this theorem , we need the extension of homothetical transformations , defined in @xcite within tubular neighborhoods , to the negatives which is done in proposition [ prophomotheties ] .
even more , we show that the homothetical transformations are diffeomorphisms that respect not only the leaves but also the types of leaves .
there are two other papers that among other things show the existence of negative homotheties , namely the one of alexandrino and tben , @xcite , and the one of lytchak and thorbergsson , @xcite .
as these proofs are contemporaneous but independent and use completely different methods , we include our proof here . with these structural results
we finally can address ourselves to morse theory of the leaves , i.e. the tautness problem . as a first step ,
we show that the leaves of a singular riemannian foliation without horizontally conjugate points are @xmath0-taut .
therefrom , we conclude that there are no exceptional leaves if the foliated manifold is simply connected . at last
, we show theorem [ theotaut ] : the leaves of a singular riemannian foliation without horizontally conjugate points are taut .
_ acknowledgements .
_ this paper essentially is my thesis which i wrote under the supervision of g. thorbergsson at the university of cologne .
he deserves my gratitude for the constantly encouraging support during the last years .
furthermore , i want to thank a. lytchak and s. wiesendorf for essential hints towards proposition [ propiso ] and theorem [ theotaut ] in this generality .
let @xmath1 be a riemannian manifold of dimension @xmath2 . a decomposition @xmath3 of @xmath4 into smooth , connected , injectively immersed submanifolds , the _ leaves , _
is called a _ singular riemannian foliation _ if the following two conditions hold : \(i ) the modul @xmath5 of smooth vector fields on @xmath4 that are tangential to the leaves is _ transitive , _
i.e. for a leaf @xmath6 through @xmath7 , and for every vector @xmath8 there is a vector field @xmath9 such that @xmath10 thus , the foliation is _ singular _ in the sense of @xcite and @xcite .
\(ii ) the metric @xmath11 on @xmath4 is _ adapted _ to @xmath3 : each geodesic in @xmath4 that hits one leaf perpendicularly remains perpendiculary to all leaves it passes through .
such a geodesic is called _ horizontal .
_ let @xmath12 denote the maximal dimension of the leaves .
a leaf is _ regular _ if its dimension is equal to @xmath13 otherwise it is _
singular_. equally , a point @xmath7 in @xmath4 is _ regular _ ( resp .
_ singular _ ) if the leaf @xmath6 is regular ( resp .
singular ) .
the manifold @xmath4 is stratified by the dimension of the leaves .
the _ regular stratum _
@xmath14 is the union of all leaves of dimension @xmath12 .
the union of all leaves of lower dimension , the singular stratum , is denoted by @xmath15 each stratum @xmath16 of leaves of dimension @xmath17 is an embedded submanifold , @xmath18 is closed , and the regular stratum is open , connected , and dense in @xmath19 for the leaf @xmath20 through @xmath21 let @xmath22 denote the normal bundle of @xmath6 in @xmath23 if @xmath24 is a relatively compact , open , and connected subset , @xmath25 denotes the bundle of open normal balls with radius @xmath26 around @xmath27 in @xmath28 for @xmath26 sufficiently small , the normal exponential map is a diffeomorphism of @xmath25 onto its image in @xmath29 which is called a _ ( distinguished ) tubular neighborhood of @xmath27 ( resp . around @xmath7 ) with radius @xmath30 _ @xmath31 a _ tube _ of radius @xmath32 around @xmath27 is the union of all points in @xmath33 whose distance to @xmath27 is equal to @xmath34 @xmath35 let @xmath36 denote the projection along horizontal geodesics .
we call @xmath37 a _ slice . _
if @xmath38 and @xmath26 sufficiently small , the leaf @xmath39 cuts the slice transversally .
def2 ( 48,25)@xmath7(95,35)@xmath6(45,35)@xmath40(17,42)@xmath33(36,19)@xmath27(62,15)@xmath41 by the following lemma [ lemlocaltube ] , the connected component of that contains @xmath42 lies in the tube @xmath43 such a component @xmath44 is called a _
plaque_. for more details we refer to @xcite , chapter 6 , wherefrom we just cite the following facts : [ lemlocaltube ] \(i ) the projection @xmath45 is a surjective submersion , and the fiber @xmath46 meets all the plaques in the tubular neighborhood for all @xmath47 .
\(ii ) the distance between neighboring leaves is locally constant for the topology of leaves : @xmath48 for all @xmath49 in @xmath50 for every point in @xmath14 there exists an open neighborhood @xmath51 called _ simple , _ such that the projection @xmath52 onto the manifold of plaques is a submersion .
the manifold @xmath53 can be endowed with a riemannian metric @xmath54 such that the transversal part of @xmath11 is the pullback of @xmath54 via @xmath55 and if @xmath56 is a horizontal geodesic in @xmath57 , its projection @xmath58 is a geodesic in @xmath53 .
let us have a look at a plaque @xmath59 by the composition of homothetical transformations in the normal bundle @xmath60 with @xmath61 we can define a _
homothetical transformation _
@xmath62 with respect to @xmath63 for all positive @xmath64 within the domain @xmath65 and [ lemposhom ] the homothetical transformation @xmath62 sends plaque to plaque and therefore respects the singular foliation @xmath3 .
if @xmath56 is a horizontal geodesic , it follows by this lemma that within a tubular neighborhood of @xmath66 @xmath67 if @xmath68 is simple , the belonging of points to their plaque in @xmath57 uniquely defines diffeomorphisms between local transversals .
given @xmath69 and @xmath70 in the same plaque , we call the corresponding diffeomorphism _ sliding along the leaves from @xmath69 to @xmath70 , _ see @xcite , section 1.7 . by coverings with simple sets ,
this definition can easily be extended to arbitrary points in a regular leaf and corresponding local transversals that are sufficiently small .
let @xmath71 denote a path from @xmath69 to @xmath70 and let @xmath72 denote the germ of this diffeomorphism at @xmath69 along @xmath73 note that @xmath72 does neither depend on the choice of the covering nor on the choice of local transversals , and concerning @xmath71 only on its homotopy class .
this germ is called _ sliding along the leaves from @xmath69 to @xmath70 along @xmath71 . _ for every point @xmath74,$ ] the slice of minimal radius of the covering tubes @xmath75 is a local transversal which then defines for every @xmath76 a unique , continuous path @xmath77
\to l_y$ ] from @xmath42 to its image under sliding along @xmath71 contained in @xmath78 whereby @xmath79 for all @xmath80.$ ] thus , in the following we sometimes speak about sliding of distances instead of points , which can be used not only within regular leaves and simple neighborhoods but by lemma [ lemlocaltube ] always as long as the path @xmath71 is completely contained in a tubular neighborhood .
we briefly recall the notions above .
for details see @xcite , section 10.4 , and @xcite : let @xmath81 be a submanifold with tangent bundle normal bundle @xmath82 and normal exponential map if @xmath83 is a critical point of @xmath84 we call @xmath83 a _ focal vector _ and @xmath85 a _ focal point .
_ let @xmath56 denote a geodesic with @xmath86 an _
@xmath87-jacobi field along @xmath56 _ is a jacobi field @xmath88 along @xmath56 such that @xmath89 where @xmath90 denotes the shape operator .
@xmath88 is a @xmath87-jacobi field if and only if it is the variational vector field of a variation @xmath91 of @xmath92 where all paths @xmath91 are geodesics with @xmath93 the following holds for a geodesic @xmath56 perpendicular to @xmath87 in @xmath66 : [ cols="<,^ , < " , ] the _ multiplicity of @xmath94 _ is equal to the dimension of the kernel of @xmath95 this also equals the dimension of the space of jacobi fields along @xmath56 vanishing in @xmath96 let now @xmath81 be properly immersed .
let @xmath97 denote the geodesic starting in the direction of @xmath98 furthermore , let @xmath99 be the unit normal bundle of @xmath100 then the _ cut locus function _ is defined by @xmath101 ; \sigma(v ) = \sup \{t \in { \mathbf{r}}\mid d(\gamma_v(t ) , n ) = t\}.\ ] ] the _ cut locus of @xmath87 _ is defined to be the set @xmath102 we call @xmath103 _ minimal _ , if @xmath104 .
the associated geodesic @xmath105}$ ] and each of its parametrizations of constant speed are called _ minimal geodesics . _
a minimal vector @xmath83 is a _ cut vector , _ if there is a different minimal @xmath106 with the same end point ; @xmath85 is then called a _ cut point . _ then , @xmath107 the following proposition holds : [ propcutlocus ] \(i ) each point of @xmath108 is a cut point or a focal point .
\(ii ) the cut locus function @xmath109 is upper semi - continuous , and \(iii ) if @xmath87 is properly immersed , @xmath109 is continuous and @xmath108 is closed in @xmath19
in general , the leaves of a singular riemannian foliation are embedded submanifolds . for the following ,
we additionally assume the embeddings to be proper .
it is easy to see that a tubular neighborhood then contains only finitely many plaques of every meeting leaf . as a counterexample
take the one - dimensional , irrational foliation of a torus . for a singular riemannian foliation with properly embedded leaves
, we can show the existence of global tubular neighborhoods , which will be a useful tool later : [ theotube ] let @xmath4 be complete and @xmath110 be a singular riemannian foliation with properly embedded leaves .
then , for every leaf @xmath81 there is an @xmath111 such that the normal exponential map @xmath112 is a diffeomorphism .
the proof is devided into two steps : \(i ) the distance of the leaves is globally constant , i.e. : if @xmath113 \to m$ ] is a horizontal geodesic that is locally minimizing between @xmath114 and @xmath115 then for every @xmath116 there is a horizontal geodesic of length @xmath117 from @xmath69 to @xmath115 and this geodesic , too , is locally minimizing between @xmath114 and @xmath118 hereby , we call @xmath56 _ locally minimizing , _ if there is a neighborhood @xmath119 around @xmath66 and a neighborhood @xmath120 around @xmath121 such that @xmath122 \(ii ) the distance between a leaf @xmath87 and its cut locus @xmath108 is bounded away from zero . for every lower bound @xmath123 it follows by definition of @xmath108 that the map @xmath124 is a diffeomorphism
. we prove step ( i ) as follows : with regard to @xmath125 it is clear that this property is closed .
that it is open , too , does not follow directly from the existence of local tubular neighborhoods as , in general , @xmath66 is not contained in such a neighboorhood of a plaque of @xmath115 for example if @xmath66 is singular but @xmath126 regular .
so , we have to show that there is a neighborhood @xmath127 of @xmath66 such that @xmath128 for all @xmath129 for this purpose take a covering of @xmath56 by tubular neighborhoods as follows : tube1 ( 40,79)@xmath66(40,67)@xmath130(40,58)@xmath131(40,51)@xmath132(40,40)@xmath133(40,30)@xmath134(40,15)@xmath121(78,79)@xmath135(78,15)@xmath136(30,80)@xmath69(30,75)@xmath137(31,65)@xmath138 around the points @xmath66 and @xmath139 let the plaques @xmath119 and @xmath120 be centers of tubular neighborhoods .
the centers of all neighborhoods of the covering shall intersect @xmath140 furthermore , choose a minimal subdivision @xmath141 where @xmath142 is such a centering plaque for all @xmath143 i.e. each point on @xmath56 lies either in one or in the intersection of two consecutive tubular neighborhoods . let @xmath144 $ ] be chosen in such a way that @xmath145 and @xmath146 lies in the intersection : @xmath147 now , let @xmath127 be a neighborhood sufficiently small of @xmath66 and @xmath129 as @xmath56 is contained completely in the finite , open covering of tubular neighborhoods , there exists a @xmath57 so small that the following construction of a broken geodesic is possible within the covering . in the neighborhood
@xmath148 there is a minimal geodesic @xmath149 \to { \rm tub}(p_0),$ ] with @xmath150 and @xmath151 and the length of @xmath152 is equal to the length of @xmath56 restricted to @xmath153.$ ] similarly , there is a minimal geodesic @xmath138 from @xmath154 to @xmath155 of length @xmath156}).$ ] proceeding like that , we obtain a broken geodesic @xmath157 of length @xmath158 which starts in @xmath159 and ends in @xmath160 thus , the distance between @xmath69 and @xmath161 has to be less equal than or equal to @xmath162 but @xmath56 is locally minimizing and equality follows by @xmath163 to prove step ( ii ) , let @xmath164 be arbitrary .
furthermore , choose @xmath165 in such a way that there is a plaque @xmath166 with tubular neighborhood @xmath167 in addition , we can assume that the leaf @xmath87 meets the set @xmath168 just once , namely as centering leaf .
now , let us assume that @xmath169 as the leaves are properly embedded submanifolds , proposition [ propcutlocus ] tells us that the cut locus function @xmath170 is continuous .
thus , there exists a leaf @xmath171 with @xmath172 where @xmath173 is a connected component of @xmath174 there exists a point @xmath175 with @xmath176 and by ( i ) , there exists a horizontal geodesic @xmath56 that realizes the distance @xmath177 and meets the leaf @xmath171 at length @xmath178 but @xmath56 can not be minimal between @xmath175 and @xmath171 because the cut locus of @xmath87 meets @xmath56 previous to @xmath179 thus , @xmath180 . sliding along the leaves with this local distance from @xmath171 to @xmath87 up to @xmath63 which is possible by ( i ) shows that there is
a plaque of @xmath171 lying within the tubular neighborhood @xmath181 on the @xmath182-sphere of @xmath100 but @xmath183 lies on the @xmath184-sphere around @xmath59 once again , sliding along the leaves with distance @xmath182 up to @xmath173 shows that there must be one more plaque of @xmath87 within the neighborhood @xmath185 a contradiction to the choice of @xmath26 !
tube2 ( 63,34)@xmath108(87,19)@xmath17(87,8)@xmath56(7,36)@xmath181(92,24)@xmath171(96,18)@xmath87(67,16)@xmath171(23.5,22)@xmath184(31,18.5)@xmath182(24,28.5)@xmath186 so , @xmath187 has to hold . in @xcite , the notion of for riemannian foliations
is defined by sliding along the leaves ( see subsection [ ss : sliding ] here ) on the regular stratum @xmath188 if @xmath71 is a loop in @xmath69 and if we choose the local transversals @xmath189 then @xmath72 is the germ of a local diffeomorphism with fixed point @xmath190 the composition of two loops @xmath191 in @xmath69 is @xmath192 thus , we obtain a group homomorphism @xmath193 whose image is the _ holonomy group of @xmath171 in @xmath69 . _ by identification of a local transversal with the quotient @xmath194 the holonomy group is independent of @xmath195 a regular leaf is called _ exceptional _ , if its holonomy is non - trivial
. furthermore , we call a regular leaf @xmath87 _ specificly regular _ , if there is a @xmath7 in @xmath87 and a local tubular neighborhood around @xmath7 wherein every meeting leaf has exactly one plaque .
note that only under the condition of being properly embedded , leaves with trivial holonomy are just the same as specificly regular leaves as the following lemma will show .
thus , we denote this stratum of leaves by @xmath196 [ lemhol ] let @xmath197 be singular riemannian with properly embedded leaves .
then , a regular leaf @xmath87 has non - trivial holonomy in @xmath69 if and only if for every global tubular neighborhood of @xmath87 there exists a leaf meeting the associated slice @xmath198 more than once , i.e. if and only if @xmath87 is notspecificly regular . to become acquainted with exceptional and specificly regular leaves , we state some easy facts about them previous to the proof of the lemma .
\(i ) for a leaf @xmath171 , the number of components of the intersection @xmath199 is independent of @xmath200 \(ii ) the holonomy of a leaf is non - trivial in every point of the leaf if it is non - trivial in an arbitrary point .
\(iii ) for specificly regular leaves , the number of components of the intersection as stated in ( i ) is constant along connected components of @xmath201 \(iv ) the plaques of a leaf @xmath202 lying in a sufficiently small , global tubular neighborhood of @xmath87 are lying in _
one _ tube around @xmath87 because of globally constant distance .
the first item results from theorem [ theotube ] and locally constant distances of leaves , the second is a direct conclusion of the first . to see the third item
, we consider a tubular neighborhood of a specificly regular leaf @xmath202 so small that it is contained completely in the global tubular neighborhood of @xmath100 now , combine the local description of a riemannian foliation with sliding along @xmath202 to see that the third property is open on @xmath201 closedness follows since the complement is open .
`` sufficiently small '' in the fourth item is easily seen to be less than one half of the radius of the global tubular neighborhood ( cf . argument ( ii ) in the proof of theorem [ theotube ] ) .
_ proof of lemma [ lemhol ] .
_ `` @xmath203 '' assume that there is a tubular neighborhood with slice @xmath198 of @xmath69 such that every meeting leaf has exactly one plaque in it .
as the leaves have the same dimension as @xmath204 the intersection of a leaf with the slice is one single point , and sliding along the leaves just produces the identity map .
`` @xmath205 '' consider a leaf @xmath87 with global tubular neighborhood of radius , say , @xmath206 and let @xmath69 be a point of @xmath100 further , let @xmath202 be a leaf with at least two intersection points with the slice @xmath207 denote these points by @xmath208 and @xmath209 sliding along the leaf @xmath87 of the distance @xmath210 to @xmath211 therefore defines a non - trivial element in the holonomy group of @xmath69 in @xmath100 as @xmath202 was arbitrary , the claim follows .
+ now , we will show that the exceptional leaves are not as bad as they may seem at a first glance : the stratum @xmath212 of leaves of trivial holonomy ( i.e. specificly regular leaves ) is open , connected , and dense in @xmath213 beyond , the stratum of exceptional leaves @xmath214 owns at least locally a submanifold structure .
[ propregdense ] let @xmath110 be a singular riemannian foliation with properly embedded leaves .
then , the set of specificly regular leaves is open and dense in @xmath4 . the stratum of regular leaves @xmath14 is open and dense in @xmath19 therefore , it is sufficient to show that @xmath212 is open and dense in @xmath215 by lemma [ lemhol ] , the first claim follows .
let @xmath87 be exceptional , @xmath216 and @xmath217 such that @xmath218 is a global tubular neighborhood of @xmath100 let @xmath219 be arbitrary .
if every leaf intersected the @xmath184-slice of @xmath216 only once , @xmath87 would not be exceptional , so we can find a leaf @xmath202 intersecting @xmath220 times .
assume that @xmath202 is not specificly regular , then iterate the previous consideration as follows : all the points @xmath221 lie in a normal distance sphere of @xmath87 around @xmath190 around each of these points choose local tubular neighborhoods @xmath222 with fixed radius @xmath182 in such a way that they do not intersect , they lie within the @xmath184-neighborhood of @xmath204 and that each contains just one @xmath223 again , have a look at the neighborhoods of one third of this radius . as
@xmath202 is exceptional , there is a leaf @xmath224 meeting each neighborhood several times and therefore meeting @xmath225 more than once , say @xmath71 times .
thus , there are @xmath226 different points of @xmath224 in the @xmath184-slice of @xmath125 and \(i ) the points @xmath227 are at the same distance from @xmath228 \(ii ) all the @xmath226 points @xmath229 lie on a distance sphere of fixed radius around @xmath190 let @xmath2 be the dimension of @xmath4 and @xmath12 the dimension of the regular leaves .
then , the dimension of a normal distance sphere of a point in @xmath14 is equal to @xmath230 the different points of @xmath231 live therefore in an at most
@xmath232-dimensional set .
thus , the iteration argument has to end after @xmath233 steps which means finding a specificly regular leaf .
the leaf @xmath204 the point @xmath125 and the radius @xmath26 were chosen arbitrarly , so the claim follows . to prove that @xmath212 is connected
, we will use the structure of the exceptional stratum , which can be stated as follows : [ lemexmfd ] each connected component of @xmath234 is a submanifold of @xmath14 with smooth ( relative ) interior , namely either being an isolated exceptional leaf or locally being the pre - image of totally geodesic , convex sets of the ( local ) quotient space .
the following proof does not give any information about smoothness of the boundaries of @xmath234 in @xmath14 resp .
in @xmath19 it will follow by lemma [ lemacc ] that every component actually has no boundary ( this fact is not used in the meantime ) .
assume that there is a leaf @xmath87 , to which exceptional leaves accumulate .
it is a direct consequence that @xmath87 itself is exceptional or singular .
let @xmath235 be a global tubular neighborhood and let @xmath236 be exceptional , lying within this neighborhood .
then there exists a horizontal geodesic realising the distance between @xmath236 and @xmath100 by globally constant distance of leaves and by connectedness , we conclude that there is a differentiable strip between the two leaves consisting of minimal geodesics @xmath237 of length @xmath238 .
if @xmath202 is another leaf meeting one of these geodesics at time @xmath239 then @xmath202 meets every such geodesic at time @xmath240 .
this follows by an open - closed argument . to see that @xmath202 lies totally within this strip ,
we argue as follows : take the distance from @xmath241 to @xmath87 as well as the distance from @xmath7 to @xmath236 and look at the associated broken geodesic from @xmath242 to @xmath243 .
piecewise , it is of the same length as the geodesics of the strip , so either it must be unbroken ( and thus contained in @xmath237 ) or the distance from @xmath242 to @xmath243 must be less than @xmath238 , which is impossible .
lemma [ lemposhom ] tells us that @xmath202 has the same dimension as @xmath236 .
let @xmath244 be a global tubular neighborhood of @xmath236 and specify @xmath202 to be a leaf lying within this neighborhood .
choose @xmath245 arbitrarily and @xmath246 lying on the unique minimal geodesic from @xmath247 to @xmath87 ( see the figure below ) . within a slice through @xmath248
the point @xmath220 is the only one belonging to @xmath202 because the distance spheres @xmath249 and @xmath250 intersect only in @xmath251 because @xmath236 is exceptional , for every neighborhood @xmath252 of @xmath220 there is a specificly regular leaf which intersects the slice @xmath253 more than once .
lemumfk ( 24,23)@xmath247(39,17)@xmath220(66,12)@xmath254(60,50)@xmath255(21,39)@xmath256(45,22)@xmath257 but locally constant distance forces this leaf to meet @xmath258 several times , elsewise there would be more points of @xmath202 in @xmath259 by definition , @xmath202 must be exceptional , too .
this consideration holds for every leaf @xmath260 within @xmath261 the stratum @xmath234 is closed .
thus , a boundary leaf @xmath262 also is exceptional . assigning the previous part of @xmath236 to this boundary leaf @xmath202 and repeating the proof yields that all leaves of the strip @xmath263 between @xmath87 and @xmath236 are exceptional . in the local quotient space @xmath194
minimal geodesics connecting arbitrary points of @xmath264 thus lie completely in @xmath265 this set is locally convex .
therefore , its interior is well - defined and can be endowed with a smooth submanifold structure , see p. 139 of @xcite .
hence , the preimage @xmath266 of the submersion @xmath267 is smooth in its interior , too .
[ propregcon ] the set @xmath212 of leaves with trivial holonomy is connected .
assume @xmath268 to be not connected .
the component @xmath269 that separates the stratum @xmath212 has to be @xmath270-dimensional and without boundary .
hence , the fibers of the normal bundle of @xmath271 are one - dimensional .
first case : an exceptional leaf @xmath87 and thus all regular leaves are of dimension @xmath272 . within every local tubular neighborhood ( sufficiently small ) of @xmath273
, there is by lemma [ lemhol ] a regular leaf having several plaques within a corresponding tube .
if @xmath87 separates the specificly regular stratum , @xmath87 separates the tube .
transversally , meaning within the quotient , the tube consists of only two points , which then have to belong to different components of @xmath196 but both belong to the same leaf which is connected , a contradiction .
second case : the component @xmath271 is a union of exceptional leaves . let @xmath274 and @xmath275 be a global tubular neighborhood of @xmath276 by assumption , @xmath277 is separated by @xmath278 beyond , there is a leaf @xmath279 arbitrarily close to @xmath7 which intersects @xmath40 several times .
choose @xmath280 and a path @xmath281 \to n$ ] from @xmath282 to @xmath283 at every point @xmath74,$ ] take the distance @xmath284 this is constant along @xmath71 , hence @xmath285 as @xmath271 is a @xmath270-dimensional submanifold of @xmath261 the points @xmath69 and @xmath42 have to be in different components of @xmath286 but again , both belong to the same leaf , a contradiction . by remark ( iii ) coupled with propositions [ propregdense ] and [ propregcon ]
we conclude : [ cortype ] in a global tubular neighborhood of an exceptional leaf , all specificly regular leaves intersect a slice equally often ( and more than once ) .
the last subsection , particularly the last corollary [ cortype ] , leads to the following , well - defined notion : [ deftype ] the _ type of an exceptional leaf @xmath287 _ is the natural number @xmath288 combining the local description of a singular riemannian foliation with the tube theorem [ theotube ] and corollary [ cortype ] yields [ lemcovering ] let @xmath87 be exceptional of type @xmath289 resp .
specificly regular with global tubular neighborhood @xmath290 then , the orthogonal projection @xmath291 restricted to a leaf @xmath292 is an @xmath293-fold covering resp . a global diffeomorphism
. a first step towards the stratification shall be the following lemma generalising the claim of lemma [ lemexmfd ] as announced .
let @xmath294 be a horizontal geodesic .
[ lemacc ] if @xmath295 is an accumulation point of both exceptional points and specificly regular points along @xmath71 , so @xmath295 must be singular .
this lemma holds for singular riemannian foliations with properly embedded leaves . within the restriction to foliations without horizontally conjugate points
, we have the stronger proposition [ propiso ] . without restriction and by lemma [ lemexmfd ]
we can assume @xmath165 to be so small that @xmath296 is globally defined and @xmath297 only consists of exceptional points whereas on @xmath298 there only lie specificly regular points .
furthermore , assume @xmath114 to be exceptional ( by proposition [ propregdense ] , it is clear that @xmath114 can not be specificly regular ) .
let @xmath299 such that @xmath300 the corresponding leaf @xmath287 meets the tube @xmath296 more than once , namely @xmath301-times ( see the following figure ) .
lemhp ( 17,43)@xmath42(86,43)@xmath69(50,42)@xmath295(94,43)@xmath71(35,55)@xmath302 ( 61,79.5)@xmath303(38,9)@xmath304(-2,55)@xmath305 all intersection points @xmath306 with the slice through @xmath295 lie on the distance sphere @xmath307 we know that @xmath308 and of course @xmath309 sliding of this distance along the leaf @xmath287 to the different points @xmath306 yields the fact that the leaf @xmath310 too , meets the tube @xmath296 several times . because of the fixed distance to @xmath311 the intersection points @xmath312 of @xmath39 and the slice of @xmath295 lie on the sphere @xmath307 by @xmath313 for all @xmath143 wo conclude , that each @xmath304 is the antipode of the corresponding @xmath314 but @xmath39 is exceptional , thus we find specificly regular leaves with several plaques within local tubular neighborhoods around the @xmath315 choose such neighboorhoods so small that neither @xmath295 nor one of the @xmath316 is contained in it .
hence , a specificly regular leaf that defines the type @xmath317 must have at least @xmath318 plaques within @xmath296 around @xmath319 this contradiction shows that @xmath114 is singular . from the beginning , we have called the set @xmath234 a stratum ; now we will show that this intuitional naming is correct in the sense of the following common definition . a locally finite partition of a manifold @xmath4 into locally closed ,
connected submanifolds @xmath320 of @xmath4 , the _ strata , _ is called a _ stratification _
if the following condition holds : for every @xmath321 denote the closure of @xmath322 by @xmath323 then @xmath324 recall that in the case of a singular riemannian foliation coming from the orbit decomposition of a proper group action , the subdivision into orbit types is not only a stratification but also fulfills the tougher technical conditions of a whitney stratification .
we conjecture that this is the case in general .
the last argument to do for the stratification by types of leaves as in [ deftype ] is [ lemacctype ] if @xmath7 is an accumulation point of exceptional points @xmath325 and not singular , then @xmath326 is exceptional of type @xmath327 beyond , the function @xmath328 defined along a horizontal geodesic @xmath329 is constant except for isolated points , i.e. nearly all leaves intersecting @xmath71 are of the same type , the remaining points are of higher type . for every tubular neighborhood of @xmath7
there is an @xmath330 such that the exceptional leaves @xmath331 meet that neighborhood .
additionally , for every @xmath332 there are specificly regular leaves , meeting the neighborhood @xmath333-times , namely those who define the type of @xmath334 by assumption , the leaves are properly embedded , therefore the number @xmath335 and thus the right hand side of the inequality are finite , and the first claim follows .
now , restrict the attention to a connected subset @xmath336 and define the set of leaves of minimal type , which is open in @xmath337 using the methods of the proof of proposition [ propregdense ] , as there are taking thirds of tubular neighborhoods , sliding of distances , and iteration , we can interpret these leaves as _ regular in @xmath57 , _ i.e. they have to meet tubular neighborhoods of exceptional leaves of higher type more than once . hence
, if two different types of leaves accumulate to one point along a horizontal geodesic in @xmath338 we can carry over the lemma [ lemacc ] directly and conclude that the accumulation point has to be singular , a contradiction .
with the work already done , the technical lemma [ lemacctype ] can be reformulated as [ proptype ] the decomposition of @xmath234 with respect to the types of the leaves @xmath339 is a stratification .
molino has shown that the decomposition of @xmath340 with respect to the dimension of the leaves is a stratification .
thus , we conclude with the help of lemma [ lemacctype ] and proposition [ proptype ] : [ theostratification ] the finer decomposition of a singular riemannian foliation @xmath341 with properly embedded leaves into the sets of specificly regular leaves , strata of types of exceptional leaves , and dimensional strata of singular leaves is a stratification .
the notion of a foliation without horizontally conjugate points was introduced by lytchak and thorbergsson in @xcite as follows .
[ defwhcp ] a singular riemannian foliation is _ without horizontally conjugate points _ if the following holds for all leaves @xmath87 and all geodesics @xmath56 perpendicular to @xmath87 : every @xmath87-jacobi field along @xmath56 that is tangential to a leaf different from @xmath87 is tangential to every leaf met by @xmath140 we call a point @xmath342 _ horizontally conjugate to @xmath7 _ if there is a horizontal geodesic @xmath113 \to m$ ] with @xmath343 and an @xmath6-jacobi field @xmath88 along @xmath56 such that @xmath344 but @xmath88 is not everywhere tangential .
let @xmath345 be a singular riemannian foliation and let @xmath56 be a horizontal geodesic .
as in @xcite we define the vector space @xmath346 wherby the @xmath240 s cover the domain of definition of @xmath140 we will call a jacobi field @xmath88 along @xmath56 _ everywhere tangential , _ if for all @xmath240 it is tangential to the leaf @xmath347 thus , every @xmath348 is everywhere tangential .
at least if the foliation has no horizontally conjugate points , we can reverse this implication as follows .
[ lemjacobifield ] let @xmath345 be a singular riemannian foliation without horizontally conjugate points .
if @xmath88 is an @xmath349-jacobi field that is everywhere tangential along @xmath92 then @xmath88 is an @xmath350-jacobi field for all @xmath351 , i.e. @xmath348 . [ prop3.1 ] this lemma is a non - mentioned detail of the proof of proposition 3.1 in @xcite which states that , within the notation used above , @xmath352 in that article is shown that for every @xmath353 in the domain of @xmath56 and for every @xmath354 there is a @xmath349-jacobi field @xmath88 which is tangential to the corresponding leaf through @xmath94 for all @xmath240 . to prove this , a curve @xmath355 with @xmath356
is lifted to a neighboring leaf @xmath202 within a tubular neighborhood of @xmath357 via the local riemannian submersion @xmath358 the geodesics pointwise connecting both curves define a horizontal variation @xmath359 of @xmath140 the appropriate jacobi field @xmath88 along @xmath56 is tangential twice , namely in @xmath349 and in @xmath360 hence by assumption everywhere . as @xmath361 lies in @xmath362
there actually is a family of jacobi fields @xmath363 ( along the geodesics @xmath364 ) , that are tangential to the leaves @xmath365 for all @xmath96 _ proof of lemma [ lemjacobifield ] . _
fix @xmath240 and define @xmath366 to be a path through the variation @xmath367 at time @xmath368 . then , @xmath369 we now verify that if necessary diminuish @xmath370 for all @xmath371 the geodesics @xmath372 meet the leaf @xmath350 at time @xmath373 for this purpose let @xmath374 be a point with @xmath375 without loss of generality consider @xmath376 to be in the interval @xmath377 .
there is a neighborhood of @xmath376 in @xmath4 containing leaves of the same or higher dimension only . by maximality of the leaf dimension in @xmath378 along @xmath379
there is a @xmath380 with @xmath381 .
restrict the attention to the stratum @xmath382 to see that an everwhere tangential path can not cross several leaves : for @xmath383 there is a local , centered chart @xmath384 such that @xmath385 and if @xmath87 is a leaf with @xmath386 then @xmath387 denote the composition @xmath388 by two components @xmath389 . as @xmath390
, we conclude @xmath391 and @xmath392 for all @xmath393 let @xmath394 be maximal in the sense that @xmath395 is contained in the leaf @xmath396 this interval then is open as seen above .
we now look at the boundary point @xmath397 consider a tubular neighborhood of the leaf @xmath398 , wherein the plaques of the leaves have locally constant distance to @xmath398 .
the path @xmath399}$ ] is compact , hence @xmath71 crosses finitely many plaques during this time .
thus , there is a @xmath400 such that the distance from @xmath401 to @xmath402 is bigger than or equal to @xmath403 for all @xmath404 contradicting continuity of @xmath71 .
hence , @xmath405 as @xmath364 is perpendicular to @xmath349 for all @xmath406 @xmath364 is perpendicular to @xmath407 therefore , @xmath367 is a @xmath408-variation , and @xmath88 is a @xmath408-jacobi field .
+ the preceding considerations reveal for singular riemannian foliations [ proptangentialcurve ] a curve @xmath409 that is always tangential to the leaves , i.e. for all @xmath410 holds lies in one leaf . with regard to lemma [ lemjacobifield ] and remark [ prop3.1 ]
, we now can define negative homothetical transformations as follows .
[ prophomotheties ] if @xmath345 is a singular riemannian folation without horizontally conjugate points , then 1 . except at isolated points , @xmath411 along any horizontal geodesic @xmath140 2 .
the homothetical transformations @xmath412 also exist for negative @xmath413 and respect the foliation .
particularly , the _ reflection _ at a leaf respects the foliation . to prove the first item ,
let @xmath357 be such that the leaf through @xmath357 is of maximal dimension ( along the geodesic ) ; let @xmath414 be of smaller dimension . by maximality , we can choose @xmath415 in such a way that @xmath416 now consider a local tubular neighborhood of @xmath417 of normal radius @xmath403 and its intersection with the stratum @xmath418 . following @xcite , p. 196
, this is a union of geodesic arcs of length @xmath419 perpendicular to @xmath420 as @xmath56 is perpendicular to @xmath421 in @xmath422 too , we have @xmath423 i.e. the geodesic arc @xmath424 is either contained completely in the stratum @xmath418 or not at all except the point @xmath425 the first inclusion can not be fulfilled as @xmath426 and we assumed the strata to be different . because of the existence of positive homothetical transformations , it is only left to show that there is an @xmath427 such that @xmath428 denote the maximal dimension of leaves along @xmath56 by @xmath13 and the dimension of @xmath421 by @xmath238 .
the intersection of a slice through @xmath131 with a plaque of a leaf @xmath429 is @xmath430-dimensional and lies within a @xmath431-sphere around @xmath131 .
all geodesics beginning in @xmath131 and ending in @xmath429 at the same time as @xmath56 form a @xmath430-dimensional variation of horizontal geodesics which can be continued up to infinity .
the negative starting vectors are of course tangential to the negatives of the curves and span a @xmath430-dimensional submanifold in @xmath432 within the tubular neighborhood , we can assume the existence of unique geodesics between arbitrary points , and the negative variation up to @xmath433 has to stay @xmath430-dimensional .
hence , there is only one leaf meeting the variation at time @xmath433 and this is @xmath12-dimensional which means maximal .
the second item now is easily seen : the proof of ( i ) shows the existence of the homothetical transformation @xmath434 at @xmath421 .
compose this with positive ones . [ corcovering ]
let @xmath87 be a leaf in @xmath4 with global tubular neighborhood @xmath290 then , the homothetical transformations @xmath435 ( as long as @xmath436 and @xmath182 are defined ) map leaves diffeomorphically onto each other .
consider a leaf @xmath437 by proposition [ prophomotheties ] , @xmath412 locally maps open sets of @xmath438 diffeomorphically to open sets of other leaves .
the image @xmath439 of an arbitrary path @xmath281 \to l_1 $ ] is tangential to the leaves as a path through the variation of horizontal geodesics defining @xmath440 hence , @xmath412 maps @xmath438 into one leaf @xmath441 further , @xmath442 is injective .
assume that there is @xmath443 contained in the closure of @xmath444 but not in its image .
then there is a variation through horizontal geodesics whose endpoints in @xmath445 converge to @xmath190 the limit curve is a horizontal geodesic @xmath56 beginning in @xmath438 and meeting @xmath204 too .
thus , @xmath446 is a preimage of @xmath69 with respect to @xmath412 which means that @xmath447 is open and closed in @xmath448 hence equal to @xmath441 in @xcite and @xcite is shown ( by completely different methods ) that the reflection respects the foliation in general , i.e. without the assumption of no horizontally conjugate points . by theorem [ theotube ]
we thus generalize : _ for a singular riemannian foliation with properly embedded leaves , the homothetical transformations defined within global tubular neighborhoods are diffeomorphisms between the leaves . _ due to an essential hint of a. lytchak
, we can specify the last corollary even to the types of leaves , which is stated in the next proposition .
thus , if the foliation has no horizontally conjugate points , a situation as in lemma [ lemacc ] can not occur which means : if there lie specificly regular points on a horizontal geodesic , singular points as well as exceptional points are isolated along this geodesic , and lemma [ lemacctype ] does not only hold in @xmath234 but in @xmath4 itself .
[ propiso ] if @xmath345 is a singular riemannian foliation without horizontally conjugate points and with properly embedded leaves , then the homothetical transformations respect the types of leaves , i.e. they map a leaf @xmath87 of type @xmath289 diffeomorphically onto a leaf of the same type
. as the leaves are properly embedded , the foliation is closed .
we can then use theorem 1.7 of @xcite to see that the quotient @xmath449 is isometric to a good riemannian orbifold @xmath450 without conjugate points .
the manifold @xmath87 and the discrete group @xmath451 are constructed in such a way that the stratification by types of leaves of @xmath4 determines the group @xmath451 of isometries of @xmath87 ( for details see @xcite ) .
because of our lemma [ lemacctype ] , it only remains to check wether the reflection , i.e. the homothetical transformation @xmath452 at a singular leaf respects the types of the neighboring leaves . and
this is true on @xmath4 because it is an obvious fact on @xmath87 stratified by the action of @xmath453 we close this subsection by citing a lemma proven by lytchak and thorbergsson in @xcite , subsection 4.5 , which we will need later .
[ lemtangentialjf ] let @xmath345 be a singular riemannian foliation , a horizontal geodesic
. denote by @xmath454 the space of _ horizontal jacobi fields _ along @xmath92 i.e. all jacobi fields that come from variations of @xmath56 through horizontal geodesics in which all geodesics meet the same leaves at the same time .
for every @xmath455 @xmath454 then spans the tangential space of the leaf @xmath456 @xmath457 in particular , @xmath458 for all @xmath459 the preceding section already points at the possibility to characterise singular points locally by variations through horizontal geodesics which leads to the notion of focal points . in this chapter , we will now have a deeper view onto the relation of these attributes .
for the following , let @xmath345 be a singular riemannian foliation without horizontally conjugate points ; let @xmath56 be a horizontal geodesic in @xmath4 meeting the leaf @xmath114 at time @xmath460 furthermore , consider @xmath114 to be of maximal dimension along @xmath140 then , [ lemfocalpoint ] @xmath461
@xmath205 : this can be seen in the proof of proposition [ prophomotheties ] : for a leaf @xmath462 near @xmath463 we find a @xmath430-dimensional variation of horizontal geodesics starting in @xmath94 that is everywhere tangential . a singular point on @xmath56
thus is a focal point to every leaf of higher dimension met by @xmath140 @xmath203 : let @xmath464 denote a non - trivial variation of horizontal geodesics starting in @xmath135 and meeting the leaf @xmath465 in @xmath466 the corresponding jacobi field is tangential in @xmath66 and in @xmath94 , hence everywhere .
therefore , @xmath464 is an @xmath467-variation for every leaf met by @xmath56 at time @xmath468 with focal point @xmath463 if @xmath467 is of higher dimension than @xmath347 again , take a tubular neighborhood of @xmath469 so small that there exist unique geodesics between arbitrary points , and let @xmath26 be such that @xmath470 is contained in it .
then , @xmath471 for all @xmath406 where @xmath472 denotes the normal sphere of radius @xmath26 around @xmath466 assuming @xmath350 to have maximal dimension now leads to a contradiction : the leaf @xmath473 must have the same dimension and @xmath474 while the variation is non - trivial and its geodesics are unique , so @xmath475 in a singular riemannian foliation ( without additional conditions ) , singular points are easily seen to be focal points of leaves nearby .
but in general , there are focal points in regular leaves , too !
a trivial example is the sphere @xmath476 with the standard metric , foliated by points .
geodesics are always horizontal , and conjugate points are horizontally conjugate . obviously , they are focal points to each other , and regular by definition . with regard to the last proof and the proof of proposition [ prophomotheties ] , we can improve lemma [ lemfocalpoint ] as follows .
[ propsingfocal ] for every singular point @xmath7 in @xmath4 and every regular leaf @xmath87 there is a geodesic @xmath113 \to m$ ] with @xmath477 that is perpendicular to @xmath478 and @xmath7 is a focal point of @xmath479 along @xmath56 of multiplicity @xmath480 in particular , the multiplicity of @xmath7 is independent of the choice of @xmath140 we just drop a perpendicular from @xmath7 to @xmath481 let be a geodesic beginning in @xmath482 perpendicularly to @xmath478 ending in @xmath483 as in the proof of lemma [ lemfocalpoint ] , there is a local variation of @xmath56 of maximal dimension @xmath484 which can be extended up to the leaf @xmath485 if this variation was not @xmath484-dimensional in @xmath479 the point @xmath17 would be a focal point along @xmath56 ( namely a focal point of the leaf @xmath473 in the notation used above ) which contradics the regularity of @xmath485 combining the latest findings with the general form of proposition [ prophomotheties ] part ( i ) as proven in @xcite and @xcite , yields the main theorem of this chapter , the geometric characterisation of foliations without horizontally conjugate points .
[ theowhcp ] a singular riemannian foliation has no horizontally conjugate points if and only if the singular points are exactly the focal points of leaves . to prove the if - part
, we will need the following technical lemma that holds under the assumption of singular points being the focal points .
[ lemdisturb ] let @xmath486 be a horizontal geodesic .
if we have @xmath487 such that @xmath488 is horizontally conjugate to @xmath489 along @xmath92 for every @xmath165 ( sufficiently small ) there is a @xmath490 and @xmath491 such that @xmath492 is horizontally conjugate to @xmath493 we define @xmath494 in analogy to @xcite and @xcite , we can define a bundle @xmath495 over @xmath496 by the following extension of @xmath497 at singular points : @xmath498 then , @xmath499 if @xmath240 is regular by the assumption that singular points are focal points . as we know the latter to be isolated along @xmath92 we see that @xmath495 is a smooth bundle , and @xmath500 for all @xmath240 wherby @xmath357 is regular ( see proposition 3.2 in @xcite ) .
define the horizontal part @xmath501 to be the orthogonal complement of @xmath502 in @xmath503 and denote the bundle by @xmath504 wilking showed that there is an associated morse sturm differential equation to @xmath505 in @xmath506 such that its space of solutions @xmath507 determines just the horizontal jacobi fields along @xmath56 which vanish in @xmath508 by the generalized index theorem of morse , the corresponding index form @xmath509 has @xmath510 } \mid j \in \mathcal l , j(b ) = 0\}\quad \text{and}\quad { \rm ind}(i_{a , b } ) = \sum_{t \in ( a , b ) } \nu(t).\ ] ] here , @xmath511 lie in @xmath512 , \mathbf r^k),$ ] the space of absolutely continuous , @xmath513-valued maps of @xmath514 $ ] with square integrable derivative and boundary condition @xmath515 ; @xmath516 is continuous and @xmath517 is self - adjoint for all @xmath240 ( here , we use the notation of lytchak and thorbergsson in @xcite which we refer to for details ) . without loss of generality
let @xmath488 be the first conjugate point to @xmath489 .
for every @xmath165 sufficiently small , let @xmath518 , \mathbf r^k)$ ] be a maximal subspace on which the index form is negative definite @xmath519 the morse sturm differential equation and thus the index form depend continuously on the end points of the interval .
beyond , there is a neighborhood around @xmath520 that contains leaves of same or higher dimension only , so by the index theorem the following holds for all @xmath490 sufficiently small : @xmath521 where @xmath522 , \mathbf r^k)$ ] with boundary condition @xmath523 is a subspace of dimension @xmath524 thus , continuity implies @xmath525 we can state @xmath526 more precisely : for @xmath527 there is a horizontally conjugate point @xmath492 with @xmath528 by assumption , there is no horizontally conjugate point between @xmath489 and @xmath529 for all @xmath530 and all @xmath531 , \mathbf r^k),$ ] we therefore have @xmath532 if @xmath182 in @xmath526 is chosen to be sufficiently small , continuity implies @xmath533 hence , @xmath527 has no horizontally conjugate point on @xmath534 _ proof of theorem [ theowhcp ] . _ with regard to lemma [ lemfocalpoint ] it remains to show that if focal points are singular , then the foliation is without horizontally conjugate points .
we prove this by contradiction .
assume that there are horizontally conjugate points , i.e. there is a horizontal geodesic @xmath113 \to m,$ ] points @xmath535 $ ] and a @xmath135-jacobi field @xmath88 along @xmath56 such that @xmath536 but @xmath537 by lemma [ lemtangentialjf ] , there exists a @xmath114-jacobifeld @xmath538 along @xmath56 that is everywhere tangential and @xmath539 the difference @xmath540 is a @xmath114-jacobi field that vanishes in @xmath353 but not identically .
hence , @xmath357 is a focal point of @xmath319 wilking shows in @xcite , section 3 , that then @xmath541 , i.e. @xmath540 is a @xmath349-jacobi field .
now , use lemma [ lemtangentialjf ] again : there is @xmath542 such that @xmath543 is a non - trivial @xmath349-jacobi field vanishing in @xmath460 hence , @xmath66 is a focal point of @xmath114 and thus singular . by lemma [ lemdisturb ]
, we can disturb horizontally conjugate points , so they are not isolated .
but for every @xmath490 ( notation as in [ lemdisturb ] ) , @xmath66 as well as @xmath544 lie in singular leaves and thus have to be isolated along @xmath92 a contradiction ! [ korbolton ] a singular riemannian foliation with regular leaves of codimension one has no horizontally conjugate points .
let @xmath164 be arbitrary .
bolton shows in @xcite that the critical points of the normal exponential map @xmath545 are just the singular points if the foliation is of codimension one .
a proper embedding @xmath546 is called _ taut _ ( with respect to @xmath547 ) , if the energy functional @xmath548 is @xmath547-perfect for every @xmath7 in @xmath29 that is not a focal point of @xmath204 where @xmath549 \to m \text { in } h^1\ \mid\ ( c(0 ) , c(1 ) ) \in ( n , p)\}\ ] ] denotes the hilbert space of @xmath550-paths starting in @xmath87 and ending at @xmath21 see @xcite and @xcite .
it is well known that the energy functional is a morse function whose critical points in @xmath551 are exactly the geodesics starting perpendicularly to @xmath87 and parametrized proportionally to arc length .
now , morse s index theorem tells us that the index of such a geodesic @xmath56 is equal to the sum of multiplicities of the focal points along @xmath140 if @xmath552 denotes the sublevel set of paths with energy less than or equal to @xmath553 then the number @xmath554 of critical points of index @xmath12 of @xmath555 in @xmath556 is finite .
let @xmath557 be the @xmath12-th betti number of @xmath556 with respect to @xmath547 coefficients , so for all @xmath558 we can state the weak morse inequalities @xmath559 a morse function is called _ @xmath547-perfect , _ if equality holds for all @xmath558 and all @xmath560 @xmath561 if @xmath87 is taut , the focal points of @xmath87 along a horizontal geodesic determine the @xmath547-homology of the space @xmath551 completely . to show tautness of an immersion , it is a well working strategy to construct linking cycles explicitly , as done by morse @xcite , p. 244f , and in the famous paper of bott and samelson @xcite ( for a good exposition of this theory see @xcite , for the construction of such cycles in special cases see also @xcite , @xcite , and @xcite ) .
as we will do so , too , let us shortly review this construction : the morse inequalities reveal the deep connection between the @xmath12-th homology group of @xmath556 and the geodesics of energy @xmath558 with index @xmath562 given such a geodesic , we need to construct an appropriate , non - trivial class , the linking cycle , in @xmath563 out of the data of the geodesic , i.e. focal points and their multiplicities , to obtain equality . as above
we define @xmath564 it is a well known fact that @xmath556 is homotopy equivalent to @xmath565 if @xmath558 is not critical . and if @xmath558 is critical , there is an @xmath566 such that @xmath567 where @xmath568 denotes the number of critical points of @xmath555 with index @xmath12 and value @xmath569 then the long exact homology sequence tells us that @xmath555 is perfect iff @xmath570 is surjective for all @xmath562 let @xmath56 be a critical point of @xmath555 with value @xmath571 let @xmath572 such that @xmath573 are just the focal points of @xmath87 along @xmath140 then @xmath574 to show surjectivity in @xmath526 it suffices to construct for every @xmath56 a compact , @xmath575-dimensional manifold @xmath576 and a smooth map @xmath577 whereby the following conditions hold : 1 .
@xmath578 for exactly one @xmath579 and there is a neighborhood @xmath57 of @xmath56 such that @xmath580 is an immersion , 2 .
@xmath56 is the only critical point of @xmath555 in the image of @xmath413 , and 3 .
@xmath581 for all @xmath582 by the negative gradient flow of @xmath583 the image @xmath584 can then be deformed into a smooth cycle @xmath585 in @xmath586 which represents a non - trivial class in @xmath587 conditions ( i ) and ( ii ) guarantee that the image of this class in @xmath526 is non - trivial , too .
so , for every @xmath56 we would have found a linking cycle @xmath588 which shows that @xmath589 is surjective for every @xmath562 thus , tautness of @xmath87 follows .
let @xmath87 denote a leaf and @xmath551 the corresponding space of paths ending in a point @xmath7 that is not a focal point of @xmath87 .
critical points of the energy integral @xmath555 are horizontal geodesics from @xmath87 to @xmath7 with index equal to the sum of focal points counted with multiplicities .
hence , the easiest case for which we have to construct a linking cycle is a geodesic with only one focal point that is of multiplicity one .
if all such critical points of index one are completable , we call @xmath87 _ @xmath0-taut . _ as a first step towards the general construction we show at first , let @xmath87 be a regular leaf and let @xmath590 \to m$ ] be a critical point of @xmath555 as above , i.e. a horizontal geodesic with exactly one focal point @xmath357 of multiplicity one . by theorem [ theowhcp ]
, @xmath357 is singular .
hence , by lemma [ lemacc ] and lemma [ lemacctype ] , we can conclude that except isolated points , all points along @xmath591}$ ] are of the same type .
furthermore , the isolated points are of higher type .
choose a global tubular neighborhood @xmath592 and @xmath32 such , that @xmath593 is specificly regular resp . of minimal type along @xmath140
then , @xmath594 is a compact , properly embedded submanifold of dimension @xmath595 of @xmath596 if it is not connected , we choose the component topologically a circle that containts @xmath593 and call it @xmath597 now , define @xmath598 \to m , \text { with } \left\{\begin{array}{l } f(s , t_0 ) = \gamma(t_0 ) \text { for all } s \in k\\ \text{and } k = \{f(s , t_0-\delta ) \mid s \in k\},\end{array}\right.\ ] ] to be the one - dimensional , smooth variation of horizontal geodesics of the same length as @xmath599}.$ ] this variation is unique because the paths @xmath600}$ ] are minimal from @xmath601 to @xmath357 in @xmath4 for all @xmath602 the path through the variation at time @xmath603 is tangential to the leaf @xmath596 obviously , @xmath604 is tangential to @xmath349 for all @xmath602 thus , @xmath605 and by proposition [ proptangentialcurve ] we have @xmath606 for all @xmath602 moreover , the set @xmath607 is a compact submanifold ( possibly not properly embedded ) of @xmath87 as image of @xmath608 cover @xmath609}$ ] by global tubular neighborhoods similar to the covering in the proof of theorem [ theotube ] .
for this purpose , let the centers be the leaf @xmath87 as well as all the leaves of higher type than @xmath610 and , if these tubular neighborhoods do not yet cover @xmath609}$ ] , some specificly regular leaves ( resp .
leaves of the same type as @xmath611 ) .
thus , each point @xmath612}$ ] lies within exactly one neighborhood or within the section of two consecutive neighborhoods .
denote the centering leaves by @xmath613 with @xmath614 the homotheties defined within the global neighborhoods are diffeomorphisms by corollary [ corcovering ] , and with the help of these maps , we move @xmath601 along the geodesics @xmath615 from @xmath611 to a leaf ( of the same type ) in the intersection @xmath616 then to a leaf in @xmath617 etc . up to a leaf @xmath171 in the intersection
@xmath618 the image of @xmath601 in @xmath619 i.e. the set @xmath620 is mapped to @xmath621 via the orthogonal projection @xmath622 in the last step , hence we loose embeddedness in general . now define an injective , smooth map @xmath623 by @xmath624,\\ \gamma(t ) , & t \in [ t_0 , 1 ] . \end{array}\right.\ ] ] the compact image @xmath584 is an injective , smooth cycle of dimension @xmath625 in @xmath626 consisting of broken geodesics of same length and thus of same energy .
the only non - broken geodesic is @xmath140 hence , the critical point @xmath56 is completable . for singular leaves
, we can use an idea based on proposition 2.7 in @xcite , which in the general case of foliations is due to s. wiesendorf , see @xcite : by definition , a leaf @xmath87 is taut , if the energy functional @xmath555 is perfect for @xmath7 not focal , but by a limiting argument , it suffices to show perfectness for all @xmath7 that are regular .
thus , along a horizontal geodesic , nearly all points are regular . to show that a singular leaf @xmath87 is taut
, we construct a linking cycle @xmath584 of a regular leaf @xmath171 nearby and then add to every broken geodesic in @xmath584 a minimal geodesic connecting the endpoint of the geodesic in @xmath584 with @xmath100 the corresponding map @xmath627 is injective and preserves the indices of geodesics as critical points of @xmath628 furthermore , the energy levels will just be shifted by a constant .
the new set of broken geodesics , @xmath629 by construction can be deformed into a linking cycle of @xmath100 tben has shown in @xcite that a singular riemannian foliation with properly embedded leaves and with flat sections in a simply connected symmetric space does never have exceptional leaves . as an application of the previous theorem , we generalize this to [ propsimplycon ] if @xmath4 is a complete , simply connected manifold and if @xmath345 is singular riemannian without horizontal conjugate points and with properly embedded leaves , there are no exceptional leaves in @xmath630 let @xmath87 be a leaf and @xmath631 there is the fibration @xmath632 where @xmath281 \to m$ ] is a path with @xmath633 the fiber @xmath634 over @xmath635 that is the space of paths from @xmath636 to @xmath21 is homotopy equivalent to @xmath637 which is connected as @xmath4 is simply connected .
the long exact sequence of homotopy groups of fibrations yields that @xmath551 is connected : assume that there is an exceptional leaf @xmath638 then let @xmath202 be a leaf with trivial holonomy meeting a tubular neighborhood of @xmath639 several times . under the boundary condition
@xmath640 the energy integral thus must have at least two local minima .
now use the minimax principle : there is a critical point of index one , i.e. a geodesic @xmath71 from @xmath202 to @xmath69 with just one focal point , and this critical point connects two components of @xmath641 in @xmath642 by theorem [ theo0taut ] , there is a one - dimensional linking cycle @xmath643 topologically a circle , in @xmath644 with @xmath645 and equality holds if and only if @xmath646 this leads to a contradiction because @xmath647 is connected but , as @xmath648 is a representative in @xmath649 that maps non - trivially to @xmath650 , @xmath647 has to meet both components of @xmath641 that are only connected by @xmath73 at first , let @xmath87 be regular . for each focal point @xmath69 of @xmath87 along a horizontal geodesic @xmath651
we find a corresponding manifold @xmath652 of dimension @xmath653 within a neighboring leaf @xmath171 that is specificly regular resp . of minimal type
along @xmath654 as in the proof of theorem [ theo0taut ] and a global variation @xmath655 of geodesics such that @xmath656 for all @xmath657 because of proposition [ propsingfocal ] .
proposition [ propiso ] tells us that leaves of different dimension or different type than @xmath171 are isolated along @xmath654 and therefore , we can carry over the whole proof of the previous theorem : compactness of the image of @xmath652 in @xmath87 follows analogously as long as we choose all these singular and exceptional leaves of higher type along @xmath654 to be centers of tubular neighborhoods of the covering .
hence , the only work to do regarding our theorem [ theotaut ] is an iteration of this idea . as in @xcite
, we can use a fiber bundle construction . for all @xmath659 let @xmath660 \to m$ ] denote the @xmath661-dimensional variation of geodesics from @xmath87 to @xmath662 parametrized by @xmath663 ( similar to the proof of theorem [ theo0taut ] ) .
this variation consists of all minimal geodesics between @xmath663 and @xmath662 continued up to the length @xmath664}).$ ] in this situation holds [ lemfixedfocal ] if @xmath665 is a focal point of @xmath87 along @xmath92 then @xmath666 is a focal point of @xmath87 of same multiplicity along @xmath667 for @xmath668 and for all @xmath669 moreover , we have @xmath670.\ ] ] as remarked in @xcite , focal points along geodesics within such a variation are not fixed in general ( i.e. in an arbitrary singular riemannian foliation ) , neither are their multiplicities . although easily proven with the help of section [ s : whcp ] here , the lemma is essential to guarantee that the iterated bundle will be a manifold of dimension @xmath575 as we will see now .
knowing that along all geodesics of the variations @xmath673 there are the same focal points at the same time as along @xmath56 we can define @xmath674 where @xmath675 denotes the manifold of dimension @xmath676 which focalizes to the point @xmath677 along the geodesic @xmath678,$ ] for all @xmath679 and by the lemma [ lemfixedfocal ] , @xmath601 is a compact manifold , a fiber bundle , of dimension @xmath680 now define the map @xmath623 as follows : @xmath681 , i = 1 , \ldots , k\\ \gamma|_{[t_1,1]}(t ) , & t \in [ t_1 , 1 ] .
\end{array}\right.\ ] ] the cycle @xmath584 thus consists of broken geodesics , all of same length ; the only one that is non - broken and therefore the critical point of @xmath555 is @xmath56 itself .
furthermore , @xmath584 has dimension @xmath575 and is injective in the path space because each step of the iteration , i.e. of the bundle construction , yields only different broken geodesics .
hence , all requirements for a linking cycle for @xmath56 are fullfilled . by arbitraryness of @xmath92
the claim follows for regular leaves @xmath100 for a singular leaf @xmath87 on @xmath56 , the argumentation is the same as above in the proof of theorem [ theo0taut ] : assume @xmath7 to be regular and construct the linking cycle of a regular leaf @xmath171 near @xmath87 . then , add the geodesics connecting the endpoints of the cycle in @xmath171 with @xmath87 .
the new manifold of broken geodesics fulfills all conditions for a linking cycle for @xmath140 assume that the leaf @xmath87 in @xmath682 is a @xmath683-torus , and the critical point in @xmath551 is a horizontal geodesic @xmath56 as in the figure on the right .
the index of @xmath56 is two , because there are the two focal points @xmath684 and @xmath685 each of them with multiplicity one .
the focal manifold @xmath686 can be identified with the outer latitude of the torus .
there is an one - dimensional variation @xmath689 through geodesics starting perpendicularly in @xmath690 and ending in the focal point @xmath691 on each of the geodesics lies exactly one focal point like @xmath685 on @xmath140 | for a singular riemannian foliation whose leaves are properly embedded , we show in the first part of this article the existence of global tubular neighbourhoods , and we develop a global description of the foliation as stratification by types of leaves .
the second part deals with the further restriction to a foliation without horizontal conjugate points , introduced by lytchak and thorbergsson , which in the special case of an isometric group action equals the concept of variationally completeness .
therefrom , we deduce a global geometric description the focal points of the leaves are exactly the singular points as well as a topological one : tautness of the leaves . |
several studies have contributed to show the importance of black hole rotation in agn jet formation .
agn jet classification mainly relies on orientation effects and relativistic doppler boosting .
however , they can not explain neither the dichotomy between radio loud and radio quiet agn , nor the difference between fri and frii jets .
there are two main theories to interpret the different characteristics of radio loud and radio quiet galaxies .
the morphological differences may be explained by the different physical properties of the environment in which the relativistic jet propagates @xcite .
the dichotomy can also be explained by involving a difference in the nature of the central engine , the spin of the central black hole , the accretion rate and the jet composition @xcite .
the discovery of two sub - classes of frii galaxies does not allow to solve the problem of the dichotomy .
the hybrid morphology radio sources have two radio lobes which exhibit a different fr morphology and can not be explained without external medium or jet power differences between the two sides of the host galaxy @xcite . in double - double radio galaxies
multiple pairs of lobes are seen and are interpreted as different episodes of jet activity @xcite leading to the possibility that jet interruption occurs in radio galaxies .
finally , a combination of external and engine factors has to be invoked to explain the fri / frii dichotomy , as we have suggested in @xcite .
however , in this paper we could not study the effect of the black hole spin because the analytical model was based on a schwarzschild metric . the dichotomy between radio loud and radio quiet sources
has a result of the spin of the black hole has been explored analytically and numerically @xcite .
moreoever @xcite have recently demonstrated that there is a strong correlation between the power of the jet and the spin of the black hole .
the spin of the black hole may also explain the high precession that is observed in some jets @xcite .
in fact , as already suggested by @xcite , the rotational energy of the central black hole is a tremendous spring for energy invoked to explain emission of plasma flows .
they were the first to propose a magnetospheric model in the force free limit as a source for extracting rotational energy from the black hole .
this model has been extensively discussed in the frame of ideal mhd @xcite .
the energy extracted under the form of poynting flux depends on the rotational speed of the magnetic fieldlines , the spin of the black hole , and as shown recently by @xcite , on plasma injection on magnetic field lines .
other 1d models allowed to study force free magnetospheres around black holes , first in schwarzschild metrics by @xcite and in kerr metrics by @xcite , with a more complexe magnetic configuration and taking into account the connexion between the magnetosphere and the accretion disk .
this last study shows that the higher the black hole spin the smaller the magnetic dead zone of the magnetosphere .
s. komissarov was the first to perform numerical simulations of the blandford znajek scenario @xcite .
besides the strong acceleration that they undergo , agn outflows appear to be highly collimated .
the question we want to address is whether there is a correlation between the collimation of the jets and the black hole spin .
we propose a model for a rotating black hole based on our previous steady axisymmetric analytical model @xcite .
the problem of collimation in all its complexity must include the interaction between the jet and the external medium ( see e.g. levinson & begelman @xcite , and references therein ) . in the following ,
we focus on the self - collimation processes where it is the combination of gas pressure and magnetic fields that acts to confine the flow , as illustrated in fig.[fig1 ] .
magnetic - self confinement requiere an ordered magnetic field anchored onto the black hole or the inner disk region .
the poloidal component of the lorentz force has a collimating effect @xcite .
therefore , we need to measure separately the contribution of the gas pressure and magnetic fields . in the framework of the @xmath0-self - similar model of @xcite , we solve the grmhd equations in the kerr background . in this model ,
two parameters give the magnetic and pressure confinement , as explained in @xcite .
this model has been successfully applied to model agn spine jets @xcite and the criterion for magnetic collimation derived in the case of a non rotating black hole . in this paper
we investigate the effect of the black hole spin on the efficiency of the magnetic confinement .
illustration of the self - collimation processes . ]
the paper is organized as follows .
the model is introduced in sect.[model - des ] . using the @xmath1 formalism @xcite we present the general equations that describe an axisymmetric , stationary , ideal mhd flow in the gravitational potential of a kerr black hole .
the complete set of 3 + 1 equations is given by @xcite in schwarzschild geometry .
we compile in sect.[equations ] the complete set in kerr geometry and obtain an original formulation of the energy and momentum equations . in sect.[model ]
we present the mathematical formalism and the assumptions leading to the self - similar model , an extension of the non relativistic meridionnaly self- similar model @xcite to the case of relativistic jets around rotating black holes . in sect.[results ] we present the collimation criterion established by @xcite extended to this metric and show that the rotation of the black hole induces a more efficient magnetic collimation of the jet .
the common picture for the structure of relativitic jets is the two - component model @xcite , where a relativistic @xmath2 plasma is accelerated in the polar region of the central black hole and is surrounded by a baryonic component coming from the accretion disk . in agn ,
pair injection in the black hole magnetosphere may arise from annihilation of mev photons if the accretion rates are sufficient , or from pair cascades induced by a potential drop in charge - starved regions @xcite . when the jet is emitted in the region very close to the central engine
, frame - dragging effects may play an important role on self - collimation processes . to adress this question we therefore need to model the inner part of the jet close to the polar axis .
the dynamics and the geometry of a magnetized , relativistic plasma flow around a rotating black hole are described by exact solutions to the general relativistic mhd equations in the kerr metric .
a standard treatment to reduce the stationnary and axisymmetric grmhd equations to a system of ordinary differential equations is to adopt a self - similar geometry _
i.e. _ to make the assumption of a scaling law of one of the variable as function of one of the coordinates ( see @xcite for a general classification of self - similar models ) .
analytical solutions using radial self - similarity have been derived from grmhd equations in order to model disk winds @xcite .
however those models fail to describe the regions close to the rotational axis .
a meridional self - similar treatment ( i.e. , similar in the @xmath0-direction ) is necessary to model the spine jet where the outflow is rather driven by the thermal pressure . in those models , the @xmath0-dependence is prescribed _ a priori _ while the radial dependence is derived from the mhd equations .
this modeling does not require the use of a polytropic equation of state .
the local equation of state can be derived a posteriori from the calculated solutions . for our investigation , we chose the analytical model for stellar jets developed by @xcite . in this model ,
outflow solutions are super - alfvnic , and one class of the solutions provides self - confined jets @xcite .
this model has been extended to the general relativistic case in the schwarzschild geometry @xcite and applied to agn jets .
the collimation of the relativistic solutions results from the distribution of the total electromagnetic energy across the jet , as compared to the corresponding distribution of the thermal and gravitational energies .
the fri / frii dichotomy was explained by the magnetic rotator efficiency @xcite . in this paper , we extend the model to the kerr geometry to study the effect of frame - dragging on jet collimation .
the kerr metric is axisymmetric and hence adapted to this assumption . in the following , we make a taylor expansion of all physical quantities with small @xmath0 .
such a treatment allows to study the physical properties of the outflow close to its rotational axis but does not imply any restriction on the black hole rotation going from zero to maximum value .
in the following we use the 3 + 1 decomposition of the kerr spacetime to write the grmhd equations .
the rotating black hole is described by two parameters , @xmath3 and @xmath4 , where @xmath5 is the mass of the black hole and @xmath6 its specific angular momentum , respectively .
the kerr metric writes , @xmath7 where @xmath8 is the lapse function and @xmath9 the angular velocity of the zero angular momentum observer ( hereafter zamo ) as seen from infinity , with @xmath10 , @xmath11 , @xmath12 .
the 4-velocity of a particle can be decomposed as @xmath13 where @xmath14 are the components of the 3-velocity @xmath15 relative to the zamo , and @xmath16 is the lorentz factor . in the following , we use cgs units .
the letters @xmath17,@xmath18,@xmath19 ... represent indices in absolute space and run from 1 to 3 while @xmath20,@xmath21 , @xmath22 ... represent indices in 4-dimensional spacetime and run from 0 to 3 .
+ the stress energy tensor is @xmath23 .
the electromagnetic stress - energy tensor @xmath24 , @xmath25 satisfies to the maxwell equations , @xmath26}=0 \,.\label{tmax}\end{aligned}\ ] ] the electric current density 4-vector @xmath27 is only defined by eq .
[ tmaxs ] .
we assume that the plasma is infinitly conducting .
the electric field is null in the comoving frame , @xmath28 and thus the magnetic field is frozen to the plasma , @xmath29})=0\label{tgel}\,.\end{aligned}\ ] ] in the 3 + 1 form the maxwell equations ( [ tmaxs ] ) and ( [ tmax ] ) writes @xcite , @xmath30 where @xmath31 are , respectively , the electric and magnetic fields measured by a zamo .
the ohm s law ( [ frozen ] ) rewrites , @xmath32 @xmath33 is the energy - momentum tensor for a perfect fluid , @xmath34 where @xmath35 is the proper particle number density , @xmath36 the specific enthalpy per particle , @xmath37 the internal energy density ( including rest - mass energy per particle ) and @xmath38 the isotropic pressure , that is the sum of the kinetic pressure and the pressure associated with the mhd waves . the particles constituting the jet plasma may have two origins .
the particles may either come from the accretion disk and thus be of hadronic origin , or due to pair creation via annihilation processes close to the black hole horizon .
we assume that the particle number is conserved , @xmath39 the jet dynamics is governed by the momentum equation , @xmath40 the first law of thermodynamics is obtained by projecting the conservation of the energy - momentum tensor along the fluid 4-velocity , @xmath41 the basic steady equations governing the kinematics of the outflow ( [ tmasse ] ) and ( [ tforce ] ) are in the 3 + 1 form , @xmath42 assuming infinite conductivity , the contribution of the electromagnetic field is null in eq.([first_principle ] ) .
the first law of thermodynamics becomes @xmath43 because of flux - freezing , in steady axisymmetric outflows , streamlines and magnetic fieldlines are roped on the same flux tubes of constant mass and magnetic flux .
the poloidal components of the velocity and magnetic field can be derived from a stream function @xmath44 and a magnetic flux function @xmath45 , @xmath46 the flow along any magnetic flux tube is given in terms of four constants of motion , the particle flux per unit magnetic flux , @xmath47 the angular velocity of the fieldlines , @xmath48 the total angular momentum , @xmath49 the total energy , @xmath50 from eq . ( [ isok ] ) we deduce the bulk velocity of a fluid particle , @xmath51 by combining eqs .
( [ angular]-[bernoulli ] ) we obtain : @xmath52\ , , \label{emlo}\ ] ] with @xmath53 finally , we deduce the values of the toroidal velocity @xmath54 and the relativistic enthalpy @xmath55 where @xmath5 is the poloidal alfvnic mach number as defined in @xcite , see also @xcite , @xmath56 at the critical point where the denominators of eq .
( [ v_phi ] ) and eq.([enthalpy ] ) vanish , the alfvnic mach number takes the value , @xmath57 it is easily to deduce from the previous equations that at this point , @xmath58 the light cylinder is defined as the surface where @xmath59 . from eqs .
( [ ohm ] ) and ( [ isok ] ) we obtain the poloidal electric field @xmath60 .
the position of the two light surfaces is thus given by @xmath61 . in the following ,
we assume that we can neglect the effect of the electric field compared to the magnetic field , @xmath62 therefore we shall consider @xmath63 in eqs .
( [ v_phi]-[enthalpy ] ) .
the alfvn regularity condition rewrites @xmath64 as a consequences , the field lines can not cross the light cylinder , in contrast with radially self similar disk wind solutions @xcite .
we are limited to describing jets possessing a weak rotation velocity compared to the speed of light ; our solutions are pressure driven .
the typical radius of the spine jet emerging of the black hole is a few dozen schwarzschild radii .
in this section , we present the mathematical formalism of the @xmath65self - similar model . according to the notations used in @xcite all physical quantities are normalized in units of the alfvn quantities at the pole ( subscript @xmath66 ) .
we have assumed as in previous papers that all physical quantities can be developed to first order in @xmath67 , which is equivalent to assuming small colatitudes .
the model hypotheses lead to 4 normalized functions from which two control the flow geometry , * the function @xmath68 gives the variations of the cylindrical radius @xmath69 with the distance to the origin , @xmath70 . * the function @xmath71 is the expansion factor .
it measures the angle between the poloidal magnetic field @xmath72 and the radial direction @xmath73 @xcite .
two functions control the flow dynamics , * the first is the poloidal alfvn mach number @xmath74 ; * the second is @xmath75 which measures the pressure along the polar axis .
+ the model possesses six free parameters , * @xmath76 is the latitudinal variation of density . for @xmath77 ( @xmath78 ) density increases ( decreases ) going out from the axis .
* @xmath79 is the latitudinal variation of the pressure . for @xmath80 ( @xmath81 ) pressure increases ( decreases ) going out from the axis . *
@xmath82 measures the rotation of the flow at the alfvn surface .
it is also a measure of the magnetic lever arm .
* @xmath83 is the ratio of the escape velocity in units of the alfvn speed at the alfvn surface .
* @xmath84 is the normalized schwarszchild radius . *
@xmath85 measures the black hole spin . for the maximal rotation value @xmath86 , @xmath87 .
the normalized spherical radius and the normalized spin parameter are , respectively , @xmath88 the strength of the gravitational potential is given by the polar escape speed at the alfvn point in units of @xmath89 , @xmath90 the normalized schwarzschild radius introduced by @xcite is given by the parameter @xmath84 , @xmath91 which is also the escape speed in units of the speed of light .
combining eqs .
( [ defnu ] ) and ( [ defmu ] ) we get a condition that restrict the parametric space to @xmath92 to recover the @xmath0-self - similar model @xcite , we must expand the metric to first order in @xmath93 .
the redshift factor and the frame - dragging angular velocity write , @xmath94 @xmath95\,.\end{aligned}\ ] ] let us introduce a normalized magnetic flux function @xmath67 .
the magnetic flux potential can be separated into a function of @xmath70 times a function of @xmath0 . to first order in @xmath67 ,
keeping the dipolar term , we have , @xmath96 as pointed out in sect.[model - des ] , in meridionally self - similar flows , the @xmath65dependance is prescribed and the bernoulli and grad - shafranov equations are solved for @xmath97 .
hence @xmath97 measures the magnetic flux relative derivative and its profile determines the geometry of the streamlines .
this function contains the normalized cylindrical radius @xmath68 and is related to the expansion factor @xmath71 as described below .
we get for the poloidal magnetic field from eq .
( [ bp ] ) , @xmath98 which gives keeping only first order terms , @xmath99 the form of the magnetosphere is determined by the explicit dependence of the magnetic flux function @xmath100 on its variables .
using stokes s theorem , eq.([bp ] ) leads to @xmath101 , where @xmath102 is the _ cylindrical _ radius at the alfvn point , in other words the magnetic lever arm .
we introduce the function @xmath68 which represents the normalized cylindrical radius of a given flux tube , @xmath103 we assume that @xmath67 has a dipolar latitudinal dependence @xmath104 that is consistent with the taylor development of the cylindrical radius , @xmath105 combining eqs.([alf_radius]-[cyl_radius ] ) we obtain the magnetic flux function , @xmath106 we introduce the expansion factor @xmath71 that is the logarithmic derivative of @xmath67 with respect to the radius @xmath70 , @xmath107 .
@xmath71 reflects the degree of the flow collimation ( see @xcite for more details ) .
the crucial assumption relates to the shape of the alfvn surface . in the absence of an accretion disk the shape of this surface
emerges as ellipsoidal in the numerical modeling of @xcite . in our model
, we assume that this critical surface is spherical , @xmath108 .
the alfvn regularity condition becomes @xmath109 . using eq.([mach ] )
we find the equipartition relation between electromagnetic and kinetic energy , @xmath110 the assumption of a spherical alfvn surface implies a separation of the variables in the expression of the free function @xmath111 ( see eq.[mach ] ) .
making a first order expansion in @xmath67 , @xmath112 where @xmath76 is a free parameter describing the deviation from spherical symmetry of the ratio @xmath113 .
conversely to the classical model it is not the deviation of the mass density itself as in @xcite ) . following eq.([thermo ] ) the @xmath0-dependance of the gas pressure is similar to that of the density to enthalpy ratio , @xmath114 with @xmath115 a constant , @xmath116 a free parameter describing the deviation from spherical symmetric pressure , and @xmath75 a normalized function .
the function @xmath117 expresses both the total angular momentum loss rate per unit of magnetic flux and the poloidal electric current carried by the outflow @xcite .
@xmath118 using eq.([psia ] ) we expand the angular momentum @xmath119 to first order , @xmath120 the functions @xmath119 and @xmath121 are free while their ratio is fixed by the regularity condition ( [ regular ] ) .
we deduce from eq.([regular]-[l ] ) the bernoulli integral and the isorotation law , @xmath122 as in @xcite the @xmath82 parameter measures the strength of the magnetic torque . in the following we introduce a new parameter @xmath123 to introduce the normalized frame - dragging potential @xmath124 in our equations . from the previous assumptions we get the expressions of the magnetic and velocity fields as well as the density and the enthalpy functions : @xmath125 \label{hgw } \,,\\
h_{}\gamma n&=&h_{\star } \gamma_\star n_\star \frac{h_{\star}^2}{m^2}\left[1+\delta\alpha -\frac{\mu \lambda^2}{\nu^2 } \left(\aleph^2\frac{n_b}{d}+\frac{\bar{\omega}}{\lambda}\right ) \alpha \label{hgn } \right ] \,.\end{aligned}\ ] ] where @xmath126
we solve equations ( [ ode1]-[ode2 ] ) .
the numerical procedure is given in the appendix .
the jet is launched from a region very close to the black hole , typically 2 gravitational radii .
we present solutions corresponding to the following set of parameters : @xmath127 , @xmath128 , @xmath129 , @xmath130 , @xmath131 , and for a black hole spin @xmath132 .
fig.[f2 ] shows the topology of the velocity and magnetic field .
fig.[f3 ] shows the effect of the black hole rotation on the jet base .
fig.[f4 ] displays the corresponding lorentz factor .
the asymptotic lorentz factor is larger in the case of a limiting solution , i.e. when the pressure @xmath133 is the lowest one that gives a cylindrical solution .
fig.[cylradius ] ( left panel ) displays three solutions to the grmhd equations , corresponding to the same physical parameters , except for the spin of the central engine .
( @xmath134 , @xmath128 , @xmath135 , @xmath136 , @xmath137 ) .
the upper jet is launched by a schwarzschild black hole while the other two solutions by a kerr black hole ( @xmath138 , @xmath139 , respectively ) .
the jet solution in the schwarzschild case is asymptotically cylindrical because we have chosen the minimum value of @xmath140 giving the limiting solution ( see @xcite for details ) . for a lower initial pressure the jet would decollimate while with a higher initial pressure the jet would undergo recollimation with oscillations .
an oscillatory jet width is a basic feature of such recollimating jets @xcite .
oscillations occurs because of the interplay of the centrifugal force and the total ( pinching plus pressure ) magnetic force , always acting in opposite directions along the cylindrical radius @xmath69 .
the transverse pressure gradient remaining similar this proves that the rotation of the central black hole enhances the efficiency of the magnetic field to collimate .
we can show in a different manner the same result now using the kerr solution of fig .
[ f2 ] and lowering the initial value of the pressure by lowering @xmath133 until we get the limiting solution for a value @xmath141 .
[ cylradius ] ( right panel ) displays two jet morphologies which are almost identical but clearly the same rate of collimation is obtained in the second case for a lower pressure as @xmath142 . in @xcite ,
a general criterion for the jet collimation has been established , based on the variation of the energy across poloidal fieldlines .
it gives an important extra parameter @xmath143 which provides the efficiency of the magnetic rotator to collimate the flow .
we follow the same procedure as in @xcite to derive the magnetic collimation parameter @xmath143 in the kerr metric . after substituting @xmath35 from eq .
( [ mach ] ) and derivating with @xmath67 constant , @xmath144 , eq .
( [ thermo ] ) can be re - written as @xmath145 \right|_\alpha \label{integral1 } \,,\end{aligned}\ ] ] where @xmath146 is proportional to the energy _ per unit volume _ of the fluid reduced to the thermal content .
thus @xmath147 in essence measures the variation between the total energy and the thermal energy of the fluid .
the form of the pressure is @xmath148 .
we also know the @xmath0 dependence in all quantities in the expression for @xmath149 , and after expanding with respect to @xmath93 we find @xmath150 .
( [ integral1 ] ) gives @xmath151 eliminating @xmath152 we get the integral @xmath153 which is a local measurement of the magnetic efficiency to collimate the flow but does not remain constant along the flow .
after substituting the expressions for @xmath154 and @xmath155 , we arrive at @xmath156 we refer to @xcite for a detailed parametric analysis and a discussion on the different solutions which are inferred from the modeling . here
we only consider for a first approach @xmath157 .
this does not necessarily implies that the flow is pressure confined
. it may be either magnetically confined or pressure confined depending on whether the efficiency of the magnetic rotator prevails or not to the thermal confinement .
a negative @xmath143 implies that the source is an inefficient magnetic rotator , which needs the help of the gas pressure to collimate the outflow @xcite .
conversely , a positive @xmath143 is the sign of an efficient magnetic rotator with a strong magnetic contribution to collimation .
it is simple to express this constant at the base of the flow @xmath158 assuming the poloidal velocity is negligible there , such that the alfvnic mach number @xmath159 .
then @xmath143 takes the following form at the source boundary , @xmath160 where @xmath161 is the poynting flux and @xmath162 is the rotational energy per particle .
the following term is very similar to the nonrelativistic one except for the metric , @xmath163 it measures the excess or the deficit of gravitational energy per unit mass which is not compensated by the thermal driving @xcite .
the last term @xmath164 represents the coupling between the orbital angular momentum of the fluid particle and the frame dragging of the kerr black hole .
we see in eq .
( [ epsilon_source ] ) that the magnetic collimation parameter seems to be larger in the kerr metric , because of the new term @xmath165 and also because @xmath166 is smaller as we increase @xmath85 .
in order to check this assumption , we calculated solutions with the same physical parameters , except for the black hole spin .
fig.[f6 ] shows the evolution of @xmath143 with the rotation of the black hole for two different sets of parameters corresponding to the solutions presented above .
the magnetic collimation efficiency increases linearly with the black hole spin .
we may expect that it comes with the linear increase of the roping of the magnetic field at the base of the flow due to the drift of the rotation coordinate .
we extended the meridional self - similar model of @xcite to the grmhd case in the kerr spacetime .
we presented the first axisymmetric analytical mhd solutions ( 2.5d ) for an outflow accelerated near the polar axis of a rotating black hole . in order to keep the self - similarity in colatitude
, we had to assume an expansion of all equations including the metric with the colatitude @xmath0 .
as in the self - similar model in the schwarzschild geometry @xcite , we made the assumption of a non relativistic rotational velocity to neglect the effects of the light cylinder on the alfvn surface .
we derived exact solutions to the grmhd equations where the collimation derive from a combination of pressure and magnetic forces .
the contribution of the different mechanisms to the collimation of the outflow in the context of meridional self - similar models was studied in previous papers @xcite . in this paper
, we made an extension of the magnetic collimation parameter for a rotating black hole in order to investigate the effect of the black hole spin on the flow collimation .
the collimation efficiency , given by the variation of the specific energy across streamlines , has five contributions ; each one gives the variation - in units of the volumetric energy of the magnetic rotator - between any streamline and the polar axis of : the kinetic energy , the volumetric gravitational energy , the poynting flux , the thermal content and the rotational energy associated with the frame - dragging potential .
the kinetic and magnetic contents depends strongly on the physics of the central engine . in the picture
the jet is driven by the blandford - znajek mechanism , the injected power goes as @xmath167 , _
i.e. _ there is an energy deficit near the polar axis , which would have the effect of increasing the magnetic collimation parameter @xmath143 .
the thermal content may depend on plasma injection along the magnetic field lines , and a complete picture should include the disk wind component , which could inhibit the lateral expansion since it is believed to be more dense than the leptonic spine jet .
the rotation of the black hole could also play a role since the location of the innermost stable orbit depends on the spin parameter .
the efficiency of self - collimation in relativistic jets can be compromised by the existence of the eletric fields , which are negligible far from the light cylinder .
we considered only the region close to the polar axis where the effect of the decollimating electric fields are negligible while the frame - dragging play an important role .
we found that the magnetic collimation efficiency is enhanced by the rotation of the black hole , and increases linearly with the black hole spin in our model .
this gain is due ( i ) to the fact that the volumetric gravitational energy @xmath166 , which has a negative contribution to the collimation efficiency , is smaller in the kerr metric , and ( ii ) to the presence of the new term @xmath168 , associated with the frame - dragging . as our model is valid only for small colatitude angles ,
the contribution of this term is then under - estimated .
for convinience we use the following notations : @xmath169 , @xmath170 . under the assumptions of axisymmetry and meridional self - similarity , the grmhd equations reduce to the following four ordinary differential equations for @xmath75 , @xmath74 , @xmath68 and @xmath71 : @xmath171 where we have defined : @xmath172 \nonumber\\&&+\frac{{h_{}^2 m^2}}{h_{\star}^2 } \left[\frac{f-2}{r}\left(1+\kappa\frac{x_{_+}}{g^2}-\frac{\sigma^2}{2 x_{_+}}\right ) -\frac{f}{\sqrt{x_{_+ } } } + \frac{h_{}^2f^2}{4 } \left ( \frac{r}{x_{_+}}+\frac{\mu}{h_{}^2}\frac{x_{_-}}{x_{_+}^2 } \right ) + \frac{h_{}^2f^3}{8\sqrt{x_{_+ } } } -\frac{h_\star^2}{h^2}\frac{\lambda^2\aleph^2\mu}{\nu^2}f\sqrt{x_{_+}}\right ] \nonumber\\&&-\frac{d g^2}{2{h _ { } } ^2 m^2}\frac{x_{_-}}{x_{_+}}\left(\nu^2 h_{\star}^4-\mu\frac{m^4}{g^4}\right)\left[\delta-\kappa+\frac{\sigma^2 g^2}{x_{_+}^2}\left(\frac{1}{2}-h^2\frac{\sigma^2 - 3r^2}{x_{_-}}\right)-\frac{2\mu \lambda^2 \aleph^2}{\nu^2}\left(\frac{n_b}{d}+\frac{\bar{\omega}}{\lambda \aleph}\right)\right ] \nonumber\\&&-\frac{m^2\mu}{2h_{}^2}\frac{x_{_-}}{x_{_+}^2}d + \lambda^2 \aleph^2 \frac{x_{_-}}{x_{_+}}\frac{n_b n_v}{d^2}\mu + \frac{h^2 g^2 d}{2}\kappa\pi\left[\frac{h^2}{h_{\star}^2}\frac{f}{d}\sqrt{x_{_+}}-2r-\frac{x_{_+}}{r}(f-2)\right ] \nonumber\\&&+ \frac{\lambda \aleph\sqrt{\mu}}{\nu } \frac{\sigma g^2}{h_{\star}m^2}\left(\nu^2 h_{\star}^4-\mu\frac{m^4}{g^4}\right ) \frac{\sigma^2 - 3 r^2}{x_{_+}^2}n_v + \frac{h_{}^2}{h_{\star}^2}\frac{f m^2}{2 } \frac{r\sigma^2}{x_{_+}^{\frac{5}{2 } } } \left(\mu - h_{\star}^4\nu^2\frac{g^4}{m^4}\right ) \nonumber\\&&-\lambda^2
\aleph^2 h^2 \frac{n_b}{d}(f-2 ) \frac{x_{_+}}{r } + \lambda^2\aleph^2 r\left[2m^2+h^2\left(\frac{f\sqrt{x_{_+}}}{r}-2\right)\right]\left(\frac{n_b^2}{d^2}-\frac{h^2}{2 m^2}\frac{n_v^2}{d^2}\right ) \,,\end{aligned}\ ] ] @xmath173 \nonumber\\ & & + \frac{h_{}^2}{h_{\star}^2 } \left(\upsilon-\frac{h_{}^2f^2}{4}\right ) \left[\frac{f^2}{r}+f\left(\frac{r}{x_{_+}}-\frac{2}{r}\right)-\frac{2}{h_{}^2\sqrt{x_{_+ } } } { - 4\frac{h_\star^2}{{h_{}}^4}\frac{\lambda^2\aleph^2\mu}{\nu^2}\sqrt{x_{_+}}}\right ] \nonumber\\&&+\left(\upsilon-\frac{h_{}^2f^2}{4}\right)\frac{\mu f}{h_{\star}^2}\frac{x_{_-}}{x_{_+}^2 } + \frac{2\pi g^2\kappa\sqrt{x_{_+}}}{h_{\star}^2 } \left [ \upsilon-\frac{h_{}^2f^2}{4 } \left(\frac{2}{f}\left[\frac{r}{\sqrt{x_{_+}}}-\frac{\sqrt{x_{_+}}}{r}\right]+\frac{\sqrt{x_{_+}}}{r}\right ) \right ] \nonumber\\&&-\frac{g^2 f}{2 m^2 h_{}^2h_{\star}^2}\frac{x_{_-}}{x_{_+}}\left(\nu^2 h_{\star}^4-\mu\frac{m^4}{g^4}\right)\left[\delta-\kappa+\frac{\sigma^2 g^2}{x_{_+}^2}\left(\frac{1}{2}-h^2\frac{\sigma^2 - 3r^2}{x_{_-}}\right)-\frac{2\mu \lambda^2 \aleph^2}{\nu^2}\left(\frac{n_b}{d}+\frac{\bar{\omega}}{\lambda \aleph}\right)\right ] \nonumber\\&&+\frac{4\lambda^2\aleph^2}{h_{}^2 } \left(\frac{n_b^2}{d^2}-\frac{h_{}^2}{2 m^2 } \frac{n_v^2}{d^2}\right)\left(\upsilon\sqrt{x_{_+}}-\frac{h^2 r f}{2}\right)-\lambda^2\aleph^2 \frac{h^2}{h_\star^2}\frac{x_{_+}}{r}f(f-2)\frac{n_b}{d^2 } -\frac{\mu f m^2}{2 h_{}^2 h_{\star}^2}\frac{x_{_-}}{x_{_+}^2 } \nonumber\\&&+\frac{2 m^2 \sigma^2 } { h^2 h_{\star}^2 x_{_+}^{\frac{3}{2}}}\left[\left(1-\frac{\nu^2 h_{\star}^4g^4r}{m^4 x_{_+}}\right ) \upsilon-\frac{h_{}^2f^2}{4}\right ] + \lambda^2\aleph^2\frac{\mu}{h_{\star}^2}\frac{x_{_-}}{x_{_+}}\frac{n_b n_v}{d^3}f + \mu\frac{f^2 m^2}{2 h_{\star}^2}\frac{r\sigma^2}{x_{_+}^{\frac{5}{2 } } } \nonumber\\&&+ \frac{\lambda \aleph\sqrt{\mu}}{\nu } \frac{g^2 f\sigma}{m^2h_{\star}^3}\left(\nu^2 h_{\star}^4-\mu\frac{m^4}{g^4}\right ) \frac{\sigma^2 - 3 r^2}{x_{_+}^2}\frac{n_v}{d } \,,\end{aligned}\ ] ] @xmath174 at the alfvn radius , the expansion factor is the solution of the second degree polynomial @xmath175 , with @xmath176 - \lambda^2 \aleph^2_\star x_{+,\star}^{3/2 } p ' \left [ 2 h^2_\star - \mu \lambda_\star \right ] \nonumber \\ & & - \lambda^2 \aleph^2_\star x_{+,\star}^{3/2 } \left [ 2 h^4_\star - 2 \mu
\lambda_\star h^2_\star + \frac{\mu^2 \lambda^2_\star}{2}\right]\,.\end{aligned}\ ] ] where @xmath177 , @xmath178 , @xmath179 .
assuming self - similarity , the pde system of grmhd equations is transformed into an ode system where the functions depends only on the normalized radial distance @xmath70 .
four coupled equations constitute the system : the three equations given in appendix [ appendixa ] determine the unknown functions @xmath75 , @xmath71 and @xmath180 and the function @xmath68 is related to @xmath71 through eq .
( [ f - g ] ) .
we have shown that @xmath71 , @xmath68 are dimensionless functions related to the shape of the jet while @xmath181 , @xmath75 describe the physics of the magnetosphere .
we briefly present the method for the numerical integration of eqs .
( [ ode1 ] ) - ( [ ode2 ] ) .
we start integrating the equations from the alfvn critical surface . in order to calculate the toroidal components of the fields , i.e. @xmath182 and @xmath183
, we apply lhospital s rule at this point , @xmath184 } { p-\mu\frac{1-\sigma^2}{(1+\sigma^2)^2 } } \ , , \quad p=\left.\frac{{\rm d } m^2}{{\rm d } r } \right|_{\star}\,.\ ] ] to avoid kinks in the fieldline shape , we need to satisfy a regularity condition @xcite .
this means that eq .
( [ dfdr ] ) should be regular at @xmath185 .
as in @xcite this extra requirement is equivalent to @xmath186 which eventually gives a second order polynomial equation for the alfvnic expansion factor @xmath187 , @xmath188 with the expressions for the coefficients @xmath189 , @xmath190 and @xmath191 given in the appendix [ appendixa ] . once we have determined the regularity conditions at the alfvn point , we integrate downwind and upwind and cross all the other existing critical points as in the non relativistic case .
notice that solutions depend also on @xmath192 , i.e. the pressure at the alfvn surface .
as in the classical case its value has been chosen such that the total gas pressure is always positive .
more details on the numerical technique can be found in @xcite .
g. v. bicknell , astrophys .
j. , suppl .
* 101 * , 29 ( 1995 ) d. s. de young , astrophys .
j. , lett .
, * 405 * , l13 ( 1993 ) r. a. laing , p. parma , & h. r. de ruiter , mon . not .
r. astron .
soc . , * 306 * , 513 ( 1999 ) gopal - krishna & p. j. wiita , astron .
astrophys . , * 363 * , 507 ( 2000 ) z. meliani , b. giacomazzo & r. keppens , astron .
astrophys . ,
* 491 * , 391 ( 2008 ) s. a. baum , e. l. zirbel & c. p. odea , astrophys .
j. , * 451 * , 88 ( 1995 ) c. s. reynolds , t. di matteo , a. c. fabian et al . , mon . not .
r. astron .
soc . , * 283 * , 111 ( 1996 ) d. l. meier , astrophys . j. , * 522 * , 753 ( 1999 ) z. meliani & r. keppens , astrophys . j. , * 705 * , 1594 ( 2009 ) j. c. mckinney & r. d. blandford , mon . not .
r. astron .
soc . , * 394l * , 126 ( 2009 ) m. cegowski , m. p. gawroski , m. kunert - bajraszewska , astron . astrophys . , * 557 * , a75 ( 2013 ) c. brocksopp , c. r. kaiser , a. p. schoenmakers , a. g. de bruyn , mon . not .
r. astron .
, * 410 * , 484 ( 2011 ) c. konar , m. j. hardcastle , mon . not .
r. astron .
soc . , * 436 * , 1595 ( 2013 ) z. meliani , c. sauty , k. tsinganos , e. trussoni & v. cayatte , astron . astrophys . , * 521 * , 67 ( 2010 ) a. tchekhovskoy , r. narayan & j. c. mckinney , astrophys . j. , * 711 * , 50 ( 2010 ) r. narayan , & j. e. mcclintock , mon . not .
r. astron .
, * 419 * , 69 ( 2012 ) m. c. begelman , r. d. blandford & m. j. rees , nature , * 287 * , 307 ( 1980 ) r. d. blandford and r. l. znajek , mon . not .
r. astron .
soc . , * 179 * , 433 , ( 1977 ) e. s. phinney , phd thesis , cambridge university ( 1983 ) m. camenzind , astron . astrophys . , * 162 * , 32 ( 1986 ) m. takahashi , s. nitta , y. tatematsu & a. tomimatsu , astrophys . j. , * 363 * , 206 ( 1990 ) i. okamoto , mon . not
r. astron .
, * 254 * , 192 ( 1992 ) c. fendt , 1997 , astron .
astrophys . , * 319 * , 1025 v. s. beskin , i. v. kuznetsova , nuova cimento b , * 115 * , 795 ( 2000 ) n. globus and a. levinson , phys . rev
. d , * 88 * , 4046 g ( 2013 ) d. a. uzdensky , astrophys . j. , * 603 * , 652 ( 2004 ) d. a. uzdensky , astrophys . j. , * 620 * , 889 ( 2005 ) s. komissarov , mon . not .
r. astron .
* 326 * , l41(2001 ) s. komissarov , mon . not .
r. astron .
soc . , * 350 * , 1431 ( 2004 ) s. komissarov , mon . not .
r. astron .
, * 350 * , 427 ( 2004 ) s. komissarov , mon . not
r. astron .
, * 359 * , 801 ( 2005 ) a. levinson & m. c. begelman , astrophys .
j. , * 764 * , 148l ( 2013 ) k. l. chan & r. n. henriksen , astrophys .
j. , * 241 * , 534 ( 1980 ) c. sauty & k. tsinganos , astron .
, * 287 * , 893 ( 1994 ) c. sauty , k. tsinganos & e. trussoni , astron .
astrophys . , * 348 * , 327 ( 1999 ) k. s. thorne , d. a. macdonald and r. h. price , _ black holes : the membrane paradigm _ ( yale university press , 1986 ) c. m. mobarry & r. v. e. lovelace , astrophys
. j. , * 309 * , 455 ( 1986 ) h. sol , g. pelletier & e. asseo , mon . not .
r. astron .
* 237 * , 411 ( 1989 ) a. levinson & f. rieger , astrophys .
j. , * 730 * , 123 ( 2011 ) n. vlahakis & k. tsinganos , mon . not .
r. astron .
, * 298 * , 777 ( 1998 ) z. meliani , c. sauty , n. vlahakis , k. tsinganos & e. trussoni , astron . astrophys . ,
* 447 * , 797 ( 2006 ) e. breitmoser & m. camenzind , astron .
astrophys . ,
* 361 * , 207 ( 2000 ) f. c. michel , astrophys .
j. , * 158 * , 727 ( 1969 ) n. vlahakis & a. knigl , astrophys .
j. , * 596 * , 1104 ( 2003 ) n. vlahakis & a. knigl , astrophys .
j. , * 596 * , 1080 ( 2003 ) n. vlahakis & a. knigl , astrophys . j. , * 605 * , 656 ( 2004 ) t. sakurai , astron . astrophys .
, * 152 * , 121 ( 1985 ) j. heyvaerts & c. norman , astrophys .
j. , * 347 * , 1055h ( 1989 ) | we present a model for the spine of relativistic mhd outflows in the kerr geometry .
meridional self - similarity is invoked to derive semi - analytical solutions close to the polar axis .
the study of the energy conservation along a particular field line gives a simple criterion for the collimation of jets .
such parameter have already been derived in the classical case by sauty et al .
1999 and also extended to the schwarzschild metric by meliani et al . 2006 .
we generalize the same study to the kerr metric .
we show that the rotation of the black hole increases the magnetic self - confinement . + * pacs numbers : * 04.70.-s , 47.75.+f , 95.30.qd |
we will discover a large class of positive finite presentations of the @xmath0-braid group . roughly speaking , a presentation in this class
has a set of generators determined by a connected graph with @xmath0 vertices immersed in a plane in such a way that each pair of edges intersects at most once .
our class of presentations includes all known positive finite presentations such as the artin presentation@xcite , the band - generator presentation in @xcite and all of sergiescu s presentations in @xcite .
our first objective is to find a ( minimal ) collection of positive relations among generators given by a graph as above so that it becomes a presentation of the @xmath0-braid groups .
the semigroup of all positive words plays a crucial role in the solutions to the word problem and the conjugacy problem for the @xmath0-braid group given by garside@xcite , thurston@xcite , and elrifi - morton@xcite .
their solutions are based on the property that the semigroup embeds in the whole group .
birman - ko - lee s recent work@xcite also requires this embedding property in order to give a fast solution to the word problem in the band - generator presentation .
our second objective is to prove that with a few exception all presentations in our class have the embedding property .
the artin and the band - generator presentations are the only nontrivial exceptions .
consequently these two among all presentations are natural and interesting for further study .
the @xmath0-braid group @xmath1 is the group of isotopy classes of orientation - preserving automorphisms of an @xmath0 punctured plane @xmath2 that fix the set of punctures and the outside of a disk containing all punctures .
it is customary to place punctures on the @xmath3-axis and equally spaced . ignoring the punctures ,
such an automorphism is a homeomorphism of @xmath2 permuting the punctures and so is isotopic to the identity map of @xmath2 .
a geometric braid is determined by taking the trace of the punctures under this isotopy .
when the punctures are not in the customary location , there is a choice of homeomorphisms of @xmath4 that send the punctures to their customary location and a geometric braid is uniquely determined up to conjugation in this case .
since the symmetric group over @xmath0 letters is naturally a quotient of @xmath1 , an automorphism that exchanges two punctures is a candidate for a generator of @xmath1 and a typical automorphism of this type can be easily depicted as an arc connecting the two punctures so the automorphism exchanges two punctures clockwise or counterclockwise inside an annulus lying along the arc .
see figure [ f : string ] for an example .
this type of braids will be called the _ half twist along the given arc_. therefore by regarding the punctures as vertices and a collection of these arcs as edges , we have a graph @xmath5 immersed in @xmath2 satisfying the following properties : 1 .
there are no loops ; 2 .
edges have no self - intersections ; 3 . two distinct edges intersect either at interior points transversely or at common end points
there is at least a vertex inside a pseudo digon " , that is , a region cobounded by two subarcs from two edges so that corners of the region are either two interior intersection points or one vertex and one interior intersection point . if the number of intersections among edges is minimized , for example , by making all edges geodesics in the hyperbolic structure of the punctured plane , the condition ( 4 ) holds automatically .
two immersed graphs are regarded as equivalent if one graph can be transformed to the other by an orientation - preserving homeomorphism of @xmath4 .
thus we may assume that vertices lie on the customary locations if necessary . throughout this article ,
graphs mean immersed graphs in @xmath6 satisfying the condition discussed .
if graphs are embedded in @xmath4 , we call them _
planar_. the set of edges in a graph @xmath5 with @xmath0 vertices corresponds to a set of elements in @xmath1 as in figure [ f : string ] .
two equivalent graphs determine two sets of braids that differ by an inner automorphism of @xmath1 .
if the set corresponding to edges generates @xmath1 , we say that @xmath5 _ generates _ @xmath1 and we are interested in graphs that generate @xmath1 . since an edge of @xmath5 corresponds to a transposition in the symmetric group , a graph that generates @xmath1 must be connected in order to permute any two vertices .
but there are connected graphs that do not generate @xmath1 .
the problem of deciding when an immersed graph generates the braid group is considered elsewhere@xcite .
we now discuss how a braid word corresponds to an edge of a graph with @xmath0 vertices in the customary location .
the artin generator @xmath7(or @xmath8 ) for @xmath9 corresponds to the clockwise(counterclockwise , respectively ) half twist along the straight edge joining the @xmath10-th and the @xmath11-st vertices .
given an edge of a graph , we simplify the edge via a sequence of half twists @xmath7 or @xmath8 for @xmath9 until it becomes a straight edge corresponding to , say , @xmath12 .
the required sequence can be expressed as a word @xmath13 in artin generators in which a half twist applied later is written on the left .
then the edge that we started with corresponds to the braid word @xmath14 .
the word @xmath13 is not uniquely expressed but @xmath13 as an element of @xmath1 is well - defined .
the inverse of this correspondence is similar . given a conjugate @xmath14 , apply a sequence of half twists determined from right to left by @xmath15 to the straight edge @xmath12 .
consequently each edge of a graph with @xmath0 vertices is uniquely represented by an element in @xmath1 that can be written as a conjugate of an artin generator .
the following lemma summarizes this discussion .
a braid @xmath16 in @xmath1 is the half twist along an arc joining two punctures in the plane if and only if @xmath16 can be written as a conjugate of an artin generator or its inverse . from now on as long as no confusion arises
, we will not distinguish three concepts , namely an edge of a graph , the half twist along the edge , and a conjugate word expressing the edge .
since any two artin generators are conjugate each other , any half twist along an arc corresponds to a conjugate of a fixed artin generator .
the following is immediate from the above lemma .
[ conjugate ] a graph @xmath5 of @xmath0 vertices generates @xmath1 if and only if each artin generator can be expressed as @xmath17 for some edge @xmath18 and some word @xmath13 on edges of @xmath5 . in the view of the above corollary , it is important to know how the conjugate of an edge by another edge in a graph looks like .
let @xmath18 and @xmath16 be words in @xmath1 that express edges of a graph .
if the edges @xmath18 and @xmath16 do not intersect each other and are not adjacent , they commute and so @xmath19 . if the edges @xmath18 and @xmath16 do not intersect each other and are adjacent , the three edges @xmath18 , @xmath16 , and @xmath20 form a triangle counterclockwise and the three edges @xmath18 , @xmath16 , and @xmath21 form a triangle clockwise as in figure [ f : triangle ] .
we note that there should be no other vertex inside the triangles and this can be achieved by drawing thin triangles inside a sufficiently small neighborhood of the union of two edges @xmath18 and @xmath16 .
when the edges @xmath18 and @xmath16 intersect at an interior point , one can still describe the edges @xmath22 and @xmath23 but we do not need them in this article . give a set @xmath24 of generators of the @xmath0-braid group @xmath1 , a _ positive word _
in @xmath24 is a product of positive powers of generators in @xmath24 .
a presentation of a group is _ finite _ if there are finitely many generators and relations .
a presentation is _ positive _ if all defining relations are equations of positive words in generators .
we now introduce the graphs that corresponds to some of known positive finite presentations of the braid groups .
the set of artin generators @xmath25 of @xmath1 forms the graph in figure [ f : artin ] and a minimal set of defining relations is given by @xmath26 the graphs in figure [ f : artin ] will be called the _ artin _ graph .
sergiescu @xcite showed how a finite presentation of @xmath1 can be obtained from any planar connected graph with @xmath0 vertices .
he described a sufficient set of positive relations that depend only on the geometry of a given planar graph .
we will give a minimal set of positive relations as a corollary of the main theorem in 3 .
let @xmath0 be the number of vertices in @xmath5 . by choosing a fixed @xmath0 points in @xmath27 and a homeomorphism of @xmath27 that sends vertices of @xmath5 to the @xmath0 fixed points , the group @xmath28 introduced in @xcite
is identified with the @xmath0-braid group @xmath1 .
these presentations of @xmath1 will be called _
sergiescu s presentations . _
the relations in a sergiescu s presentation are highly redundant but they are useful when we need to find a relation locally given by a planar graph .
recently a new presentation of @xmath1 called the band - generator presentation has been developed by birman - ko - lee@xcite .
this presentation has @xmath29 generators @xmath30 for @xmath31 corresponds to the graph in figure [ f : band ] and defining relations : @xmath32 the artin generator and the band - generators are related as @xmath33 and @xmath34 the set of band - generators are depicted as the graph in figure [ f : band ] that will be called the _ inner complete _ graph .
the relations @xmath35 will be called a _
triangular _ relation . a triangular relation
is derived whenever a new generator is introduced by means of a conjugation of an edge by an adjacent edge .
triangular relations serve as building blocks of positive relations in the braid groups .
after an immersed graph is turned into a planar graph by regarding all interior intersections as vertices , a region bounded by a closed edge - path is called a _ pseudo face _ if the edge - path contains at least one vertex that is an interior intersection of two edges as in figure [ f : pseudo ] .
a graph @xmath5 is said to be _ linearly spanned _ if it is connected and there is no vertex in any pseudo face of @xmath5 .
a connected subgraph of a linearly spanned graph is clearly linearly spanned .
two edges of a linearly spanned graph intersect each other at most once since any pseudo digon cobounded by two edges can not exist in a linearly spanned graph .
all planar graphs and all subgraphs of the inner - complete graph are linearly spanned .
for example the graph on the left in figure [ f : linear ] is neither planar nor a subgraph of the inner - complete graph but it is linearly spanned , and the graph on the right is not linearly spanned the following lemma justifies the terminology `` linearly spanned '' . a linearly spanned graph that is a tree is equivalent to a subgraph of the inner - complete graph .
choose a point @xmath3 far away from the graph . since the graph is a tree and no vertices
are surrounded by edges , there are @xmath0 arcs that join @xmath3 to each vertex and are disjoint each other and are disjoint from the graph as in figure [ f : tree ] .
choose a new horizontal axis disjoint from the graph and move each vertex along each arc by a homeomorphism so that it lies on the new axis as figure [ f : tree ] .
then the result is a subgraph of the inner complete graph .
a linearly spanned graph @xmath5 with @xmath0 vertices generates @xmath1 .
a maximal tree of @xmath5 is equivalent to a connected subgraph of the inner - complete graph from which we can obtain the inner complete graph by adding missing edges that are conjugates of an existing edge by an adjacent edge . [ tree ]
let @xmath36 be a linearly spanned tree with @xmath0 vertices .
then @xmath36 generates @xmath1 with @xmath37 positive relations described as in the proof below .
induction on @xmath0 .
it is trivial when @xmath38 .
suppose a linear spanned tree @xmath39 with @xmath40 vertices generates @xmath41 with @xmath42 positive relations .
we add a new vertex and a new edge @xmath18 to @xmath39 . for each edge @xmath16 of @xmath39
, we will have a new positive relation so that @xmath43 new positive relations will be added . in the following we use tietze transformations @xcite that add and delete a generator(s ) denoted by @xmath44 or @xmath45 to utilize triangular relations . 1 .
if @xmath18 and @xmath16 have no intersection , add the positive relation @xmath46 2 . for each set of edges @xmath47 simultaneously adjacent to @xmath18 at a vertex of valency @xmath48 as in figure
[ f : induct1](a ) , add @xmath49 positive relations @xmath50 3 .
if @xmath18 and @xmath16 intersect and form a pseudo face as in figure [ f : induct1](b ) , add the positive relation @xmath51 that is derived from @xmath52 where @xmath53 4 .
if @xmath18 and @xmath16 intersect and form a pseudo face as in figure [ f : induct2](a ) , add the positive relation @xmath54 that is derived from @xmath55 where @xmath53 5 .
if @xmath18 and @xmath16 intersect and form a pseudo face as in figure [ f : induct2](b ) , add the positive relation @xmath56 that is derived from @xmath57 where @xmath58 the following is the main theorem of this section and the known presentation mentioned in the previous section can be obtained from this .
[ circuit ] let @xmath5 be a linearly spanned graph with @xmath0 vertices
. then @xmath5 generates @xmath1 with @xmath59 positive relations described as in the proof below where @xmath5 has @xmath60 edges .
choose a spanning tree @xmath36 of @xmath5 .
then a linearly spanned tree @xmath36 generates @xmath1 with @xmath37 positive relations as in lemma [ tree ] .
when each edge @xmath18 in @xmath61 are added , a circuit is formed and the circuit give a new positive relation as follows : 1 . if @xmath18 forms a circuit with no intersection with other edges as in figure [ f : cutopen ] , add the positive relation @xmath62 where we regard that the circuit bounds a polygonal disk by cutting open all edges inside the circuit as in figure [ f : cutopen ] .
2 . if @xmath18 forms a circuit with some intersections with other edges as in figure [ f : face ] , add the positive relation @xmath63
two positive words @xmath64 , @xmath65 in a positive presentation will be said to be _ positively equivalent _ if they are identically equal or they can be transformed into each other through a sequence of positive words such that each word of the sequence is obtained from the preceding one by a single direct application of the defining relations . and we will write @xmath66 if @xmath64 and @xmath65 are positively equivalent . given a positive finite presentation @xmath67 of @xmath1 , let @xmath68 be the free semigroup generated by @xmath69 modulo @xmath70 . if any two equivalent positive words @xmath64 and @xmath65 are positively equivalent , then we say that the semigroup @xmath68 _ embeds _ in @xmath1 or the presentation @xmath67 has the _ embedding property_. a set @xmath69 of generators is said to have the _ embedding property _ if a presentation @xmath67 has the embedding property for some finite set @xmath70 of positive relations .
the _ embedding problem _ of a graph that generates @xmath1 with positive relations is to decide whether the set of generators given by the graph has the embedding property .
thus a graph does not have the embedding property if and only if no finite set of positive relations over the set of generators given by the graph form a presentation with the embedding property .
garside@xcite showed that the artin presentation has the embedding property and birman - ko - lee@xcite showed that the band - generator presentation has the embedding property .
we will show that linearly spanned graphs with more than 3 vertices do not have the embedding property except these two presentations .
there are 4 possible linearly spanned graphs with 3 vertices as in figure [ f:3vertex ] .
the first two graphs give the artin and the band - generator presentations of @xmath71 .
the last two graphs have multiple edges .
we do not know whether these graphs have the embedding property .
in particular the following presentation with 4 defining relations : @xmath72 may have the embedding property .
but all known techniques as in @xcite fail to apply to this example .
we will try to avoid these unknown exceptions in the following discussion by allowing no multiple edges .
[ circle ] let @xmath73 be a connected full subgraph of a linearly spanned graph @xmath5 . if a graph @xmath5 has the embedding property and there is a circle @xmath74 such that @xmath74 contains @xmath73 inside and all vertices in @xmath75 lie outside , then @xmath73 also has the embedding property .
let @xmath69 and @xmath76 be the set of generators given by @xmath5 and @xmath73 , respectively and let @xmath70 be a finite set of positive relations on @xmath69 such that @xmath67 has the embedding property .
let @xmath77 be the set of relations on @xmath76 which is a full " subset of @xmath70 in the sense that @xmath77 contains all relations in @xmath70 written on @xmath76 .
we will prove by contradiction that @xmath78 has the embedding property .
suppose that there exists a pair of positive words @xmath79 on @xmath76 such that @xmath80 are equivalent in the braid group but are not positively equivalent under @xmath77 . since @xmath81 has the embedding property , @xmath66 under @xmath70 .
thus there is a sequence of positive words @xmath82 over @xmath69 such that @xmath83 under @xmath70 and each positive equivalence is obtained by one direct application of relations in @xmath70 .
the sequence must contain at least one positive word , say @xmath84 , that is not written solely on @xmath76 , otherwise @xmath66 under @xmath77 .
in the view of lemma [ tree ] and theorem [ circuit ] , it is impossible that @xmath70 contains any relation @xmath85 such that @xmath86 is a positive word over @xmath76 and @xmath13 is a positive word over the edges that are not incident to any vertex of @xmath73 . furthermore @xmath73 is a full subgraph of @xmath5 .
thus @xmath84 must contain an edge that joins a vertex @xmath87 in @xmath73 to a vertex @xmath88 not in @xmath73 .
let @xmath16 be such an occurrence that comes last in @xmath84 .
since @xmath73 is connected , the vertex @xmath87 is joined to another vertex @xmath89 in @xmath73 by an edge @xmath18 .
as an automorphism of the punctured plane , @xmath64 does not change the circle @xmath74 because the edges for @xmath64 never touch @xmath74 . on the other hand we will show @xmath84 must change @xmath74 and
this is a contradiction because @xmath64 and @xmath84 are isotopic as automorphisms of the punctured disk .
the automorphism @xmath84 is the composition of the counterclockwise half twists along edges in @xmath84 .
then the counterclockwise half twist along @xmath16 creates an intersection @xmath3 of @xmath74 with the edge @xmath18 .
the intersection @xmath3 can disappear only via a clockwise half twist along an edge incident at either @xmath89 or @xmath87 as in figure [ f : circle ] .
but all of half twists in @xmath84 are counterclockwise because @xmath84 is a positive word over edges , an intersection between two adjacent edges always creates a pseudo face that must contain a vertex and so there is no intersection between adjacent edges in a linearly spanned graph .
suppose @xmath5 has more than one intersection .
figure [ f : oneinter ] shows a typical situation with one intersection and another intersection @xmath3 .
the edge @xmath18 must join vertices @xmath90 and @xmath91 , otherwise @xmath3 is an intersection between two adjacent edges .
one can easily check that all possibilities of completing @xmath18 create a pseudo face containing at least a vertex .
[ length ] let @xmath5 be a linearly spanned graph and @xmath69 be the set of generators given by edges of @xmath5 .
suppose that for each @xmath92 , there are positive words @xmath93 and @xmath94 over @xmath69 for @xmath95 in @xmath69 such that let @xmath101 be any positive finite presentation of the braid group .
the hypothesis ( iii ) implies that any shorter positive relation than @xmath85 itself can not make @xmath13 positively equivalent to @xmath86 .
choose a large @xmath102 such that @xmath85 is longer than any relation in @xmath70 .
then @xmath13 is not positively equivalent to @xmath86 over @xmath70 .
the following theorem completely determines when a linearly spanned graph with 4 vertices and no multiple edges has the embedding property . in the proof given below , @xmath103 or
@xmath104 will denote a positive word over the generators @xmath105 . in order to show that all other linearly spanned graphs with 4 vertices do not have the embedding property , we appeal to lemma [ length ] . to check the hypothesis ( iii ) of the lemma
, we use the fact that the band - generator presentation has the embedding property so that the positive equivalence in the presentation is the same as the equivalence in the braid group .
we also utilize the left and right cancellation theorem and the left and right canonical forms in the band - generator presentation in @xcite .
planar graphs with 4 vertices can be divided further into three types : graphs containing neither a triangle nor a rectangle , graphs containing at least a rectangle , and graphs containing at least a triangle but no rectangle , where a triangle ( or a rectangle ) is non - degenerate , that is , must have 3 ( or 4 , respectively ) vertices and must contain no other vertices inside .
[ [ i - graphs - containing - a - vertex - adjacent - to - all - of - remaining - three - vertices ] ] * ( i ) graphs containing a vertex adjacent to all of remaining three vertices * + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + we may have two possible graphs as in figure [ f : star1 ] up to equivalence .
for each of these two graphs , let @xmath106 and @xmath107 .
we will show that @xmath13 and @xmath86 satisfy lemma [ length ] .
add the edges @xmath108 so that these edges together with @xmath109 form an inner - complete graph
. then @xmath110 and so @xmath85 .
suppose @xmath111 for some positive word @xmath112 in the band - generator presentation
. then @xmath113 by the left cancellation in the band - generator presentation . but @xmath114 may only start with @xmath44 or @xmath115 which commute .
thus @xmath112 can not be written over @xmath109 . similarly @xmath116 .
thus @xmath13 satisfies ( iii ) of lemma [ length ] . due to the triangular relation @xmath117 , @xmath13 can not be equivalent to a positive word over @xmath118 that contains @xmath119 since @xmath13 is not positively equivalent to a positive word in the band - generator presentation that contains the subword @xmath120 .
thus @xmath121 also satisfy the hypothesis of lemma [ length ] over @xmath118 .
[ [ ii - graphs - containing - no - vertex - adjacent - to - all - of - remaining - three - vertices ] ] * ( ii ) graphs containing no vertex adjacent to all of remaining three vertices * + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + suppose @xmath126 for some positive word @xmath112 in the band - generator presentation .
then @xmath127 by the left cancellation in the band - generator presentation .
thus @xmath112 can not be written over @xmath128 . similarly @xmath129 .
thus @xmath13 satisfies ( iii ) of lemma [ length ] . due to the triangular relation @xmath130 , @xmath13 can not be equivalent to a positive word over @xmath131 that contains @xmath132 since @xmath13 is not positively equivalent to a positive word in the band - generator presentation that contains the subword @xmath133 .
thus @xmath121 also satisfy the hypothesis of lemma [ length ] over @xmath131 .
let @xmath137 .
. the right canonical form of @xmath139 is @xmath140 .
so @xmath139 may finish with @xmath141 over the band - generators and so @xmath44 or @xmath45 must appear in @xmath112 .
. in this case triangles must share at least 1 edge since we consider only 4 vertices . however
if two triangles share 2 edges then it must contain a multiple edge .
thus we have two possible graphs as in figure [ f : trian1 ] .
for these two graphs , let @xmath143 and @xmath144 .
the left canonical form of @xmath146 is @xmath147 and so it can not start with @xmath119 over the band - generators .
thus @xmath148 .
the right canonical form of @xmath149 is @xmath150 and so it can not finish with @xmath151 . thus @xmath152 . due to the triangular relation @xmath153
, @xmath13 can not be equivalent to a positive word over @xmath131 that contains @xmath154 since @xmath13 is not positively equivalent to a positive word in the band - generator presentation that contains the subword @xmath133 .
thus @xmath121 also satisfy the hypothesis of lemma [ length ] over @xmath131 .
the edges @xmath151 and @xmath115 that intersect each other are transformed to the diagonals of a rectangle .
then 4 more edges are needed to make an inner - complete graph .
we have four distinct types of graphs , depending on how many edges are missing from the inner - complete graph .
let @xmath160 .
then @xmath161 . to remove @xmath162 ,
@xmath163 must start with @xmath164 .
but the left canonical form of @xmath163 is @xmath165 and so it may start only with @xmath166 . thus @xmath167 .
similarly @xmath168 . the left canonical form of @xmath174 is @xmath175 and
so it can not start with @xmath162 over the band - generators .
thus @xmath176 .
the right canonical form of @xmath177 is @xmath178 and so it can not finish with @xmath119 . thus @xmath179 .
the left canonical form of @xmath185 is itself and so it can not start with @xmath162 over the band - generators .
thus @xmath186 .
let @xmath187 then @xmath188 and @xmath189 commute , so @xmath162 can not be removed .
thus @xmath190 .
we have already discussed about linearly spanned graphs with 3 vertices .
let @xmath5 be a linearly spanned graph with more than 3 vertices that has the embedding property .
first we choose a connect full subgraph @xmath73 with 4 vertices from @xmath5 so that there is a separating circle satisfying the hypothesis of theorem [ circle ] .
choose 4 vertices that form a connected subtree in a spanning tree of @xmath5 .
take the full subgraph with these 4 vertices . if there is no other vertices in faces of this full subgraph , then this is a desired full subgraph .
if there is other vertices in the faces of this full subgraph and none of them is adjacent to the chosen 4 vertices , then there is an edge - path starting at a vertex @xmath87 on a face and ending at one of the 4 vertices such that the edge - path intersect the full subgraph of the 4 vertices and so @xmath87 is contained in a pseudo face . since @xmath5 is linearly spanned , this can not happen .
thus at least one of vertices on faces , say @xmath88 , is adjacent to one of the 4 vertices
. then we have the less number of unwanted vertices contained in faces of a new connected full subgraph that is obtained by replacing one of the 4 vertices by @xmath88 . by repeating this process
, we eventually obtain a connect full subgraph @xmath73 with 4 vertices @xmath191 that can be separated by a circle from other vertices of @xmath5 .
theorem [ circle ] and theorem [ 4vertex ] say that @xmath73 must be either the artin graph or the inner - complete graph with 4 vertices .
choose a vertex @xmath192 in @xmath75 that is adjacent to one of @xmath191 .
if @xmath73 is an inner - complete graph , then each full subgraph with the 4 vertices consisted of @xmath192 and any three vertices from @xmath191 must be an inner - complete graph by theorem [ circle ] and theorem [ 4vertex ] .
thus the full subgraph with the 5 vertices @xmath193 is an inner - complete graph . by repeating this process , @xmath5 itself
eventually becomes an inner - complete graph .
we now suppose that @xmath73 is the artin graph with 4 vertices @xmath191 where @xmath194 are of valency 1 and the other vertices are of valency 2 .
choose a vertex @xmath192 in @xmath75 that is adjacent to any one of @xmath191 .
by theorem [ circle ] and theorem [ 4vertex ] , @xmath192 can be adjacent to either one or both of vertices @xmath194 .
if @xmath192 is adjacent to both vertices , we stop the process .
if @xmath192 is adjacent to one vertex , say @xmath195 , then @xmath196 form an artin graph and we repeat this process .
eventually we see that @xmath5 must contain one of the following two types of graphs as a full subgraph separated by a circle as in theorem [ circle ] unless @xmath5 itself is an artin graph .
the proof will be completed when we show that both types of graphs do not have the embedding property .
the graph of this type is depicted in figure [ f : kcycle ] .
let @xmath197 and @xmath198 .
then one can show that @xmath121 satisfy the hypothesis of lemma [ length ] by a similar but longer argument as for figure [ f : cycle ] .
thus @xmath5 does not have the embedding property we now show that @xmath203 suppose that @xmath204 then @xmath205 where @xmath206 . due to the triangular relation @xmath207 , the positive word @xmath208 must end with @xmath209 in order for @xmath210 to disappear .
we will show that @xmath208 can not end with @xmath209 by an induction on @xmath211 .
clearly @xmath212 can not end with @xmath164 in the band - generator presentation .
it is easy to check that @xmath214 ends with @xmath209 if and only if @xmath215 ends with @xmath209 if and only if @xmath216 ends with @xmath217 in the band - generator presentation . by the induction hypothesis
, @xmath216 can not end with @xmath217 .
similarly we have that @xmath218 consequently @xmath219 satisfy the hypothesis of lemma [ length ] .
mm e. artin , _ theori der zpfe _ ,
hamburg * 4 * ( 1925 ) , 4772 .
j. s. birman , _ braid groups and mapping class group _ , annals of math .
studies , no . * 82 * , princeton university press , 1974 .
j. s. birman , k. h. ko and s. j. lee , _ a new approach to the word and conjugacy problem in the braid groups _ , adv . math .
* 139 * ( 1998 ) 322353 a. h. clifford and g. b. preston , _ algebraic theory of semi - groups _
survey 7 ( 1961 ) .
e. a. elrifai and h. r. morton , _ algorithms for positive braids _ , quart
. j. math .
oxford * 20 * ( 1994 ) 479497 d. b. a. epstein ( with cannon , holt , levy , patterson and thurston ) , _ word processing in groups _ , jones and bartlett , boston , mass .
( 1992 ) , isbn 0 - 86720 - 244 - 0 f. a. garside , _ the braid group and other groups _ , quart . j. math .
oxford * 20 * ( 1969 ) 235254 .
j. w. han and k. h. ko , conjugate presentations of the braid groups , preprint .
e. s. kang , k. h. ko and s. j. lee , _ band - generator presentation for the 4-braid group _ , topology appl .
* 78 * ( 1997 ) 3960 j. h. remmers , _ on the geometry of semigroup presentations
_ , advance math . * 36 * , ( 1980 ) , 283296 .
v. sergiescu , _ graphes planaires et prsentations des groupes de tresses _ , math .
z. * 214 * ( 1993 ) 477490 .
h. tietze , _ ber die topologischen invarianten mehrdimensionalen mannigfaltigkeiten _ monatsh . f. math
. u. physik * 19 * ( 1908 ) | a large class of positive finite presentations of the braid groups is found and studied .
it is shown that no presentations but known exceptions in this class have the property that equivalent braid words are also equivalent under positive relations . |
for over the last forty years , the kondo effect has occupied a central place in condensed matter physics . while earlier studies of the effect have focused on its single - channel version realized in dilute magnetic alloys and valence - fluctuating systems , later attention has largely shifted to its more exotic multichannel variants where deviations from conventional fermi - liquid behavior can be found .
the overscreened kondo effect is a paradigmatic example for quantum criticality in quantum - impurity systems . besides the possible relevance to certain heavy fermion alloys , @xcite ballistic metal point contacts , @xcite scattering off two - level tunneling systems , @xcite and
the charging of small metallic grains , @xcite the overscreened kondo effect is one of the rare examples of interaction - driven non - fermi - liquid behavior that is well understood and well characterized theoretically .
@xcite the underscreened kondo effect , which might be realized in ultrasmall quantum dots with an even number of electrons , @xcite is a prototype for yet another form of unconventional behavior that of a singular fermi liquid .
@xcite these vastly different ground states , as well as that of an ordinary fermi liquid , can all be described within the single framework of the multichannel kondo hamiltonian , which is among the simplest yet richest models for strong electronic correlations in condensed - matter physics .
the multichannel kondo hamiltonian of eq .
( [ xxz - kondo ] ) describes the spin - exchange interaction of a spin-@xmath5 local moment with @xmath4 identical , independent bands of spin-@xmath0 electrons . for antiferromagnetic exchange ,
the low - energy physics features a subtle interplay between @xmath4 and @xmath5 , which could be summarized as follows .
@xcite for @xmath13 , the impurity spin is exactly screened .
a spin singlet progressively forms below a characteristic kondo temperature @xmath14 , leading to the formation of a local fermi liquid . for @xmath15
, the impurity spin is overscreened by the @xmath4 conduction - electron channels .
the system flows to an intermediate - coupling , non - fermi - liquid fixed point , characterized by anomalous thermodynamic and dynamical properties .
the elementary excitations are collective in nature , in contrast to the concept of a fermi liquid . for @xmath16 ,
the impurity spin is underscreened .
the low - energy physics comprises of quasiparticle excitations plus a residual moment of size @xmath17 .
however , it differs from a conventional fermi liquid in the singular energy dependence of the scattering phase shift and divergence of the quasiparticle density of states .
@xcite such behavior was recently termed a singular fermi liquid .
@xcite similar qualitative behavior is found for ferromagnetic exchange with arbitrary @xmath4 and @xmath5 , except that the residual moment has the full spin @xmath5 .
inherent to some of the leading scenarios for the realization of the multichannel kondo model @xcite is a large spin - exchange anisotropy .
an xxz anisotropy is well known to be irrelevant both in the single - channel ( @xmath18 , @xmath19 ) and the two - channel ( @xmath20 , @xmath19 ) cases .
the accepted phase diagram for these two models consists of an antiferromagnetic and a ferromagnetic domain , separated by a kosterliz - thouless line that traces @xmath21 at weak coupling . here
@xmath7 and @xmath8 are the longitudinal and transverse exchange couplings , respectively . as long as one lies within the confines of the antiferromagnetic domain , the system flows to the isotropic spin - exchange fixed point regardless of how large the exchange anisotropy is .
far less explored is the role of spin - exchange anisotropy for either @xmath9 or @xmath22 .
the stability of the overscreened fixed point against weak spin - exchange anisotropy has been analyzed in ref . using conformal field theory . for @xmath23 and either @xmath19 or @xmath24 ,
the non - fermi - liquid fixed point was found to be stable against a weak spin - exchange anisotropy .
this has led to the perception that spin - exchange anisotropy is irrelevant for these values of @xmath4 and @xmath5 .
in contrast , spin - exchange anisotropy is a relevant perturbation at the overscreened fixed point for all other values of @xmath25 ( assuming @xmath26 ; for @xmath27 it is a marginal perturbation ) , @xcite though the nature of the anisotropic fixed points was never explored . likewise unexplored
is the effect of spin - exchange anisotropy on the underscreened fixed point for @xmath28 . in this paper
, we revisit the phase diagram of the multichannel kondo model with an xxz spin - exchange anisotropy .
we find a far more complex picture than previously appreciated , including new coupling regimes where an xxz anisotropy substantially alters the low - energy physics .
our main findings are as follows . * for @xmath19 and @xmath22 ,
a second ferromagnetic - like domain is found deep in the antiferromagnetic regime .
the new domain extends above a ( typically large ) critical longitudinal coupling @xmath2 , and is separated from the conventional antiferromagnetic ( non - fermi - liquid ) domain by a second kosterliz - thouless line . * for spin @xmath9 and arbitrary @xmath4 , spin - exchange anisotropy is relevant near the free - impurity fixed point . for sufficiently small @xmath29
( @xmath30 is the conduction - electron density of states per lattice site ) , the system flows to a line of stable ferromagnetic - like fixed points with a residual isospin-@xmath0 moment . for sufficiently small @xmath31 ,
the flow is either to a conventional fermi liquid with no residual degeneracy ( for half - integer @xmath5 ) , or to a @xmath4-channel kondo effect with an effective spin-@xmath0 local moment ( for integer @xmath5 ) . here by sufficiently small @xmath32 and @xmath33 we mean the limit @xmath34 with any fixed ratio @xmath35 .
the closer @xmath36 is to one , the smaller the couplings must be for these results to apply .
* for @xmath6 and a sufficiently large @xmath3 , the system flows to a line of stable ferromagnetic - like fixed points with a residual isospin-@xmath0 moment . only for @xmath13 and @xmath37
are the exactly screened and overscreened fixed points stable against such a large anisotropy , which is marginal for @xmath38 . and @xmath4 .
( a ) @xmath19 and @xmath1 .
in addition to the well - established ferromagnetic domain for @xmath39 , a second ferromagnetic - like domain extends to the right and below the red line stretching from @xmath40 . within bosonization ,
@xmath41 is given by eq .
( [ j - ast ] ) .
( b ) @xmath42 .
in addition to the conventional ferromagnetic domain for @xmath39 , the system flows to a line of stable ferromagnetic - like fixed points with a residual isospin-@xmath0 local moment in two opposite limits : ( i ) for sufficiently small @xmath43 , ( ii ) for sufficiently large @xmath44 ( @xmath45 being the conduction - electron bandwidth ) . for weak @xmath46 ,
the system flows either to a fermi - liquid fixed point with a frozen impurity ( for integer @xmath5 ) , or to an overscreened @xmath4-channel , spin-@xmath0 kondo effect ( for half - integer @xmath5 ) .
the above weak - coupling behavior is generic to @xmath9 .
it extends to all cases described below , irrespective of @xmath4 .
( c ) @xmath47 .
in contrast to ( b ) , the isotropic overscreened fixed point is stable against a weak spin - exchange anisotropy ( i.e. , in the vicinity of the overscreened fixed point ) , as well as against a sufficiently large @xmath44 .
( d ) @xmath48 .
the exactly screened fixed point , corresponding to strong coupling , is stable against a sufficiently large anisotropy , @xmath44 .
however , it is unstable at weak coupling , as described above .
( e ) @xmath28 .
all underscreened cases flow to a line of stable ferromagnetic - like fixed points with a residual isospin-@xmath0 local moment both for sufficiently large @xmath49 and for sufficiently small @xmath43.,width=642 ] a compilation of these results is presented in fig .
[ fig : fig1 ] , in the form of phase diagrams for the different categories of @xmath5 and @xmath4 .
the phase diagrams for @xmath9 are still incomplete .
there remain extended coupling regimes where the low - energy physics is yet to be determined . to obtain these results ,
we employ a combination of perturbative renormalization - group ( rg ) techniques , abelian bosonization , a strong - coupling expansion in @xmath12 , and explicit numerical renormalization - group @xcite calculations . using perturbative
rg we first analyze in sec .
[ sec : perturbative - rg ] the limit of weak coupling .
in addition to the standard rg equations for the dimensionless couplings @xmath32 and @xmath33 , a new hamiltonian term proportional to @xmath50 is generated for @xmath9 and @xmath51 . here
@xmath52 is the @xmath53 component of the impurity spin .
depending on the sign of @xmath10 , the new hamiltonian term favors either the maximally polarized impurity states ( @xmath54 ) or the minimally polarized ones .
this leads to a qualitative distinction between @xmath43 and @xmath46 , and to the different low - temperature behaviors described above .
proceeding with abelian bosonization , we next show in sec .
[ sec : mapping ] that the anisotropic multichannel kondo hamiltonian with @xmath19 and @xmath23 possesses an exact mapping between the two sets of coupling constants @xmath55 and @xmath56 , where @xmath57 the above mapping is restricted to values of @xmath7 where the right - hand side of eq .
( [ mapping ] ) does not exceed @xmath58 .
for @xmath20 , this condition limits the validity of the mapping to @xmath59 , in which case eq .
( [ mapping ] ) simplifies to @xmath60 for @xmath1 , the mapping extends to negative values of @xmath7 , relating @xmath61 to @xmath62 and vice versa [ see eq .
( [ j_min ] ) for definition of @xmath63 .
thus , in contrast to common perception , the multichannel kondo hamiltonian with @xmath19 and @xmath1 possesses a line of stable ferromagnetic - like fixed points for @xmath64 .
note , however , that @xmath41 exceeds the bandwidth for intermediate values of @xmath4 , and is pushed to weak coupling for @xmath65 .
interestingly , the mapping of eq .
( [ mapping ] ) is ingrained in the anderson - yuval approach to the multichannel kondo problem , devised in refs . and .
in fact , it was already recognized by fabrizio _
et al . _ for @xmath20 , @xcite but
was never appreciated to our knowledge for @xmath1 . here
we provide an explicit operator mapping between the two sets of model parameters , for arbitrary @xmath23 .
since the critical coupling @xmath41 predicted by eq .
( [ mapping ] ) is typically larger than the bandwidth , one might question the applicability of bosonization and its unbound linear dispersion . could it be that @xmath66 when the conduction electrons are placed on a lattice ? to eliminate this concern , a strong - coupling expansion in @xmath12 is carried out in secs .
[ sec : sc - s=1/2 ] and [ sec : sc - general - s ] , first for @xmath19 and then for arbitrary @xmath5 .
the strong - coupling expansion not only confirms the existence of the new line of stable ferromagnetic - like fixed points for @xmath19 and @xmath1 , but further extends this result to arbitrary @xmath5 and @xmath4 obeying @xmath6 . for @xmath13 and @xmath37
, it establishes the irrelevance of such a large anisotropy . as an explicit demonstration of these ideas , the phase diagram of the @xmath19 , @xmath67 model
is studied in sec .
[ sec : nrg ] using wilson s numerical renormalization - group ( nrg ) method .
@xcite the nrg results confirm the phase diagram inferred from eq .
( [ mapping ] ) , including the order of magnitude of the critical coupling @xmath41 .
an extension of the mapping of eq .
( [ mapping ] ) to a spin - one impurity is presented in turn in appendix [ app : s=1 ] .
we conclude in sec .
[ sec : discussion ] with a discussion and summary of our results .
we begin our discussion with the limit of weak coupling , which is treated using perturbative rg . in the standard notation
, the multichannel kondo hamiltonian reads @xmath68 s_{z } \nonumber \\ & + & \frac{j_{\perp}}{2n } \sum_{n=1}^{k } \sum_{\vec{k } , \vec{k } ' } \left [ c^{\dagger}_{\vec{k } n \uparrow } c_{\vec{k } ' n \downarrow } s^{- } + c^{\dagger}_{\vec{k } n \downarrow } c_{\vec{k } ' n \uparrow } s^{+ } \right ] \!.
\label{xxz - kondo}\end{aligned}\ ] ] here , @xmath69 creates an electron with wave number @xmath70 and spin projection @xmath71 in the @xmath72th conduction - electron channel ; @xmath73 is a spin-@xmath5 operator ; @xmath7 and @xmath8 are the longitudinal and transverse kondo couplings , respectively ; and @xmath74 is the number of lattice sites . as a first step toward devising a perturbative rg treatment of the hamiltonian of eq .
( [ xxz - kondo ] ) , we convert the model to dimensionless form . to this end
, we introduce the fermion operators @xmath75 which represent the @xmath76 conduction - electron modes that couple to the impurity within the energy shell @xmath77 . here
@xmath45 is the conduction - electron bandwidth and @xmath78 is the conduction - electron density of states per lattice site : @xmath79 the dimensionless operators @xmath80 have been normalized to obey canonical anticommutation relations : @xmath81 assuming a box density of states @xmath82 and omitting all modes that decouple from the impurity , eq .
( [ xxz - kondo ] ) is recast in the form @xmath83 , where @xmath84 is the dimensionless hamiltonian @xmath85 s_z \nonumber \\ & + & \frac{\tilde{j}_{\perp}}{2 } \sum_{n=1}^{k } \int_{-1}^{1}\!d\varepsilon \int_{-1}^{1}\!d\varepsilon ' \left [ a^{\dagger}_{\varepsilon n \uparrow } a_{\varepsilon n \downarrow } s^{- } + { \rm h.c . }
\label{dimensionless - h}\end{aligned}\ ] ] here @xmath86 and @xmath87 are the dimensionless kondo couplings . focusing on @xmath88 , we treat the hamiltonian of eq .
( [ dimensionless - h ] ) using perturbative rg .
the rg transformation consists of the following three steps .
suppose that the bandwidth has already been lowered from its initial value @xmath45 to some value @xmath89 ( @xmath90 ) .
further lowering the bandwidth to @xmath91 requires the elimination of all @xmath92 degrees of freedom in the interval @xmath93 .
this goal is accomplished using anderson s poor - man s scaling .
@xcite at the conclusion of this step , all integrations in eq .
( [ dimensionless - h ] ) have been reduced to the range @xmath94 .
the rg transformation is completed by ( i ) rescaling @xmath95 to account for the reduced bandwidth , and ( ii ) restoring the original integration ranges in eq .
( [ dimensionless - h ] ) by converting to @xmath96 this latter step allows us to write @xmath97 the dimensionless hamiltonian is recast in this manner in a self - similar form , but with renormalized couplings .
these obey a set of coupled differential equations , specified below . for @xmath19
we recover the familiar rg equations @xmath98 showing that spin - exchange anisotropy is irrelevant for weak @xmath99 , irrespective of the number of channels @xmath4 .
a different qualitative picture emerges for spin @xmath5 larger than one - half . here a new hamiltonian term @xmath100 is generated within @xmath84 .
starting from zero , @xmath101 renormalizes according to @xmath102 , \label{d - delta - dl}\ ] ] which supplements eqs .
( [ rg - z ] ) and ( [ rg - perp ] ) for @xmath103 and @xmath104 . restricting attention to @xmath9
, we now analyze the ramifications of the new coupling @xmath101 as a function of @xmath5 , @xmath4 , and the bare kondo couplings @xmath7 and @xmath8 .
since @xmath105 is conserved under the rg , a negative ( positive ) @xmath101 is generated if the bare kondo couplings satisfy @xmath106 ( @xmath107 ) .
for a given ratio @xmath108 and sufficiently small @xmath109 , the coupling @xmath110 approaches unity well before any kondo temperature can be reached .
this has the effect of freezing all but the lowest - lying spin states . for @xmath106 ( negative @xmath101 ) ,
only the maximally polarized states @xmath54 are thus left . for @xmath107 ( positive @xmath101 ) ,
the picture depends on @xmath5 : for half - integer @xmath5 , the two degenerate states @xmath111 are selected ; for integer @xmath5 , only the state @xmath112 remains .
depending on which case is realized , a different effective low - energy hamiltonian emerges .
consider first the case @xmath106 ( negative @xmath101 ) .
introducing the isospin operators @xmath113 @xmath114 the resulting low - energy hamiltonian contains the term @xmath115 \tau_z , \label{h_jz_1}\ ] ] which follows from projection of the spin - exchange interaction of eq .
( [ dimensionless - h ] ) onto the @xmath54 subspace . here
@xmath116 is the running coupling constant at the scale where @xmath110 approaches one .
in contrast to eq .
( [ h_jz_1 ] ) , terms that flip the isospin @xmath117 involve the creation and annihilation of at least @xmath11 conduction electrons .
this follows from the fact that a flip in @xmath117 changes @xmath52 by @xmath118 .
since the hamiltonian conserves the total spin projection of the entire system ( impurity plus conduction electrons ) , such a process must be accompanied by an opposite spin flip of @xmath11 conduction electrons in the vicinity of the impurity . as result ,
isospin - flip terms have the scaling dimension @xmath119 about the free - impurity fixed point , rendering them irrelevant .
the resulting fixed - point structure corresponds then to a line of ferromagnetic - like fixed points with a residual isospin-@xmath0 local moment , characterized by a finite @xmath103 . note that this picture is insensitive to the sign of the kondo couplings .
it equally applies to positive and negative @xmath7 and @xmath8 . proceeding with the case @xmath107 ( positive @xmath101 )
, we first consider half - integer @xmath5 . here
the two states selected by @xmath101 are @xmath111 .
similar to eqs .
( [ tau - z - def ] ) and ( [ tau - perp - def ] ) , we introduce the isospin operators @xmath120 @xmath121 which now relate to the states @xmath122 .
projection of eq .
( [ dimensionless - h ] ) onto the @xmath111 subspace yields the following effective spin - exchange interaction : @xmath123 \tau_z \nonumber \\ & + & \gamma \frac{\tilde{j}_{\perp}}{2 } \sum_{n=1}^{k } \int_{-1}^{1}\!d\varepsilon\ !
\int_{-1}^{1}\!d\varepsilon ' \left [ a^{\dagger}_{\varepsilon n \uparrow } a_{\varepsilon n \downarrow } \tau^{-}\ ! + { \rm h.c . } \right ] \label{h_j_2}\end{aligned}\ ] ] with @xmath124 .
once again , @xmath103 and @xmath104 in eq .
( [ h_j_2 ] ) are the running coupling constants at the scale where @xmath101 approaches one .
equation ( [ h_j_2 ] ) has the form of a @xmath4-channel kondo hamiltonian with an effective isospin-@xmath0 local moment . since latexmath:[$|\tilde{j}_z| <
\gamma antiferromagnetic domain . hence , irrespective of the original spin @xmath5 , the system flows to the overscreened fixed point of the @xmath4-channel kondo effect with spin @xmath126 .
( for @xmath18 , the flow is to the strong - coupling fixed point of the conventional one - channel kondo effect ) . excluding the case @xmath127 with antiferromagnetic exchange , the resulting low - energy physics differs markedly from that of the isotropic model , whether ferromagnetic or antiferromagnetic .
most strikingly , the underscreened fixed point for @xmath128 and @xmath129 gives way to an overscreened fixed point for any given ratio @xmath130 and sufficiently small @xmath109 .
of the different possible cases for @xmath101 and @xmath5 , the simplest picture is recovered for @xmath107 and integer @xmath5 . here
the impurity spin is frozen in the @xmath112 state , loosing its dynamics .
this results in a conventional fermi - liquid fixed point with neither non - fermi - liquid characteristics nor a residual local - moment degeneracy .
as is evident from the discussion above , there is a qualitative difference between @xmath19 and @xmath9 concerning the effect of spin - exchange anisotropy for weak coupling . in this section
we focus on @xmath19 and use abelian bosonization to derive the mapping of eq .
( [ mapping ] ) .
our starting point is the continuum - limit representation of the multichannel kondo hamiltonian @xmath131 \nonumber \\ & + & \sum_{n = 1}^{k } \frac{j_z a}{2 } s_z\ !
\left [ \psi^{\dagger}_{n \uparrow}(0 ) \psi_{n \uparrow}(0)\ ! -\
! \psi^{\dagger}_{n \downarrow}(0 ) \psi_{n \downarrow}(0 ) \right ] - g_i h s_z \nonumber \\ & - & \frac{g_e h}{2 } \sum_{n = 1}^{k } \int_{-\infty}^{\infty } \left [ \psi^{\dagger}_{n \uparrow}(x ) \psi_{n \uparrow}(x ) - \psi^{\dagger}_{n \downarrow}(x ) \psi_{n \downarrow}(x ) \right ] dx , \nonumber\end{aligned}\ ] ] written in terms of the left - moving one - dimensional fields @xmath132 with @xmath133 and @xmath134 . here
, @xmath73 is a spin-@xmath0 operator ; @xmath135 is a short - distance cutoff corresponding to a lattice spacing ; @xmath136 is a fictitious coordinate conjugate to @xmath137 ; and @xmath138 is an applied magnetic field . for the sake of generality ,
we allow for different impurity and conduction - electron land @xmath139-factors , @xmath140 and @xmath141 , respectively . to treat the hamiltonian of eq .
( [ h ] ) , we resort to abelian bosonization . according to standard prescriptions ,
@xcite @xmath76 boson fields are introduced one boson field @xmath142 for each left - moving fermion field @xmath143 .
the fermion fields are written as @xmath144 where the @xmath145 fields obey @xmath146 = -i \delta_{n n ' } \delta_{\sigma \sigma ' } \pi\ , { \rm sign } ( x - y ) .\ ] ] the ultraviolet momentum cutoff @xmath147 is related to the conduction - electron bandwidth @xmath45 and the density of states per lattice site @xmath30 through @xmath148 and @xmath149 , respectively .
the operators @xmath150 in eq .
( [ psi - via - phi ] ) are phase - factor operators , which come to ensure that the different fermion species anticommute .
our explicit choices for these operators are @xmath151 } , \ ] ] where @xmath152 is the number operator for electrons in channel @xmath153 with spin projection @xmath71 . in terms of the boson fields ,
the multichannel kondo hamiltonian takes the form @xmath154 } s^{- } + { \rm h.c . }
\right \ } \nonumber \\ & + & \delta_z\frac{a}{\pi^2 \rho } \sum_{n = 1}^{k } \left [ \nabla \phi_{n \uparrow}(0 ) - \nabla \phi_{n \downarrow}(0 ) \right ] s_z - g_i h s_z \nonumber \\ & - & \frac{g_e h}{4\pi } \sum_{n = 1}^{k } \int_{-\infty}^{\infty } \left [ \nabla \phi_{n \uparrow}(x ) - \nabla \phi_{n \downarrow}(x ) \right ] dx , \label{h - bosonized}\end{aligned}\ ] ] where @xmath155 is the parallel - spin phase shift induced by @xmath7 in the absence of @xmath8 .
note that @xmath156 is bounded in magnitude by @xmath58 , which stems from the cutoff scheme used in bosonization .
although the bosonic hamiltonian of eq . ( [ h - bosonized ] ) does support larger values of @xmath157 , this parameter must not exceed @xmath58 in order for @xmath158 to possess a fermionic counterpart of the form of eq .
( [ h ] ) .
we shall return to this important point later on . at this stage
we manipulate the bosonic hamiltonian of eq .
( [ h - bosonized ] ) through a sequence of steps .
we begin by converting to @xmath76 new boson fields , consisting of @xmath159 \label{phi_s - def}\ ] ] plus @xmath160 orthogonal fields : @xmath161 with @xmath162 .
the orthogonal fields @xmath161 are formally expressed as @xmath163 where the real coefficients @xmath164 obey @xmath165 @xmath166 = 0 .\ ] ] the precise form of the coefficients @xmath164 is of no practical importance to our discussion , and need not concern us .
their choice is not unique . in terms of the new fields
, the combinations @xmath167 take the form @xmath168 where @xmath169 is some linear combination of the fields @xmath161 with @xmath162 .
once again , the precise form of the @xmath169 combinations is of no real significance to our discussion
. we shall only rely on them being orthogonal to @xmath170 . using the new boson fields defined above , the hamiltonian of eq .
( [ h - bosonized ] ) is converted to @xmath171 dx \nonumber \\ & + & \sum_{n = 1}^{k } \frac{j_{\perp}}{2 } \left \ { e^{i\sqrt{\frac{2}{k } } \phi_s(0 ) + i\varphi_n(0 ) } s^{- } + { \rm h.c . }
\right \ } \nonumber \\ & + & \delta_z \frac{a}{\pi^2 \rho } \sqrt{2k } \nabla \phi_{s}(0 ) s_z - g_i h s_z \nonumber \\ & - & \frac{g_e h}{4\pi } \sqrt{2 k } \int_{-\infty}^{\infty } \nabla \phi_{s}(x ) dx .
\label{h - via - phi_s}\end{aligned}\ ] ] next the canonical transformation @xmath172 with @xmath173 $ ] is applied to obtain @xmath174 dx \nonumber \\
& + & \sum_{n = 1}^{k } \frac{j_{\perp}}{2 } \left \ { e^{-i\sqrt{\frac{2}{k } } \phi_s(0 ) + i\varphi_n(0 ) } s^{- } + { \rm h.c . } \right \ } \nonumber \\ & + & \left ( \delta_z - \frac{\pi}{k } \right ) \frac{a}{\pi^2 \rho } \sqrt{2k } \nabla \phi_{s}(0 ) s_z - ( g_i - 2g_e ) h
s_z \nonumber \\ & - & \frac{g_e h}{4\pi } \sqrt{2 k } \int_{-\infty}^{\infty } \nabla \phi_{s}(x ) dx . \label{h'}\end{aligned}\ ] ] here we have omitted a constant term from @xmath175 , and made use of the identity @xmath176 in writing the term @xmath177 .
finally , the transformation @xmath178 is introduced , which yields @xmath179 dx \nonumber \\ & + & \sum_{n = 1}^{k } \frac{j_{\perp}}{2 } \left \ { e^{i\sqrt{\frac{2}{k } } \phi_s(0 ) + i\varphi_n(0 ) } s^{- } + { \rm h.c . }
\right \ } \nonumber \\ & + & \delta'_z \frac{a}{\pi^2 \rho } \sqrt{2k } \nabla \phi_{s}(0 ) s_z + ( 2 g_e - g_i ) h s_z
\nonumber \\ & + & \frac{g_e h}{4\pi } \sqrt{2 k } \int_{-\infty}^{\infty } \nabla \phi_{s}(x ) dx \label{h - final}\end{aligned}\ ] ] with @xmath180 .
the hamiltonian of eq .
( [ h - final ] ) is identical to that of eq .
( [ h - via - phi_s ] ) , apart from a renormalization of certain parameters : @xmath181 , @xmath182 , and @xmath183 . as long as @xmath184
, one can revert the series of steps leading to eq .
( [ h - via - phi_s ] ) , to recast the hamiltonian of eq .
( [ h - final ] ) in fermionic form .
the end result is just the original multichannel kondo hamiltonian of eq .
( [ h ] ) with the following renormalized parameters : @xmath185 @xmath186 @xmath187 for zero magnetic field , this establishes the mapping of eq .
( [ mapping ] ) , including the restriction to values of @xmath7 where the right - hand side of eq .
( [ mapping ] ) does not exceed @xmath58 .
the latter condition is just a restatement of the requirement @xmath184 .
we now analyze in detail the ramifications of this restriction as a function of the number of channels @xmath4 . for @xmath18 ( single - channel case )
, @xmath188 exceeds @xmath58 for all @xmath189 .
thus , the mapping of eq .
( [ mapping ] ) does not apply to the single - channel kondo effect , in accordance with known results . for @xmath20 ( two - channel case ) ,
the required condition is met for all @xmath59 , mapping weak to strong coupling and vice versa [ see eq .
( [ 2chk ] ) ] .
hence , the mapping of eq .
( [ mapping ] ) can be viewed as an anisotropic variant @xcite of the weak - to - strong - coupling duality of nozires and blandin .
@xcite the most interesting case occurs for @xmath1 , when eq .
( [ mapping ] ) extends to all @xmath190 with @xmath191 in particular , the range @xmath64 with @xmath192 is mapped onto the negative - coupling regime @xmath193 and vice verse .
thus , the kosterliz - thouless line separating the antiferromagnetic and ferromagnetic domains is duplicated from @xmath21 to @xmath194 with @xmath195 here we have assumed @xmath196 , in order for the weak - coupling parameterization of the kosterliz - thouless line to apply . the resulting phase diagram for @xmath1 is plotted in fig .
[ fig : fig1](a ) . for @xmath197 ,
the multichannel kondo hamiltonian flows to a line of stable ferromagnetic - like fixed points , rendering the non - fermi - liquid fixed point of the model unstable against a large enough anisotropy .
this behavior contradicts the common perception of spin - exchange anisotropy as irrelevant for @xmath19 .
note that the same phase diagram emerges from the renormalization - group equations derived by ye using the equivalent anderson - yuval approach .
@xcite however , the newly found line of stable ferromagnetic - like fixed points has eluded ye .
a potential concern with the above picture for @xmath19 has to do with the validity of the bosonization approach used .
since @xmath41 exceeds the bandwidth @xmath45 for intermediate values of @xmath4 , one might wonder to what extent is bosonization ( or the anderson - yuval approach for that matter ) justified for such strong coupling . in fact , the very usage of the continuum - limit hamiltonian of eq .
( [ h ] ) with its unbounded linear dispersion can be called into question .
could it be that @xmath198 when the conduction electrons are placed on a lattice ? not withstanding the observation that @xmath41 for @xmath65 is pushed to weak coupling , to firmly establish the phase diagram of fig .
[ fig : fig1](a ) one must show that the new line of stable ferromagnetic - like fixed points extends to proper lattice models for the underlying conduction bands .
this is the objective of the following analysis , which focuses on the limit of a large longitudinal coupling , @xmath199 . as a generic lattice model for the conduction bands ,
we consider a spin-@xmath0 impurity moment coupled to the open end of a semi - infinite tight - binding chain with @xmath4 identical conduction - electron species : @xmath200 \nonumber \\ & + & \frac{j_{\perp}}{2 } \sum_{n = 1}^{k } \left [ f^{\dagger}_{0n\downarrow}f_{0n\uparrow}s^+ + { \rm h.c . }
\right ] \nonumber \\ & + & \frac{j_z}{2 } \sum_{n = 1}^{k } s_z\ ! \left [ f^{\dagger}_{0 n \uparrow } f_{0 n \uparrow } - f^{\dagger}_{0 n \downarrow } f_{0 n \downarrow } \right ] - g_i h s_z \nonumber \\ & - & \frac{g_e h}{2 } \sum_{j = 0}^{\infty } \sum_{n = 1}^{k } \left [ f^{\dagger}_{j
n \uparrow } f_{j n \uparrow } - f^{\dagger}_{jn\downarrow } f_{jn\downarrow } \right ] .
\label{h - lattice}\end{aligned}\ ] ] here @xmath201 and @xmath202 , respectively , are the on - site energies and hopping matrix elements along the chain . any lattice model with identical noninteracting bands
can be cast in the form of eq .
( [ h - lattice ] ) using a wilson - type construction .
@xcite different lattice models are distinguished by the tight - binding parameters @xmath201 and @xmath202 , the largest of which determines the bandwidth @xmath45 .
for example , particle - hole symmetry demands that @xmath203 for all sites along the chain . in the following
we assume a large longitudinal coupling , @xmath199 , and expand in powers of @xmath12 to derive an effective low - energy hamiltonian at energies far below @xmath7 . for @xmath204 ,
the fermionic degrees of freedom at site zero bind tightly to the impurity so as to minimize the @xmath7 interaction term .
there are two degenerate ground states of this interaction term : @xmath205 corresponding to the total spin projections @xmath206 .
all excited eigenstates of the @xmath7 interaction term are thermally inaccessible , being removed in energy by integer multiples of @xmath207 . defining a new isospin operator @xmath117 according to @xmath208
[ compare with eqs .
( [ tau - z - def])([tau - perp - def ] ) and eqs .
( [ tau - z - def2])([tau - perp - def2 ] ) ] , the effective low - energy hamiltonian takes the form @xmath209 , where @xmath210 is the tight - binding hamiltonian of the truncated chain with site @xmath211 removed , @xmath212 is the magnetic - field term @xmath213 \label{h - mag}\ ] ] with @xmath214 , and @xmath215 contains all finite - order corrections in either @xmath12 or @xmath8 ( or both ) .
the explicit form of the hamiltonian term @xmath215 depends on the number of conduction - electron channels @xmath4 . for @xmath18 and up to linear order in @xmath12
, it takes the form @xmath216 , \ ] ] where @xmath217 equals @xmath218 here we have omitted the redundant channel index @xmath72 from within @xmath219 , i.e. , we have set @xmath220 .
since the @xmath217 term is exactly marginal , the low - temperature physics is governed for @xmath221 by the spin - flip term @xmath8 , which favors the singlet state @xmath222 $ ] for @xmath223 . in this manner one
recovers the characteristic spin singlet of the single - channel kondo effect , restoring thereby su(2 ) spin symmetry as @xmath224 .
a local magnetic field breaks the emerging su(2 ) spin symmetry as it physically should by introducing weak spin - dependent scattering at the open end of the truncated chain .
the situation is somewhat different for @xmath20 . in this case
@xmath215 acquires the form @xmath225 \nonumber \\ & + & \lambda_{\perp } \sum_{n = 1}^{2 } \left [ f^{\dagger}_{1 n\uparrow } f_{1 n\downarrow } \tau^- + { \rm h.c . } \right ] , \label{h_int-2ck}\end{aligned}\ ] ] where @xmath217 is still given to order @xmath12 by eq .
( [ lambda_z ] ) , and @xmath226 scales as @xmath227 here we have omitted higher order interaction terms within @xmath215 , and restricted attention to particle - hole symmetry . away from particle - hole symmetry ,
an additional potential - scattering term of the form @xmath228 is generated at second order in @xmath12 .
equation ( [ h_int-2ck ] ) has the same exact form as the original spin - exchange interaction in eq .
( [ h - lattice ] ) , but with two modifications : ( i ) the coupling constants @xmath7 and @xmath8 have been pushed to weak coupling ; ( ii ) the site @xmath211 has been replaced with @xmath229 .
the physical content of the local moment @xmath117 has also changed .
importantly , since @xmath230 one lies within the confines of the antiferromagnetic domain , rendering @xmath215 a relevant perturbation .
hence , in accordance with the bosonization treatment of sec .
[ sec : mapping ] , the strong - coupling limit @xmath231 maps onto weak coupling , extending the duality of nozires and blandin @xcite to large spin - exchange anisotropy . moreover , since @xmath232 for conventional lattice models , then @xmath217 of eq .
( [ lambda_z ] ) is comparable to @xmath233 of eq .
( [ 2chk ] ) .
the main difference as compared to bosonization pertains to the transverse coupling @xmath8 , which renormalizes to @xmath234 in the strong - coupling expansion , but is left unchanged within bosonization . excluding this rather minor discrepancy , the two approaches are in close agreement with one another .
the crucial difference for @xmath1 has to do with the dynamic components of @xmath215 , responsible for flipping the isospin @xmath117 . to see this we note that the total spin projections @xmath235 of the states @xmath236 and @xmath237 differ by @xmath238 . since the hamiltonian of eq .
( [ h - lattice ] ) preserves the overall spin projection of the entire system , then the flipping of @xmath239 from plus to minus or vice versa must be accompanied by an opposite spin flip of @xmath238 electrons along the truncated chain .
hence the leading dynamic term in @xmath215 only shows up at order @xmath240 , taking the form @xmath241 \label{lambda_p - k}\ ] ] with @xmath242 in contrast to the dynamic part , the leading static component of @xmath215 remains given by the @xmath217 term of eq .
( [ h_int-2ck ] ) , except for the summation over @xmath72 which now runs over all @xmath4 channels . to linear order in @xmath12
the coupling @xmath217 is independent of @xmath4 , being given by eq .
( [ lambda_z ] ) .
once again , an additional potential - scattering term is generated at order @xmath243 away from particle - hole symmetry . since the @xmath226 term of eq .
( [ lambda_p - k ] ) involves the creation and annihilation of @xmath244 electrons at the open end of the truncated chain , it has the scaling dimension @xmath245 with respect to the @xmath246 `` free '' hamiltonian .
hence this term is irrelevant for @xmath1 .
the same holds true of all higher order dynamical terms generated , as these likewise contain at least @xmath244 creation and @xmath244 annihilation operators of electrons localized along the truncated chain .
similar to the case @xmath9 and @xmath106 in sec .
[ sec : perturbative - rg ] , the resulting fixed - point hamiltonian for @xmath221 corresponds then to a finite longitudinal coupling @xmath217 but zero @xmath226 , in perfect agreement with the results of bosonization .
in fact , an identical scaling dimension @xmath247 is obtained in the anderson - yuval approach for @xmath248 , @xcite reinforcing the qualitative agreement between the two approaches .
most importantly , a line of stable ferromagnetic - like fixed points is seen to exist for any lattice model with sufficiently large @xmath249 , just as predicted by bosonization .
evidently , the strong - coupling expansion confirms the results of bosonization for all three cases : @xmath18 , @xmath20 , and @xmath1 .
moreover , it provides a transparent physical picture for the source of distinction between the three cases .
it all boils down to the nature of the local spin configurations selected by a sufficiently large @xmath249 .
we therefore conclude that fig .
[ fig : fig1](a ) correctly describes the phase diagram of the spin-@xmath0 multichannel kondo model with @xmath1 conduction - electron channels .
our discussion in the previous two sections was confined to a spin-@xmath0 impurity .
an appealing feature of the strong - coupling expansion in @xmath12 is that it can easily be generalized to arbitrary spin @xmath5 .
this is the goal of the present section .
the basic considerations for arbitrary @xmath250 are quite similar to those for @xmath19 .
as before , the @xmath7 interaction term possesses two degenerate ground states .
we label these states according to @xmath251 for @xmath252 , and @xmath253 for @xmath16 . with this convention
, @xmath254 has the total spin projection @xmath255 . defining the isospin @xmath117 according to eq .
( [ tau - def ] ) , the effective low - energy hamiltonian is written as @xmath209 , where @xmath210 is the tight - binding hamiltonian of the truncated chain with site @xmath211 removed , @xmath212 is the magnetic - field term of eq .
( [ h - mag ] ) , and @xmath215 contains all finite - order corrections in either @xmath12 or @xmath8 ( or both ) .
the sole modification to @xmath212 of eq .
( [ h - mag ] ) is in the effective @xmath139-factor @xmath256 , which changes from @xmath257 for @xmath19 to @xmath258 for arbitrary @xmath5 . here
the plus ( minus ) sign corresponds to @xmath259 ( @xmath16 ) .
similar to the case @xmath19 , the delicate interplay between @xmath4 and @xmath5 enters through the hamiltonian term @xmath215 .
let us separate the discussion of the dynamic and static components of @xmath215 , as these depend differently on @xmath5 and @xmath4 .
the leading static component of @xmath215 remains given by the @xmath217 term of eq .
( [ h_int-2ck ] ) , except for the summation over @xmath72 which now runs over all @xmath4 conduction - electron channels .
the coupling @xmath217 does depend on @xmath260 , however only through its sign .
for @xmath252 it is given to order @xmath12 by eq .
( [ lambda_z ] ) , corresponding to an antiferromagnetic interaction .
for @xmath16 it acquires an additional minus sign , corresponding to a ferromagnetic interaction .
apart from its overall sign , the leading static component of @xmath215 is independent of @xmath5 . moving on to the dynamic part of @xmath215 ,
its leading - order component displays a more elaborate dependence on @xmath4 and @xmath5 .
as noted above , the total spin projections of the states @xmath236 and @xmath237 differ by @xmath261 . consequently
, the flipping of @xmath239 from up to down or vice versa must be accompanied by an opposite spin flip of @xmath261 electrons along the truncated chain , else the total spin projection of the system is not conserved .
this consideration dictates the following form for the leading dynamical term : @xmath262 , \ ] ] where @xmath263 is a channel - symmetric operator that creates @xmath261 spin - up electrons and annihilates @xmath261 spin - down electrons at , or close as possible to , the open end of the truncated chain .
the operator @xmath264 is generally too complicated to write down .
it greatly simplifies in two cases . for @xmath13
, @xmath264 reduces to the unity operator ; for @xmath19 and @xmath23 , it is given by eq .
( [ lambda_p - k ] ) .
as for the coupling @xmath226 , it scales differently for @xmath265 and @xmath266 . for @xmath265 , @xmath226 behaves as @xmath267 for @xmath266 , it involves a higher power of @xmath12 , as electrons farther into the chain must participate in the flipping of @xmath239 . with the possible exception of @xmath13 ,
the transverse coupling @xmath268 is parametrically smaller than @xmath269 , a fact that will have important implications later on . as a function of @xmath5 and @xmath4
, @xmath226 has the scaling dimension @xmath270 with respect to the @xmath246 `` free '' hamiltonian . for @xmath221 ,
this yields the following classification of the low - energy physics .
\(i ) @xmath48 in this exactly screened case , @xmath226 acts as a local transverse magnetic field , lifting the two - fold degeneracy of @xmath254 .
although the local state favored by @xmath226 is generally not an su(2 ) spin singlet , the low - energy physics is identical to that for isotropic antiferromagnetic exchange , as can be seen from the boundary condition imposed on the truncated chain .
specifically , a local fermi liquid progressively forms below the kondo temperature @xmath271 .
\(ii ) @xmath272 in this overscreened case , the hamiltonian term @xmath215 has the same exact form as eq .
( [ h_int-2ck ] ) , except for the summation over @xmath72 which now runs over all @xmath4 conduction - electron channels .
hence the system is described by a weakly coupled @xmath4-channel kondo hamiltonian with @xmath273 .
the effect of the large spin - exchange anisotropy in the original hamiltonian is to reduce the effective impurity moment from @xmath5 to @xmath274 .
based on the perturbative rg analysis of sec .
[ sec : perturbative - rg ] , the resulting hamiltonian flows to the overscreened fixed point of the @xmath4-channel , spin-@xmath0 kondo hamiltonian , which is equivalent in turn to that of the @xmath4-channel kondo model with @xmath272 .
as in the exactly screened case discussed above , a large spin - exchange anisotropy @xmath199 is seen to be irrelevant for @xmath272 .
\(iii ) @xmath275 similar to the previous case , the system is described by a weakly coupled @xmath4-channel kondo hamiltonian with @xmath276 . however , the effective coupling is now ferromagnetic : @xmath277 .
consequently , the flow is to a line of stable ferromagnetic - like fixed points with finite @xmath217 but zero @xmath226 .
the resulting low - energy physics is that of a singular fermi liquid @xcite plus a residual isospin @xmath117 .
it differs from that of an isotropic spin - exchange interaction only in the residual @xmath217 interaction . since the latter term is marginal , so is the large spin - exchange anisotropy for this underscreened case .
, as a function of @xmath4 and @xmath5 . for @xmath278 and @xmath279 ,
the low - energy physics is stable against such a large spin - exchange anisotropy .
the fixed point remains that of an exactly screened ( @xmath278 ) or an overscreened ( @xmath279 ) kondo effect . for @xmath6 ,
the low - energy physics is unstable against such a large spin - exchange anisotropy .
the system flows to a line of stable ferromagnetic - like fixed points with a residual isospin-@xmath0 local moment . for @xmath280 ,
a large spin - exchange anisotropy is marginal .
the system still flows to a line of stable ferromagnetic - like fixed points , however these differ from the isotropic underscreened fixed point only by the marginal operator @xmath217 ( see text).,width=302 ] \(iv ) @xmath6 in this effectively underscreened case , the scaling dimension @xmath281 exceeds one .
hence @xmath226 is irrelevant , as are all higher order dynamical terms generated within @xmath215 .
the system thus flows to a line of stable ferromagnetic - like fixed points , characterized by a finite @xmath217 but zero @xmath226 .
the low - energy physics is again that of a ( potentially singular ) fermi liquid plus a residual isospin @xmath117 . for @xmath282 ,
this differs markedly from the overscreened fixed point of the isotropic spin - exchange model .
also for @xmath283 this differs from the underscreened fixed point of the isotropic model , as the residual local - moment degeneracy is two instead of @xmath284 .
importantly , in all cases the resulting low - energy physics is insensitive to the sign of @xmath260 , in stark contrast to the isotropic case .
a large spin - exchange anisotropy of the form @xmath199 is therefore relevant with respect to the isotropic fixed point for all @xmath6 .
the phase diagram of the multichannel kondo hamiltonian with @xmath44 is summarized in fig .
[ fig : fig2 ] as a function of @xmath4 and @xmath5 .
excluding the exactly screened line @xmath278 and the overscreened line @xmath279 , the model flows to a line of stable ferromagnetic - like fixed points with a residual isospin-@xmath0 local moment for all @xmath4 and @xmath5 . for @xmath280 ,
the ferromagnetic - like fixed points and the isotropic underscreened fixed point are equivalent .
they only differ by the marginal operator @xmath217 . for @xmath285 ,
the ferromagnetic - like fixed points and the isotropic underscreened fixed point are distinctly different , possessing a different residual degeneracy .
although the strong - coupling expansion unequivocally confirms the existence of a line of stable ferromagnetic - like fixed points for large @xmath7 and @xmath6 , it can not access the entire phase diagram of the anisotropic multichannel kondo hamiltonian .
in particular , the second kosterliz - thouless line predicted by bosonization for @xmath19 and @xmath1 [ see fig . [
fig : fig1](a ) ] lies beyond the scope of this approach . in this section ,
we use wilson s numerical renormalization - group ( nrg ) method @xcite to conduct a systematic study the phase diagram of the anisotropic multichannel kondo hamiltonian with @xmath19 and @xmath67 . in conventional formulations of the nrg ,
@xcite one considers a particular choice for the tight - binding parameters in eq .
( [ h - lattice ] ) , given by @xmath203 and @xmath286 with @xmath287 @xmath288 here @xmath289 is a discretization parameter . for @xmath290 ,
the resulting model describes an impurity spin locally coupled to @xmath4 identical conduction bands with a symmetric box density of states @xmath291 . for @xmath289 ,
the model represents an impurity spin coupled to a logarithmically discretized version of the same bands .
@xcite the first step in the nrg approach is to recast the hamiltonian @xmath158 as the limit of a sequence of dimensionless hamiltonians @xmath292 : @xmath293 with @xmath294 .
\nonumber\end{aligned}\ ] ] the finite - size hamiltonians @xmath292 are then diagonalized iteratively using the nrg transformation @xmath295 here the prefactor @xmath296 in eq .
( [ h - nrg ] ) guarantees that the low - energy excitations of @xmath292 are of order one for all @xmath74 .
the approach to a fixed - point hamiltonian is signaled by a limit cycle of the nrg transformation , with @xmath297 and @xmath292 sharing the same low - energy spectrum .
it practice , it is impossible to keep track of the exponential increase in the size of the hilbert space as a function of the chain length @xmath74 .
hence only the lowest @xmath298 eigenstates of @xmath292 are retained at the conclusion of each iteration .
@xcite the retained states are used in turn to construct the eigenstates of @xmath299 .
the truncation error involved can be systematically controlled by varying @xmath298 . however , since the size of the hilbert space increases by a factor of @xmath300 with each additional iteration , only a moderately small number of conduction - electron channels can be treated reliably using present - day computers , typically no more than @xmath301 or @xmath302 channels . in the following we focus on @xmath67 and @xmath19 , which is the simplest version of the multichannel kondo hamiltonian that is both amendable to the nrg and predicted to display the phase diagram of fig .
[ fig : fig1](a ) . to cope with the large computational effort involved in exploring the phase diagram of the three - channel kondo model with spin - exchange anisotropy
, we used rather large values of @xmath303 ( either @xmath304 or @xmath305 ) , and applied an alternative truncation scheme to the one customarily used .
since each of the nrg hamiltonians @xmath292 is block diagonal in the conserved quantum numbers , @xcite the computational time is mostly governed by the largest block to be diagonalized .
thus , instead of retaining a fixed number of eigenstates of @xmath292 , at each iteration we adjusted the threshold energy for truncation so that the largest block did not exceed @xmath306 states .
@xcite for @xmath304 we used @xmath307 , while for @xmath305 we set @xmath308 .
this approach gave rise to variations in the number of states retained at the conclusion of each iteration .
typically , 3000 to 4000 states were retained . in some extreme cases down to 2500 states and up to
5500 states were kept .
we have verified that the threshold energies selected in this way were sufficiently high so as not to spoil the accuracy of the calculations .
figure [ fig : fig3 ] depicts the finite - size spectra obtained for @xmath309 and different values of @xmath310 .
here @xmath304 was used . up to a critical coupling @xmath311
, the system always flows to the non - fermi - liquid fixed point of the three - channel kondo effect .
indeed , the fixed - point spectrum is independent of @xmath312 , and is identical to that for isotropic spin - exchange interaction ( left - most panel ) .
there is excellent agreement with the finite - size spectrum of the three - channel kondo effect obtained from conformal field theory , @xcite whose energy levels are marked by arrows in fig .
[ fig : fig3 ] . here
the slight splitting of nrg levels near the dimensionless energy @xmath313 is a @xmath303-dependent feature .
this splitting is reduced upon decreasing @xmath303 , and should disappear for @xmath290 . a different picture is recovered for @xmath314 .
( i ) as demonstrated in fig .
[ fig : fig3 ] for @xmath315 and @xmath316 , the finite - size spectrum remains fermi - liquid - like down to the lowest energies reached ( @xmath317 in some extreme runs ) , without crossing over to the overscreened fixed point of the three - channel kondo effect .
( ii ) the finite - size spectrum varies continuously with @xmath318 , suggesting the flow to a line of fixed points connected by a marginal operator .
( iii ) for @xmath319 , the quantum numbers and degeneracies of the nrg levels are in excellent agreement with those anticipated based on the strong - coupling analysis of sec .
[ sec : sc - s=1/2 ] .
( iv ) the nrg level flow confirms that channel asymmetry is a marginal perturbation for @xmath318 ( not shown ) , in stark contrast to the regime @xmath312 .
these features are all consistent with the flow to a line of stable ferromagnetic - like fixed points as predicted by bosonization . , @xmath67 , @xmath309 , and increasing values of @xmath32 .
here @xmath304 was used . for @xmath320 (
left two panels ) , the system flows to the non - fermi - liquid fixed point of the three - channel kondo effect , irrespective of @xmath7 .
there is excellent agreement with the finite - size spectrum of three - channel kondo effect obtained from conformal field theory , whose energy levels ( in units of the fundamental level spacing ) are indicated by arrows . for @xmath318 ( right two panels ) ,
the finite - size spectrum remains fermi - liquid - like down to minuscule energies , without crossing over to the non - fermi - liquid fixed point of the three - channel kondo effect .
the persisting drift of nrg levels for @xmath318 appears to be due to some weak numerical instability .
a similar drift of levels occurs for @xmath321 ( ferromagnetic coupling).,width=623 ] we do note , however , a persisting drift of the nrg levels for @xmath318 .
a similar drift of levels occurred for @xmath322 ( ferromagnetic coupling ) , and appears to be driven by some weak numerical instability .
while we can not entirely rule out an eventual crossover to the non - fermi - liquid fixed point of the three - channel kondo effect at some lower temperature , we find this scenario highly unlikely given the extremely low energy scales reached and the rapid change of behavior as @xmath7 is swept through @xmath323 .
we therefore identify @xmath324 with the phase boundary between the antiferromagnetic and the new ferromagnetic - like domain predicted by bosonization .
guided by this interpretation , we turned to explore the dependence of @xmath323 on @xmath8 . the resulting phase diagram is plotted in fig .
[ fig : fig4 ] , for @xmath305 . while points on the solid line all converged to the non - fermi - liquid fixed point of the three - channel kondo effect , points on the dashed line showed no such tendency up to @xmath325 nrg iterations ( corresponding to @xmath326 ) .
we therefore estimate the phase boundary between the antiferromagnetic and the ferromagnetic - like domains to lie in between the solid and dashed lines .
in contrast to the solid line , which definitely lies on the antiferromagnetic side of the transition , one can not guarantee that all points on the dashed line lie on the ferromagnetic - like side .
we have confirmed , however , that the phase diagram of fig .
[ fig : fig4 ] is practically unchanged upon going to @xmath327 iterations .
we further stress that the exact position of the phase boundary is in general @xmath303 dependent , although its dependence on @xmath303 appears to be weak . for @xmath304 , for example , the position of the phase boundary varied by no more than just a few percent .
we expect a similar proximity of the @xmath290 phase boundary .
, @xmath67 and @xmath305 .
points on the solid line all converged to the non - fermi - liquid fixed point of the three - channel kondo effect .
points on the dashed line showed no such tendency up to @xmath325 nrg iterations , corresponding to @xmath326 .
we estimate the phase boundary between the two domains to lie in between the solid and dashed lines.,width=283 ] the above results are clearly in good qualitative agreement with the bosonization treatment of sec .
[ sec : mapping ] .
we now turn to a more quantitative comparison . setting @xmath67 in eq .
( [ j - ast ] ) , the critical coupling @xmath41 predicted by bosonization is equal to @xmath328 .
based on the nrg results for @xmath304 and @xmath305 , we estimate the location of the critical coupling for a symmetric box density of states ( i.e. , for @xmath290 ) to be around @xmath329 .
thus , the bosonization and nrg results are within a factor of two from one another . considering that @xmath41 lies well beyond the strict range of validity of bosonization , we find this degree of agreement to be quite remarkable .
as for the shape of the phase boundary , it was predicted in sec .
[ sec : mapping ] to have the linear form @xmath330 , where @xmath331 is given by eq .
( [ c_k ] ) . as seen in fig .
[ fig : fig4 ] , the nrg phase boundary for @xmath305 is well described by a linear curve , at least up to @xmath332 .
the corresponding slope @xmath333 is indeed less than one , but is nearly three - fold larger than the bosonization result , @xmath334 .
this discrepancy can be largely accounted for by plugging the nrg value for @xmath335 into the right - hand side of eq .
( [ c_k ] ) , which yields @xmath336 .
the multichannel kondo hamiltonian is an important paradigm in correlated electron systems , with possible applications to varied systems . depending on the size of the impurity spin , @xmath5 , and the number of independent conduction - electron channels , @xmath4
, it can display either local fermi - liquid , singular fermi - liquid , or non - fermi - liquid behavior .
although the isotropic model is by now well understood , we have shown in this paper that an xxz spin - exchange anisotropy has a far more elaborate effect on its low - energy physics than previously appreciated .
below we briefly summarize our main findings and discuss their implications .
a detailed account of our results is presented in fig .
[ fig : fig1 ] .
we begin with a spin-@xmath0 impurity and with @xmath1 conduction - electron channels . from conformal field theory
it is known that the non - fermi - liquid fixed point of the corresponding kondo model is stable against a small spin - exchange anisotropy .
@xcite however , it was found to be unstable against a sufficiently large @xmath3 . in the latter regime ,
the system flows to a line of stable ferromagnetic - like fixed points with a residual isospin-@xmath0 local moment .
the phase diagram of the model thus consists of three distinct domains : the conventional ferromagnetic and antiferromagnetic ( i.e. , non - fermi - liquid ) domains , plus a second ferromagnetic - like domain located deep in the antiferromagnetic regime .
the new ferromagnetic - like domain extends above a critical longitudinal coupling @xmath41 , whose magnitude depends on @xmath4 . while @xmath41 exceeds the bandwidth for intermediate values of @xmath4 , it is pushed to weak coupling for @xmath65 .
each of the two ferromagnetic - type domains is separated from the antiferromagnetic one by a kosterliz - thouless line , as depicted in fig .
[ fig : fig1](a ) .
while we can not entirely rule out the possibility of yet another domain for sufficiently large @xmath337 , there are no indications at this point in favor of such a scenario .
proceeding to @xmath338 , spin - exchange anisotropy is known @xcite to be a relevant perturbation at the overscreened non - fermi - liquid fixed point for @xmath26 ( for @xmath339 it is a marginal perturbation ) .
the basin of attraction of the overscreened fixed point is therefore confined to the line @xmath340 .
the nature of the low - temperature fixed points for @xmath341 was never explored . as shown in secs .
[ sec : perturbative - rg ] and [ sec : sc - general - s ] , the system flows to a line of stable ferromagnetic - like fixed points with a residual isospin-@xmath0 local moment both for sufficiently small @xmath43 ( i.e. , in the limit @xmath34 for any given ratio @xmath342 ) and for a sufficiently large @xmath3 .
this suggests a single generic behavior in the entire domain @xmath43 .
further support in favor of this interpretation is provided in appendix [ app : s=1 ] , for the special case of a spin - one impurity . as for the domain @xmath107
, here information is confined to the weak - coupling regime , @xmath343 .
depending on the parity of @xmath11 , the system flows either to a conventional fermi liquid with no residual degeneracy ( integer @xmath5 ) , or to a @xmath4-channel kondo effect with an effective spin-@xmath0 local moment ( half - integer @xmath5 ) .
it remains to be seen to what extent is this behavior generic to @xmath107 . for isotropic antiferromagnetic exchange and @xmath1 , the two overscreened spins @xmath19 and @xmath272 share the same low - energy physics .
indeed , the corresponding kondo models are related for @xmath128 through a weak - to - strong - coupling duality .
it is not surprising , then , that the overscreened fixed point shows the same stability for both spins against a weak spin - exchange anisotropy .
@xcite a similar duality appears to hold also for an xxz anisotropy , at least in the range @xmath43 .
indeed , for @xmath19 the overscreened fixed point is stable at weak coupling ( assuming @xmath344 ) , giving way to a line of stable ferromagnetic - like fixed points for sufficiently large @xmath3 . for @xmath272
the roles are reversed .
the overscreened fixed point is stable against a sufficiently large @xmath3 , but is unstable for sufficiently small @xmath43 . in the latter regime ,
the system flows to the same line of stable ferromagnetic - like fixed points that is approached for @xmath19 and a large @xmath7 .
note that , for @xmath272 and small @xmath46 , the stability of the overscreened fixed point depends on the parity of @xmath4 .
the overscreened fixed point is stable for half - integer @xmath5 ( even @xmath4 ) , but is unstable for integer @xmath5 ( odd @xmath4 ) .
all cases discussed above pertain to an overscreened impurity .
we now turn to an underscreened spin , i.e. , @xmath28 . as shown in secs .
[ sec : perturbative - rg ] and [ sec : sc - general - s ] , an xxz anisotropy is a relevant perturbation both near the free - impurity fixed point and for a sufficiently large @xmath49 .
the sole exception to the rule is the case @xmath275 , where @xmath345 is a marginal perturbation in each of these limits .
consider first the range @xmath43 . here
the flow in both extremes is to the same line of ferromagnetic - like fixed points with a residual isospin-@xmath0 local moment .
as before , this suggests a single generic behavior throughout the domain @xmath43 .
similar to the case of isotropic antiferromagnetic exchange , the low - energy physics is comprised of quasiparticle excitations plus a residual local moment . however , the residual degeneracy for @xmath283 is smaller than that of the isotropic underscreened fixed point ( two versus @xmath346 ) , distinguishing the ferromagnetic - like line of fixed points from the isotropic underscreened one . of the different possible cases , the most intriguing
perhaps is that of an underscreened impurity with @xmath107 and half - integer @xmath5 . as we have shown in sec .
[ sec : perturbative - rg ] , the resulting low - energy physics is that of a @xmath4-channel , spin-@xmath0 kondo effect , at least in the limit of sufficiently weak coupling
. thus , an xxz anisotropy drives the system from underscreened to overscreened behavior .
a more complete characterization of the transition between these two distinctly different behaviors is clearly needed .
going back to @xmath19 , we conclude with a few further comments on the mapping of eq .
( [ mapping ] ) .
we first reiterate that @xmath41 lies well beyond the strict range of validity of bosonization for intermediate values of @xmath4 .
nevertheless , this approach ( and its anderson - yuval equivalent @xcite )
well describes the new ferromagnetic - like phase both qualitatively and quantitatively .
for @xmath67 , which features the largest @xmath41 and is thus the most prone to error , eq .
( [ j - ast ] ) is only a factor of two larger than the nrg estimate for a symmetric box density of states .
we find this degree of quantitative agreement to be quite remarkable . although the mapping of eq .
( [ mapping ] ) was derived in sec .
[ sec : mapping ] for a channel - isotropic model , it can easily be extended to channel anisotropy both in the spin - flip coupling , @xmath347 , and in the longitudinal coupling , @xmath348 .
the individual transverse couplings remain unchanged in the course of the mapping , whether isotropic or not . as for the individual longitudinal couplings ,
these transform according to @xmath349 where @xmath350 is the average phase shift for the @xmath4 different channels .
accordingly , the mapping of eqs .
( [ anisotropic_mapping-1 ] ) and ( [ anisotropic_mapping-2 ] ) is restricted to values of @xmath351 where the modulus of the right - hand side of eq .
( [ anisotropic_mapping-1 ] ) does not exceed @xmath58 for any of the @xmath4 channels .
observe that channel anisotropy is preserved by eqs .
( [ anisotropic_mapping-1 ] ) and ( [ anisotropic_mapping-2 ] ) , which adequately reduce to eq .
( [ mapping ] ) in the limit of isotropic couplings . finally
, we remark on the possibility of generalizing the mapping of eq .
( [ mapping ] ) to arbitrary spin @xmath5 .
two modifications appear when the same sequence of steps is applied to an impurity spin larger than one - half : ( i ) for @xmath352 , the phase shift @xmath156 in the absence of @xmath8 is a nonlinear function of @xmath52 . hence , the bosonized form of the @xmath7 interaction term is no longer linear in @xmath52 as for @xmath19 .
( ii ) the unitary transformation @xmath353 produces an additional hamiltonian term of the form @xmath354 , similar to the one generated in perturbative rg [ see eq . (
[ d - delta - dl ] ) ] . for @xmath19 ,
this term amounts to a uniform shift of the entire spectrum , which can be safely ignored .
this , however , is no longer the case for @xmath9 , where different kramers doublets are split . as a result of the former modification ,
the mapped hamiltonian no longer assumes the form of a simple spin - exchange hamiltonian for @xmath352 .
the case @xmath355 is an exception in this regard .
the mapped hamiltonian does acquire an additional @xmath354 term , but otherwise retains the form of a conventional spin - exchange interaction .
a detailed discussion of this particular case is presented in appendix [ app : s=1 ] .
stimulating discussions with natan andrei , piers coleman , andres jerez , eran lebanon , pankaj mehta , and gergely zarnd are gratefully acknowledged .
a.s was supported in part by the centers of excellence program of the israel science foundation , founded by the israel academy of science and humanities .
in this appendix , we extend the mapping of eq .
( [ mapping ] ) to a spin - one impurity .
as explained in the main text , @xmath355 is the only other spin size for which a simple spin - exchange interaction is restored at the conclusion of the mapping .
however , an additional local field proportional to @xmath50 is generated .
the basic steps of the derivation are nearly identical to those carried out in sec .
[ sec : mapping ] for @xmath19 .
only a few minor modifications appear , as specified below .
our starting point is the hamiltonian of eq .
( [ h ] ) , where @xmath73 now represents a spin - one operator . bosonizing the fermion fields according to eq .
( [ psi - via - phi ] ) , the bosonic hamiltonian assumes the form of eq .
( [ h - bosonized ] ) with one sole variation : the longitudinal spin - exchange term now reads @xmath356 s_z , \ ] ] where @xmath357 converting to the boson fields of eqs .
( [ phi_s - def ] ) and ( [ phi_mu - def ] ) and applying the transformation @xmath173 $ ] , the transformed hamiltonian @xmath172 retains the same overall form as eq .
( [ h ] ) , but with two important modifications : ( i ) the coefficient of the @xmath177 term is replaced with @xmath358 ( ii ) a new hamiltonian term @xmath359 with @xmath360\ ] ] is added to @xmath175 .
@xcite proceeding with the transformation @xmath178 and converting back to a fermionic representation , the hamiltonian @xmath175 regains the form of eq .
( [ h ] ) , but with certain renormalized parameters : @xmath361 @xmath186 @xmath362 and @xmath363 .\ ] ] here @xmath233 is determined from the equation @xmath364 which comes in place of eq .
( [ mapping ] ) for a spin-@xmath0 impurity . and
different values of @xmath65 . in the region below and to the right of each solid line , both @xmath365 and @xmath366 .
hence the system flows to the line of stable ferromagnetic - like fixed points with a residual degeneracy of two .
the fixed - point structure in the remaining portion of the domain @xmath43 , i.e. , in between the solid lines and the overscreened line @xmath367 ( marked by a dashed line ) , can not be deduced based on eqs .
( [ mapping(a)-s=1])([mapping(b)-s=1 ] ) alone.,width=302 ] as for @xmath19 , the mapping of eqs .
( [ mapping(a)-s=1])([mapping(b)-s=1 ] ) is restricted to values of @xmath7 where the right - hand side of eq .
( [ mapping(b)-s=1 ] ) does not exceed @xmath58 .
this constrains the mapping to @xmath368 ( overscreened impurity ) , and to @xmath369 with @xmath370 specifically , for @xmath26 the coupling regimes @xmath62 and @xmath371 with @xmath372 are mapped onto one another .
this should be compared with eqs .
( [ j_min ] ) and ( [ j - ast ] ) for @xmath19 . since eqs .
( [ mapping(a)-s=1])([mapping(b)-s=1 ] ) define yet another multichannel kondo hamiltonian with both spin - exchange anisotropy and a potentially competing @xmath354 term , we have no conclusive way to deduce its fixed - point structure throughout the @xmath7@xmath8 plane .
nevertheless , the behavior in one particular region is clear .
for values of @xmath7 where both @xmath373 and @xmath365 , the spin - exchange interaction and the @xmath354 term conspire to favor a ferromagnetic - like state with two - fold residual degeneracy .
here the residual degeneracy originates from the @xmath374 kramers doublet favored by @xmath365 . for such values of @xmath7
, one can safely conclude that the system flows to the line of stable ferromagnetic - like fixed points identified previously from the strong - coupling expansion of sec .
[ sec : sc - general - s ] .
these considerations provide us with the following lower bound on the basin of attraction of the ferromagnetic - like line of fixed points : @xmath375 @xmath376 .
\label{boundary - ii}\ ] ] figure [ fig : fig5 ] depicts that portion of the @xmath7@xmath8 plane defined by eqs .
( [ boundary - i ] ) and ( [ boundary - ii ] ) , for several values of @xmath65 . with increasing @xmath4 , a growing fraction of the domain @xmath43 is covered by this region , which stretches for @xmath65 from @xmath377 and upward in @xmath7 .
this result partially bridges between the strong- and weak - coupling limits ( @xmath378 and @xmath379 , respectively ) , where flow to a ferromagnetic - like state has been established in secs .
[ sec : perturbative - rg ] and [ sec : sc - general - s ] using vastly different techniques . as for the remaining fraction of the domain @xmath43 , its fixed - point structure can not be immediately deduced based on eqs .
( [ mapping(a)-s=1])([mapping(b)-s=1 ] ) alone .
99 d. l. cox , phys .
rev . lett .
* 59 * , 1240 ( 1987 ) . c. l. seaman , m. b. maple , b. w. lee , s. ghamaty , m. s. torikachvili , j .- s .
kang , l. z. liu , j. w. allen , and d. l. cox , phys . rev
. lett . *
67 * , 2882 ( 1991 ) .
d. c. ralph and r. a. buhrman , phys .
* 69 * , 2118 ( 1992 ) .
d. c. ralph , a. w. w. ludwig , j. von delft and r. a. buhrman , phys .
lett . * 72 * , 1064 ( 1994 ) .
a. zawadowski , phys .
lett . , 211 ( 1980 ) .
k. vladar and a. zawadowski , phys .
b * 28 * , 1564 ( 1983 ) ; * 28 * 1582 ( 1983 ) ; * 28 * 1596 ( 1983 ) .
t. cichorek , a. sanchez , p. gegenwart , f. weickert , a. wojakowski , z. henkie , g. auffermann , s. paschen , r. kniep , and f. steglich , phys .
lett . * 94 * , 236603 ( 2005 )
. this paper reports two - channel kondo characteristics in transport properties of thasse , presumably due to scattering off structural two - level systems .
k. a. matveev , zh . eksp .
fiz . * 99 * , 1598 ( 1991 ) [ sov .
jetp * 72 * , 892 ( 1991 ) ] .
d. berman , n. b. zhitenev , r. c. ashoori , and m. shayegan , phys .
lett . * 82 * , 161 ( 1999 ) . for a comprehensive review of the multichannel kondo effect ,
see d. l. cox and a. zawadovski , adv .
* 47 * , 599 ( 1998 ) .
a. posazhennikova and p. coleman , phys .
lett . * 94 * , 036802 ( 2005 ) .
p. mehta , n. andrei , p. coleman , l. borda , and g. zarnd , phys .
b * 72 * , 014430 ( 2005 ) .
p. coleman and c. ppin , phys .
b * 68 * , 220405(r ) ( 2003 ) .
w. koller , a. c. hewson , and d. meyer , phys .
b * 72 * , 045117 ( 2005 ) .
i. affleck , a. w. w. ludwig , h .- b .
pang , and d. l. cox , phys .
b * 45 * , 7918 ( 1992 ) .
k. g. wilson , rev .
phys . * 47 * , 773 ( 1975 ) .
m. fabrizio , a. o. gogolin , and p. nozires , phys .
b * 51 * , 16088 ( 1995 ) .
j. ye , phys .
* 77 * , 3224 ( 1996 ) .
w. anderson , j. phys .
c * 3 * , 2436 ( 1970 ) .
f. d. m. haldane , j. phys .
c * 14 * , 2585 ( 1981 ) .
p. nozires and a. blandin , j. physique * 41 * , 193 ( 1980 ) .
see eqs .
( 2 ) of ref . with @xmath380 .
careless truncation may result in spurious symmetry breaking . to avoid symmetry breaking in the course of truncation it is essential to accommodate the degenerate multiplets of @xmath292 in their entirety .
this may require a slight increase in the number of states retained at certain iterations .
in our code we exploited the conservation of the following quantities : the total electronic charge , the @xmath53 component of the overall spin of the system , and the operator @xmath381 $ ] .
we did not make use of the full su(3 ) channel symmetry of the model .
i. affleck , a. w. w. ludwig , nucl .
phys . , 641 ( 1991 ) .
@xmath382 is formally proportional to @xmath383 , which is regularized as @xmath384 . | the phase diagram of the multichannel kondo hamiltonian with an xxz spin - exchange anisotropy is revisited , revealing a far richer fixed - point structure than previously appreciated . for a spin-@xmath0 impurity and @xmath1 conduction - electron channels ,
a second ferromagnetic - like domain is found deep in the antiferromagnetic regime .
the new domain extends above a ( typically large ) critical longitudinal coupling @xmath2 , and is separated from the antiferromagnetic domain by a second kosterliz - thouless line . a similar line of stable ferromagnetic - like fixed points with a residual isospin-@xmath0 local moment is shown to exist for large @xmath3 and arbitrary @xmath4 and @xmath5 obeying @xmath6 .
here @xmath7 is the longitudinal spin - exchange coupling , @xmath8 is the transverse coupling , and @xmath5 is the impurity spin . near the free - impurity fixed - point ,
spin - exchange anisotropy is a relevant perturbation for @xmath9 and arbitrary @xmath4 .
depending on the sign of @xmath10 and the parity of @xmath11 , the system flows either to a conventional fermi liquid with no residual degeneracy , or to a @xmath4-channel , spin-@xmath0 kondo effect , or to a line of ferromagnetic - like fixed points with a residual isospin-@xmath0 local moment .
these results are obtained through a combination of perturbative renormalization - group techniques , abelian bosonization , a strong - coupling expansion in @xmath12 , and explicit numerical renormalization - group calculations . |
one - dimensional ( 1d ) systems @xcite are a model to study the fundamental processes of dynamics and ( de)coherence in interacting many - body quantum systems .
ultracold atoms in strongly elongated traps with @xmath5 ( @xmath6 , @xmath7 being the frequencies of the radial and longitudinal confinement , respectively ) offer the possibility to implement 1d quantum physics if both the temperature @xmath8 and chemical potential @xmath9 are small compared to the energy scale given by the transverse confinement : @xmath10 1d systems of ultra - cold atoms were implemented in both optical lattices @xcite and atom chips @xcite . in the limit of zero temperature
they are a realization of the lieb - liniger model @xcite of spinless bosons with contact ( point - like ) interaction , a prime example of an integrable system .
an important parameter characterizing an 1d system of bosons with point - like interactions described by the ( three dimensional ) @xmath11-wave scattering length @xmath12 is the lieb - liniger parameter @xcite @xmath13 where @xmath14 is the mass ot the bosonic atom , @xmath15 the linear density of the atoms in the 1d trap characterized by the transverse confinement frequency @xmath6 , and @xmath16 is the fundamental length scale of the localization of an atom in the transversal direction given by @xmath17 the limit @xmath18 corresponds to a weakly - interacting regime , whereas @xmath4 signifies strongly - interacting , strongly correlated ( tonks - girardeau ) regime . in an integrable system
@xcite the number of their integrals of motion equals exactly the number of their degrees of freedom .
thus such a system always `` remembers '' its initial state in the course of its dynamical evolution , and thermalization does not occur . in an integrable system the finite spread of initial energy may lead only to relaxation towards the generalized gibbs ( or fully constrained thermodynamic ) ensemble @xcite . strictly speaking ,
there is no thermalization in any _ closed _ system , but for non - integrable systems the eigenstate thermalization hypothesis @xcite holds , enabling _ dephasing _ to mimic the relaxation to the thermal equilibrium .
recently strong inhibition of thermalization , was observed on an optical lattice experiment with bosons deep in the 1d regime @xcite . on the other hand interference experiments on atom chips with pairs of weakly interacting bose gases fulfilling the conditions of eq .
( [ eq:1 ] ) are in a good agreement with the thermal - equilibrium description of the 1d atomic ensembles @xcite . in the present paper
we investigate the collisional properties of bose gases in a 1d geometry and how they contribute to thermalization .
we follow thereby a procedure outlined in our two previous works @xcite , and give a more in - depth description of the underlying theoretical considerations .
we start with the calcualtion of the freeze - out of thermalization providing two body collisions .
we then proceed to show that virtual excitations of excited states lead to effective three - body collisions , which lead to a term in the hamiltonian that breaks integrability and enables thermalization .
we then proced to estimte the effects of quantum correlations in 1d and show how they suppress the three - body term for strongly correlated 1d systems , thereby extending the time scale , on which a quasi-1d system can be _ approximately _ described as integrable . in our theoretical considerations we consider identical bosons in a tight 1d wave guide with radial confinement given by a 2d harmonic oscillator with frequency @xmath6 ( we set @xmath19 ) .
the contact interaction is described by the pseudopotential @xmath20 .
the hamiltonian of the 1d system is : @xmath21 , \label{eq:2 } \\
\hat{h}^{(r)}&=&-\frac { \hbar ^2}{2m}\left ( \frac { \partial ^2}{\partial x^2}+\frac { \partial ^2}{\partial y^2 } \right ) + \frac{m\omega _ r^2}2 ( x^2+y^2 ) .
\label{eq:3}\end{aligned}\ ] ] thereby the field operators @xmath22 are assumed to vanish for @xmath23 and to be periodic along @xmath24 with the period @xmath25 .
for the solutions to eq .
[ eq:3 ] we make the _ ansatz _ : @xmath26 where @xmath25 is the quantization length and the atomic annihilation and creation operators @xmath27 and @xmath28 obey the standard bosonic commutation rules .
@xmath29 is the normalized eigenfunction of both the radial confinement hamiltonian , @xmath30 and the @xmath24-projection of the orbital momentum , @xmath31 \varphi _ { n,\ell } ( x , y)=\ell \varphi _ { n,\ell } ( x , y).\ ] ] because we consider identical bosons the main quantum number @xmath32 , and the @xmath24-projection quantum number @xmath33 of the orbital - momentum is restricted by @xmath34 and thus has the same parity as the main quantum number .
we first look at collisions of two identical bosonic atoms that are initially in the transverse ground state of the radial confinement .
if the collision is restricted to 1d , that is both atoms remain after the collision in the transverse ground state , then there can be no energy exchange and cinsequently no thermalization .
for such two - body collisions to contribute to energy exchange and thermalization , they have to lead to a change in transverse excitation . by symmetry @xmath35
must be even . for atoms in the transverse ground state ( @xmath36 ) @xmath37 band following the above considerations their orbital - momentum quantum numbers after collision
are restricted to @xmath38 and @xmath39 .
the rate of populating the radially excited modes by pairwise atomic collisions @xmath40 , can then be estimated for a non - degenerate bose gas , using fermi s golden rule . for @xmath41
this rate is @xmath42 the dimensionless quantity @xmath43 combines two dimensionless parameters characterizing a 1d system .
@xmath44 is a measure how much the interaction energy ( the chemical potential @xmath9 ) is smaller than the energy scale given by the transverse confinement .
@xmath45 characterizes the relation between the transverse confinement and the strength of the contact interaction .
its importance can be seen when looking at how the effective 1d coupling constant @xmath46 of pairwise interacting bosonic atoms in a waveguide changes with confinement due to virtual excitation of the radial modes .
following olshanii @xcite @xmath47 $ ] , @xmath48 and increases as @xmath45 grows .
this points to the ratio @xmath49 as the measure how much the 1d approximation is violated . in a general case
we have to change in all the following expressions @xmath12 to @xmath50 $ ] .
( [ eq:20 ] ) has a also transparent physical interpretation : the rate @xmath40 is related to the 3d atomic density ( @xmath51 ) , times the @xmath11-wave scattering cross - section ( @xmath52 ) , times the exponential boltzmann factor for the fraction of atoms fast enough to scatter into higher radial modes , times the corresponding velocity of the collision ( @xmath53 ) .
looking at the scaling of eq .
( [ eq:20 ] ) one immedeately sees that the collision rate leading to thermalization ( @xmath40 ) rapidly diminishes when the temperature approaches @xmath54 and is suppressed by more then a factor of 50 ( @xmath55 ) for @xmath56 . estimating the numbers for recent experients @xcite in @xmath57rb : @xmath58 , @xmath59m@xmath60 , @xmath61 khz , @xmath62 nk ( @xmath63 ) one obtains a collision rate of @xmath64 .
ths is at least one order of magnitude too small for two body collisions to be responsible for the thermalization required in the evaporative cooling process leading to these low temperatures .
if the kinetic energy of the collision is less than @xmath65 , then the radial modes can be excited only virtually .
such processes contribute to the system dynamics in the second and higher orders of perturbation theory .
the simplest case is when after the collision the radial motion state is @xmath66 .
then only one more collision is enough to de - excite the radial mode and bring the system back on the energy shell [ see fig . [ fig:1](a ) ] .
such a process yields an effective three - body collision already in the _ second _ order of perturbation theory .
in contrast processes involving a virtual excitation to @xmath67 , @xmath68 , shown in fig .
[ fig:1](b ) , contribute only in the _ third _ order , and thus will be neglected .
we will now calculate the matrix elements for the process shown in figure [ fig:1](a ) which leads to effective three - body collisions . in our perturbation calculation
the small parameter is @xmath69 , that is the mean field interaction @xmath9 is much smaller then the energy scale @xmath70 connected to the transverse confinement . in addition , to avoid complications related to the confinement - induced resonance in 1d scattering @xcite we assume @xmath71 .
we can then rewrite the hamiltonian ( [ eq:2 ] ) as @xmath72 where all the terms irrelevant to the process under discussion ( fig .
1a ) are gathered in @xmath73 , and the radial matrix element @xmath74 is given by : @xmath75 it connects to two atoms in the ground state of the incoming channel , to one atom remaining in the same state , and the other being excited to a state with zero orbital - momentum quantum number and even main quantum number @xmath76 , @xmath77 ( remember for boson @xmath78 and @xmath33 are required to have the same parity ) .
to evaluate eq .
( [ eq:5 ] ) , we recall that the normalized radial wave functions @xmath29 of interest are real and can be expressed through laguerre polynomials @xmath79 @xmath80 since @xmath81 , we obtain @xcite @xmath82 independently of @xmath83 .
then we easily obtain the necessary matrix element as @xmath84 , where the coefficient @xmath85 is defined by the expansion @xmath86 comparing the coefficients in front of @xmath87 in the left- and right - hand sides of eq .
( [ pert4 ] ) , we obtain @xmath88 and @xmath89 in our consideration we are only interested in the case where the collision energy of the two atoms is always much smaller than @xmath90 . then , using the matrix element ( [ pert5 ] ) and adiabatically eliminating the radially excited mode operators , we obtain from the original hamiltonian eq .
( [ eq:2 ] ) an effective 1d hamiltonian : @xmath91 where we write for simplicity @xmath92 and the numerical constant @xmath93 is given by @xmath94 note that the relative contribution of the virtual states with the excitation energy higher than @xmath95 given by @xmath96 is remarkably small .
the summation in the last term of eq .
( [ eq:11a ] ) is taken over all the kinetic momenta obeying the conservation law @xmath97 introducing the field operator @xmath98 , we can rewrite eq .
( [ eq:11a ] ) as @xmath99 the first and second terms in eqs .
( [ eq:11a ] ) or ( [ eq:11b ] ) correspond to the lieb - liniger model .
the third ( _ cubic _ ) term stems from the effective three - body collisions mediated by virtually excited states .
this third ( _ cubic _ ) term in eq .
[ eq:11b ] violates the integrability in the 1d system .
the fact that the effective three - body interactions are dominated by virtual excitations of the lowest even - parity excited state may seem surprising , since the correct calculation of the effective _ two - body _
1d coupling constant requires taking into account of the infinite number of states @xcite .
however , in the latter case one deals with the calculation of the two - body wave function , which has in 1d a @xmath100 singularity ( stemming from the 3d boundary condition at @xmath101 that provides the correct asymptotic form of the scattered @xmath11-wave ) .
the removal of this singlularity yields the regular part of the two - body wave function and , through this regular part , the scattering amplitude and , hence , the effective coupling in 1d . on the other hand ,
if one tries to obtain the effective 1d interaction by adiabatic elimination of _ all _ the excited states , one gets a divergent series @xmath102 in the expression for the effective 1d coupling constant .
the aforementioned regularization of the wave functions formally corresponds to the renormalization of this divergent series via substituting it by a finite expression @xmath103 @xcite . in our case , processes related to _ three - body _ collisions do not give rise to additional singularities in the many - body wave function , and no additional regularization is needed .
the sum over all excited states thus converges .
the convergence is rapid enough to ensure fair estimation of the whole sum by its first term
. accurate calculation of the effective three - body interaction potential @xmath104 yields @xmath105 , \label{u3biiia}\ ] ] where @xmath106\sum _ { p=1}^\infty \frac { \sqrt{p}}{\sqrt{2}\ , 4^{p-1 } p\ , l_r}\exp \left ( - \frac{\sqrt{2 p } \ , |x_1-x_2|}{l_r } \right ) .
\label{yiiia}\ ] ] obviously , the sum in eq .
( [ yiiia ] ) converges and gives a sharp - peaked function rapidly ( exponentially ) decreasing at distances much larger than @xmath16 .
( [ eq:1 ] ) holds , we consider scattering events with transferred momenta much less than @xmath107 . in this case
we can use approximation @xmath108\delta ( z_1-z_2)$ ] . then , by taking the matrix element of the effective interaction @xmath104 and dividing it by 3 ! ( the number of permutation of three identical particles ) we obtain the last term in the second - quantized hamiltonian ( [ eq:10 ] ) . before
continuing we want to point out similarities with other recent works : ( 1 ) the mechanism discussed here is to a certain extent similar to the virtual association of atoms to a molecular dimer @xcite . in our discussion here ,
virtual excitation of radial modes during a two - atom collision temporarily localize the interatomic distance on the length scale @xmath109 . scattering a third atom on such a transient structure of finite size and
mass @xmath110 leads to an effective threebody collision . in ref
@xcite , collisions of a third atom bring `` virtual '' dimers , enhanced in size by a feshbach resonance down to the energy shell , thus bringing about `` quantum chemistry '' in 1d .
( 2 ) in a similar way effective three - body interactions between polar molecules emerge , due to virtual transitions to an off - resonant internal state @xcite .
the cubic term in eq . ( [ eq:11b ] ) is negative and thus supports no bound ground state .
therefore we have to consider it as a first correction term to the purely pairwise interaction energy in the effective 1d hamiltonian .
considering the mean - field limit of eq .
( [ eq:11b ] ) , @xmath111 , we get the energy density ( per unit length ) @xmath112 + \hbar \omega _
r \alpha _ s n_{1d}^2 - \frac \xi 2 \hbar \omega _
s^2 n_{1d}^3 , \label{eq : var1}\ ] ] this expansion is correct in the limit of the small linear density @xmath113 .
if we neglect the contribution of the radial levels with the main quantum number larger than 2 by setting @xmath114 , we see that eq .
( [ eq : var1 ] ) is the expansion up to the cubic term of the energy density obtained by the variational method by salasnich , parola and reatto @xcite @xmath115+n_{1d}\left ( \frac { \hbar ^2}{2m\sigma ^2}+ \frac 12 m\omega _
r^2 \sigma ^2\right ) + \frac { \hbar ^2\alpha _ sn_{1d}^2 } { m\sigma ^2}\nonumber \\ & = & \frac { \hbar^2}{2 m } \left [ \left(\frac { \partial \sqrt{n_\mathrm{1d } } } { \partial z } \right)^2 + n_\mathrm{1d}\left(\frac { \partial \theta } { \partial z } \right)^2\right ] + n_{1d}\hbar \omega _ r \sqrt{1 + 2n_{1d } \alpha _ s } .
\label{eq : var2}\end{aligned}\ ] ] here one assumes the wave function of the transversal atomic motion to be @xmath116 $ ] , with @xmath117 which minimizes @xmath118 .
we now turn to the collision rates for these effective three body collisions .
we start with calculating the rate @xmath119 for the decay of a specific state @xmath120 , @xmath121 being the vacuum state of the atomic field wherein atoms are absent ( should not be confused with the _ vacuum of elementary excitations _ ) due to three - body collisions .
the final states of the decay are @xmath122 . to make the calcualtion simple we assume the 1d bosonic gas being strongly non - degenerate ( @xmath123 much higher than the chemical potential ) and weakly - interacting .
this enables us to neglect the probability double ( and higher ) occupation of any @xmath124-mode , therefore assuming all involved atomic momenta to be different , and assume the elementary excitations coinciding with the atomic plane waves with the free - particle ( quadratic ) dispersion law .
then fermi s golden rule yields @xmath125 \left ( \frac { \xi \hbar \omega _ r\alpha _ s^2}{2l^2}\right ) ^2 \times \nonumber \\ & & \left .
| \langle k_1^\prime , k_2^\prime , k_3^\prime | \sum _ { \{q_j^\prime \ } } \hat{a}_{q_1}^\dag \hat{a}_{q_2 } ^\dag \hat{a}_{q_3}^\dag \hat{a}_{q_1}\hat{a}_{q_2}\hat{a}_{q_3}|k_1,k_2,k_3\rangle | ^2 \right| _ { k_1^\prime + k_2^\prime + k_3^\prime = k_1+k_2+k_3 } .
\label{eq : rates1 } \end{aligned}\ ] ] to account for the condition @xmath126 we add an additional integration over @xmath127 with the necessary delta - function : @xmath128 \delta \left ( \sum _ j^3k^{\prime } _ j -\sum _
j^3k_j \right ) \times \nonumber \\ & & \left ( \frac { \xi \hbar \omega _ r\alpha _ s^2}{2l^2}\right ) ^2
\left| \langle k_1^\prime , k_2^\prime , k_3^\prime | \sum _ { \{q_j^\prime \ } } \hat{a}_{q_1}^\dag \hat{a}_{q_2 } ^\dag \hat{a}_{q_3}^\dag \hat{a}_{q_1}\hat{a}_{q_2}\hat{a}_{q_3}|k_1,k_2,k_3\rangle \right| ^2 . \label{eq : rates2 } \end{aligned}\ ] ] since the atoms are indistinguishable , we need to integrate over the volume @xmath129 in the @xmath130-space that corresponds to a unique ordering of the variables , e.g. , @xmath131 .
the integral over @xmath129 of a function fully symmetric over permutations of @xmath132 amounts to @xmath133 of the integral over the whole @xmath130-space .
if all the involved momenta are different ( as is the case for a system far from degeneracy ) all possible ways of ordering three bosonic creation operators with lower indices @xmath134 and three bosonic annihilation operators with lower indices @xmath135 finally yield @xmath136 to evaluate the integral in eq .
( [ eq : rates3 ] ) , we perform an orthogonal transformation from @xmath132 to jacobi co - ordinates ( here we deal with the 1d analog of the hyperspherical co - ordinates , which are used in the three - body problem @xcite ) in the wavenumber space : @xmath137 further we introduce the hyperangle @xmath138 via @xmath139 and express eq .
( [ eq : rates3 ] ) as @xmath140 and accurate evaluation of the integral yields ( cf .
@xcite ) @xmath141 to calculate the three - body collision rate @xmath142 per atom , we need to multiply @xmath119 by the product of populations @xmath143 of the states @xmath144 , @xmath145 ( the occupation probabilities are normalized to unity , @xmath146 ) and integrate the whole @xmath135-space , divide by 3 ! to take into account the indistinguishability of the bosons .
then we obtain the number of three - body collisions per unit time in the whole system .
this number should be divided by @xmath147 to obtain the rate per atom : @xmath148 the result of eq .
( [ eq:17b ] ) may seem counterintuitive at first : the collision rate is independent of temperature , and it is proportional to the dimensional parameter @xmath149 and the radial confinement @xmath150 .
the physics behind the first observation is related to the fact that the collision kinetic energy is small compared to the virtual excitation energy .
this was one of our assumptions in deriving the effective three body collisions and is required by the condition ( @xmath151 ) to be fulfilled when building a 1d system ( eq .
( [ eq:1 ] ) ) .
consequently the composite matrix element of the second - order process should not depend in leading order on the velocities ( energies ) of the colliding particles and hence on temperature ( see eq .
( [ eq:5 ] ) ) .
in addition the phase space volume for the scattered particles is independent on the incoming momenta @xmath152 , @xmath153 , and @xmath154 .
the other terms can be motivated the following basic physics considerations : since effective three - body elastic scattering is the dominant process the scattering rate must be proportional to the 3d density squared : @xmath155 .
furthermore , the scattering rate contains the square of the matrix element corresponding to the diagram in fig .
1(a ) , where each vertex is proportional to @xmath156 , therefore this rate is proportional to @xmath157 .
the factor @xmath158 provides the correct dimensionality ( s@xmath60 ) .
we can now compare the scattering rates for the two - body collisions @xmath159 ( eq . ( [ eq:20 ] ) ) or effective three - body collisions @xmath160 ( eq . ( [ eq:17b ] ) ) and
evaluate their contributions to thermalization and the breakdown of integrability in 1d systems .
for @xmath41 we find a simple scaling : @xmath161 the relative importance of two - body collisions and the effective three - body collions mediated by _ virtual _ excitations is determined by the dimensionless quantity @xmath162 .
for large @xmath163 and small temperatures ( @xmath164 ) the three - body scattering rate due to _ virtual _ excitations dominates , and can lead to thermalization even when the thermalization processes due to two - body collisions are frozen out .
for example , in typical atom chip experiment @xcite the three body reate @xmath160 dominates above the two - body collisions at @xmath165 . a detailed comparison of the two rates @xmath159 and @xmath160 and their relation to typical experimental parameters
is given in fig .
[ fig : scatrat ] . the scattering rate due to _ virtual _ excitations of the radial modes can dominate over real excitations for typical parameters of the recent experiment @xcite . the above calculation was for a non - degenerate ultracold gas . in a degenerate gas
we need to consider the bogoliubov - type spectrum of elementary excitations @xcite , that is phononic in the long - wavelength regime , as well as the relation between atoms and elementary excitations via the bogoliubov transformations and bosonic amplification of scattering to modes , which are initially occupied . taking into account all these factors
, we find the rate of damping of a fast particle in a quasicondensate ( see also ref .
@xcite ) : @xmath166 here the momenta of the scattered elementary excitations are defined by the expressions reciprocal to eq .
( [ eq : rates4 ] , [ eq : rates4bis ] ) : @xmath167 the energy of a mode with the momentum @xmath168 is @xmath169 with @xmath170 the static structure factor of a quasicondensate is @xmath171 in equilibrium the population of the elementary mode with the momentum @xmath168 is given by the bose - einstein statistics with the mean occupation number for the mode with the momentum @xmath168 @xmath172 - 1 } .\ ] ] the initial kinetic energy @xmath173 of the fast atom is assumed here to be large compared to both the mean - field interaction energy per particle in the quasicondensate and the temperature : @xmath174 under condition ( [ eq : rates8 ] ) one of the scattered particles is always fast , and one of the three structure factors appearing in eq .
( [ eq : rates6 ] ) is always very close to 1 ( and the corresponding occupation number is close to 0 ) . in the most scattering events the other two particles are also fast , and ,
hence , @xmath175 and @xmath176 for all three particles , @xmath145 . only for the scattering events with small transferred momentum two of the structure factors
are significantly less than 1 and/or the corresponding populations approach the high - temperature limit @xmath177 . however , in the practically interesting case where @xmath178 the contribution of scattering events with small transferred momentum is relatively small , and @xmath179 the result of eq .
( [ fast - qc ] ) for a fast atom in a quasicondensate is also obtained by tan , pustilnik and glazman @xcite .
we now turn to quantify thermalization in tightly confined bosons in a quasi-1d geometry by both two body collisions and the effective interaction ( [ eq:11a ] ) .
we again for simplicity consider a non - degenerate , weakly - interacting ( the lieb - liniger parameter @xcite @xmath180 being much less than 1 ) gas of bosonic atoms .
the assumptions of weak interaction and non - degeneracy enable us to express the three - particle distribution function through the product of single - particle distribution functions @xmath181 .
in contrast , calculation of relaxation via three - body collisions of low - energy excited states in the stronger interacting regime and especially for @xmath182 , requires to consider the ( strong ) quantum correlations in the quasi-1d bosonic system .
we will discuss the effects of correlations on scattering rate @xmath142 and on thermalization in section [ sec : correlations ] we start be writing the boltzmann equation with a three - body collision integral @xcite , taking into account the indistinguishability of the particles : @xmath183 with @xmath184 eq .
( [ eq:12 ] ) can be easily understood : after integration over @xmath130 and @xmath185 the loss term in eq .
( [ eq:12 ] ) is simply @xmath186 , which is the elastic three - body collision rate per atom . on the other hand
, the three - atom state @xmath187 is populated by elastic three - body collisions from those states @xmath188 which have the same center - of - mass momentum : @xmath189 . since the kinetic energy of the relative motion ,
@xmath190 is conserved , the states @xmath188 ( from where the state @xmath191 can be populated from ) can be fully parametrized by the hyperangle @xmath192 .
we now use the following _ ansatz _ for the perturbed momentum distribution @xmath193 , \label{eq:16}\end{aligned}\ ] ] to solve eq .
( [ eq:12 ] ) .
therby @xmath194 and @xmath195 is the hermite polynomial of the 4th order .
ansatz _ is the simplest nontrivial perturbation that retains @xmath196 .
we then proceed to linearizing eq .
( [ eq:12 ] ) with respect to the perturbation amplitude @xmath197 and obtain an exponential solution @xmath198}^{3b}t)$ ] with @xmath199}^{3b}&=&c_{[4]}\frac { \hbar}{m } \left ( \frac { n_{1d}}{l_r ^2 } \right)^2 \alpha _ s ^4 = c_{[4 ] } \ , \omega_r \ , \zeta^2 \label{eq:17}\end{aligned}\ ] ] with the numerical constant @xmath200}=\frac 8{27}c_{3b } = \frac { 8\xi ^2}{3\sqrt{3}}\approx 2.04 .
\label{eq:170}\ ] ] to estimate the validity of our _ ansatz _ we note that using a higher - order hermite polynomial @xmath201 in eq .
( [ eq:16 ] ) leaving the functional dependence on the parameters of the system unchanged and leads only to a minor modification of the numerical prefactor .
for example , for @xmath202 and 6 the thermalization rates are given by @xmath203 and @xmath204 , respectively .
it is interesting to note that in these 1d systems the thermalization rate due to the effective three - body collisions ( @xmath205}^{3b}$ ] ) is about a factor 3 smaller then the collision rate ( @xmath142 ) .
this suggests that in 1d systems thermalization requires also about 3 collisions , similar to 3d @xcite .
[ fig : thermalization ] shows numerical values of @xmath205}^{3b}$ ] as a function of the 1d density of @xmath57rb atoms and the radial trapping frequency . for comparison
we calculate numerically the thermalization rate @xmath205}^{2b } $ ] for two - body collisions involving the _ real _ transitions between the ground and excited radial states .
we follow hereby the same _ ansatz _ and perturb the velocity distribution of atoms in the ground and excited state as given by eq .
( [ eq:16 ] ) , the boltzmannian distribution of overall populations between the levels being kept intact . in the parameter range of interest
we find numerically @xmath205}^{2b}\approx ( 0.33\pm 0.03)\gamma _ { 2b}$ ] , i.e. @xmath206}^{2b } \approx 0.93\ , \omega _ r
\zeta e^{-\frac{2\hbar \omega _ r}{k_bt } } . \label{eq:30}\ ] ] the ratio of the thermalization rates for the two - body and three - body processes is therefore very close to the respective ratio of the collision rates , shown in fig .
[ fig : scatrat ] .
it is interesting to note that we find for both processes that thermalization in 1d needs about 3 collisions capable to distribute energy .
this is very close to the 2.7 collisions required for thermalization in 3d @xcite . for the typical parameters of an ultracold @xmath57rb gas on an atom chip @xcite ( @xmath207 khz , @xmath208m@xmath60 ) we obtain @xmath205}^{3b}\approx 2 $ ] s@xmath60 .
this thermalization rate is temperature - independent and much larger than the one calculated from the simple two - body collisions with the energy sufficient to excite radial modes @xmath205}^{2b}\approx 3 \times 10^{-3}$ ] s@xmath60 at the lowest temperatures measured ( 30 nk ) .
the estimated @xmath205}^{3b}$ ] is consistent with the time needed for evaporative cooling of a @xmath57rb gas on an atom chip well below @xmath90 @xcite .
the thermalization rate @xmath209 } $ ] given by eq .
( [ eq:17 ] ) was calculated for a weakly - interacting , non - degenerate gas .
calculation the thermalization rate @xmath210}$ ] in a general case requires to take into account additional physics .
first we need to consider the effects of quantum degeneracy and second the fact that the disperison relations for the elementary excitations in a 1d quantum ( degenerate ) system may differ significantly compared to a free particle , especially for phonon - like excitations . these effects , together with the bosonic amplification of the scattering into thermally populated modes , tend to accelerate thermalization .
a third observation is that the three body rates require three particles to be close to the same location .
quantum mechanically this is characterized by the third order correlation function @xmath211 .
a full consideration of the above competing effects will require extended numerical analysis of many particular cases and transcends beyond the scope of the present manuscript .
we will give here physical arguments of what to expect .
we start by pointing out that the form of the secondary - quantized hamiltonian eq .
( [ eq:11b ] ) allows us to give a simple estimate of the ratio of these two rates @xmath212}}/{\gamma ^{3b}_{[4 ] } } = \varrho ( { g_3(0)}/6)^2 \ , .\ ] ] here @xmath213 is a phase - space factor accounting for the dispersion law of elementary excitation , which changes from free - particle - like to phonon - like .
but it changes the thermalization rate less dramatically than the second factor associated with the local three - body correlation function @xmath214 and will become the dominating factor when approaching the strongly correlated regime ( @xmath215 ) .
for a _ non - degenerate _ weakly - interacting bose gas @xmath216 . for a _ degenerate _
1d bose gas @xmath211 has been recently calculated for the whole range of atomic repulsion strength ( @xmath217 ) by cheianov et al .
in the zero - temperature limit @xmath211 rapidly decreases from 1 to @xmath218 as @xmath219 grows from 0 to values @xmath4 .
we now turn to the above conjecture on suppression of the thermalization by atomic correlations .
a detailed calculation can be found in ref .
@xcite , here we sketch the basic physics argument .
to look at the correlations we start by considering @xmath147 identical bosons in 1d configuration with the hamiltonian [ eq:11b ] , which , after rescaling of units , takes the form @xmath220 @xmath221 is the strength of interaction of two atoms in the tight waveguide with ground state size @xmath16 .
@xmath222 is obtained by by adiabatic elimination of transverse modes virtually excited by the 3d short - range pairwise atomic interaction @xcite .
the explicit form of @xmath222 is given by eq .
( [ u3biiia ] ) within a numerical prefactor , @xmath223{\cal u}_\mathrm{3b}$ ] .
we follow now our detailed calcultions in ref .
@xcite and estimate the three - body scattering amplitude in the presence of the delta - functional pairwise repulsive interactions .
the stronger the _ pairwise _ interparticle repulsion , the smaller is the probability of a close encounter of three particles which will result in a suppression of the three - body scattering amplitude . the simplest case is to analyze the hamiltonian ( [ eq1 ] ) for @xmath224 particles . for that purpose
we express it in hyperspherical coordinates @xmath225 , @xmath226 defined as @xcite @xmath227 and obtain for the hamiltonian @xmath228 the corresponding schrdinger equation for the three - particle wave function is @xmath229 the wavenumbers @xmath230 are defined from the set of transcendental equations @xcite , provided that the periodic boundary conditions are set on the interval of the length @xmath25 . by setting @xmath231 , we obtain a continuous spectrum , where @xmath230 s are real for repulsive interaction ( @xmath232 ) .
we can now separate the center - of - mass motion and describe the relative motion in hyperspherical coordinates .
this leads to the _ ansatz _ : @xmath233 \psi _ \mathrm{r}(r,\chi ) \ ] ] where the kinetic energy of the relative motion is given by @xmath234\ ] ] . in the adiabatic hyperspherical approximation @xcite , which holds in the long - wavelength limit @xmath235 and neglects coupling of different scattering channels as well as accumulation of phase shifts of the scattered wave due to non - adiabatic effects we get @xmath236 the hyperangular part @xmath237 of this wave function
is the eigenfunction of the hamiltonian ( [ eq3 ] ) with fixed @xmath225 , corresponding to the lowest eigenvalue @xmath238 , which is the smallest positive root of the transcendental equation @xmath239=cr/\sqrt{2 } .
\label{eq0110}\ ] ] using the regular hexagon symmetry group @xcite of the hamiltonian ( [ eq3 ] ) , we write the hyperangular part of eq .
( [ eq60 ] ) as @xmath240 where @xmath241 , & | \chi |\leq \pi /3 \\ 0 , & \mathrm{otherwise } \end{array } \right . \label{eq103}\ ] ] after integrating out the hyperangular variable , the schrdinger equation in the adiabatic hyperspherical approximation reduces to @xmath242 f_0=k^2f_0 \label{eq17}\ ] ] where @xmath243 with the boundary conditions requiring @xmath244 to be finite for both @xmath245 and @xmath246 , for the partial wave " corresponding to the lowest eigenvalue @xmath238 , whose asymptotic expressions are @xmath247 we can solving now eq .
( [ eq17 ] ) analytically in two regions , @xmath248 and @xmath249 , with @xmath238 approximated by eq .
( [ eq18 ] ) , and tailoring the solutions by quasiclassical expressions for @xmath250 in the intermediate hyperradius range . the scattering amplitude @xmath251 ( for its definition in planar geometry see @xcite )
can then be obtained from the asymptotic form of the wave function at @xmath246 @xmath252 where @xmath253 is the hankel function of the first kind and @xmath254 are the third - order bessel functions .
the behavior of @xmath244 at @xmath255 is defined by the details of the potential @xmath256 , but the result can be finally expressed via the effective vertex of the three - body elastic collisions , thus yielding @xmath257 where @xmath258 is a numerical constant and @xmath259 is defined by eq .
( [ defxi ] ) .
we use thereby the fact that @xmath256 differs significantly from zero on the length scale @xmath16 , over which a virtually excited particle can propagate , and @xmath260 . from eq .
( [ eq209 ] ) we conclude that the three - body scattering amplitude decreases in proportion to @xmath261 as @xmath262 , i.e. , when the interaction is strong enough to induce significant atomic correlations .
the three - body scattering rate in a 1d system of bosons in the case of strong pairwise interaction is suppressed by a factor @xmath263 . by averaging over collision momenta in a moderately - excited strongly - interacting state
we obtain the scattering rate suppression factor @xmath264 .
we can now compare our result with the zero - distance three - particle correlation function @xmath211 . in the strong interaction limit @xmath4 @xmath265 @xcite and gives a direct physical motivation of our original conjecture @xcite that the pairwise interactions and the quantum correlations induced by them in a strongly - interacting 1d bosonic system suppress the three - body elastic scattering rate , and , hence , thermalization , by a factor @xmath266 .
in other words , strong quantum correlations extend the time scale , on which a quasi-1d system approaching the tonks - girardeau regime can be considered as approximately integrable . moreover
, we can corroborate this conjecture by observing that the thermalization rate is proportional to the square of the matrix element of the transition operator that is proportional to @xmath267 .
since the energies of the products of the _ elastic _ three - body process are low ( of about @xmath123 ) , we may assume that the correlations in the initial and the final states are the same , and the transition matrix element can be regarded as proportional to @xmath211 that yields again the @xmath268 scaling of the rate .
in contrast to this , the _
inelastic _ three - body processes are accompanied by a large energy release , and after an inelastic collision the newly formed dimer molecule and the fast atom leave the system almost immediately . therefore the inelastic three - body relaxation rate in a 1d ultracold bose gas is proportional to the first power of @xmath211 @xcite .
we now compare the calculated thermalization rates to the quantum newton s cradle experiment @xcite where a _ lower boundary _ for the damping time of the 1d motion towards a gaussian profile was estimated .
for three different lieb - liniger parameters @xmath269 , 3.2 , 18 kinoshita et al estimate lower bounds to the thermalization time from the consistency of the the experimental momentum distributions with the experimentally observed heating during the time interval of 0.5 , 0.5 , 1.0 seconds probed .
they find one sigma lower limits of 2.6 s , 25 s , and 13 s respectively . in their experiment the motion of two groups of @xmath57rb atoms
was excited at the relative velocity equal to 4 recoil velocities , which is far above the width of the ground - state velocity distribution .
we therefore can not expect the collision rate to be suppressed in proportion to @xmath270 that , as described above for slow collisions .
instead , we have to apply the estimates for damping of a fast particle discussed in the end of section iv . in a strongly - correlated system
one has to take the particle correlations into account and eq .
[ fast - qc ] has to be multiplied by a factor @xmath271 denoting the _ two - particle _ correlation function at zero distance ( tan , pustilnik and glazman @xcite ) : @xmath272 substituting the experimental parameters of ref .
@xcite and taking the values for @xmath271 from ref .
@xcite , we obtain @xmath273 , @xmath274 , and @xmath275 for @xmath269 , 3.2 , and 18 , respectively . to compare this calculated damping rates to the experiment in @xcite
one has to consider that ( 1 ) the two colliding clouds overlap only for a very short time during each oscillation and ( 2 ) that the thermalization rate is a factor 3 longer ( section v ) . taking this into account we estimate the respective thermalization times of 2.6 s , 35 s and @xmath276 1000 s. these rates are consistent with the experimental findings of ref .
for a more detailed comparison one would need longer time scale experiments with lower intrinsic heating and more detailed calculations of the dynamics including the damping due to three body collisions discussed here and in @xcite .
a radially confined atomic gas is never perfectly 1d , and radial motion can be excited , either in reality or virtually even if eq .
( [ eq:1 ] ) holds .
this possibility leads to effective three - body collisions , which arise in the second order of perturbation theory and can be associated with virtual excitation of radial modes .
these processes lead to thermalization even when two body collisions are frozen out at @xmath277 and provide a mechanism to break integrability in 1d systems .
in other words , the freeze - out of the radial modes is only a _
necessary _ , but not sufficient condition for integrability in 1d systems .
our estimations of the relaxation rates for weakly interacting quasi-1d bose gases are consistent with recent experimental observations for weakly - interacting quasicondensates @xcite .
these effective three - body collisions can be suppressed by quantum correlations caused by strong pairwise repulsions .
if they dominate , as in a strongly correlated 1d tonks - girardeau gas , they suppress the influence of the integrability - breaking interaction term .
the thermalization rate decreases in proportion to @xmath270 as the system enters the regime of strong correlations ( @xmath4 ) , and the system ( remaining non - integrable in the strict sense ) behaves like ( almost ) integrable on time scales short compared to the inverse thermalization rate . the effective three - body collisions , and their suppression by quantum correlations should be accessible in experiments looking at the damping of fast , particle - like excitations in systems with @xmath192 varying in a broad range of values from less than @xmath278 to @xmath279 .
manz s , bcker r , betz t , koller c , hofferberth s , mazets i e , imambekov a , demler e , perrin a , schmiedmayer j and schumm t 2009 _ two - point density correlations of quasi - condensates in free expansion _ , arxiv : 0911.2376 | we study the collisional processes that can lead to thermalization in one - dimensional systems . for two body collisions excitations of transverse modes
are the prerequisite for energy exchange and thermalzation . at very low temperatures excitations of transverse modes
are exponentially suppressed , thermalization by two body collisions stops and the system should become integrable . in quantum mechanics
virtual excitations of higher radial modes are possible .
these virtually excited radial modes give rise to effective three - body velocity - changing collisions which lead to thermalization .
we show that these three - body elastic interactions are suppressed by pairwise quantum correlations when approaching the strongly correlated regime .
if the relative momentum @xmath0 is small compared to the two - body coupling constant @xmath1 the three - particle scattering state is suppressed by a factor of @xmath2 , which is proportional to @xmath3 , that is to the square of the three - body correlation function at zero distance in the limit of the lieb - liniger parameter @xmath4 .
this demonstrates that in one dimensional quantum systems it is not the freeze - out of two body collisions but the strong quantum correlations which ensures absence of thermalization on experimentally relevant time scales
. + _ short title _ : * thermalization in a quasi-1d ultracold bosonic gas * |
the problem of measuring the scientific excellence of research institutes and universities is one of current and continued importance @xcite .
it is important to be able to detect and support the most promising research groups and to have sound and robust bases to wisely plan and invest for the future .
metrics and indicators of scientific activity may be based on the volume of researchers , the number of published papers , the amount of citations to these publications , the number of phd students and on finances generated by application of scientific ideas , amongst other factors .
it is often argued that automated scientometric or bibliometric indicators may be an inadequate substitute or proxy for peer - review based evaluations of the merit of research institutes @xcite .
there is a lot of arguments against the use of indicators in isolation , suggesting they reflect only one aspect of the research and should only be used as an adjunct to peer review ( for example , see @xcite ) .
but in spite of the problems with indicators and metrics , reasonable and cheap approaches to the evaluation of scientific productivity and quality are desirable for practical purposes , not least because institutional peer review is an expensive and time - consuming exercise @xcite . here
we investigate `` absolute '' and `` specific '' measurements of scientific excellence on the scale of university research groups .
e.g. , the _ absolute _ citation count for a department is the total number of citations , irrespective of how many researchers that department contains .
the corresponding _ specific _ citation count is then the average number of citations per head ( see , for example , @xcite ) . in physics parlance these correspond respectively to the notions of `` extensive '' and `` intensive '' quantities and are usually represented mathematically by upper - case and lower - case symbols . that convention is borrowed here for absolute and specific measurements . in this paper
we compare these specific and absolute notions of quality and strength with indicators of citation impact , to determine if the latter may be used as a reliable proxy for the former .
we consider the overall _ strength _ of a research collective ( a university department , research centre , group or some variant , component or amalgam of these ) as measured through a peer - review evaluation scheme to be an absolute quantity and denote it by @xmath0 .
a peer - review measure of the _ quality _ @xmath1 of a group of researchers is then the strength per head .
similarly , the total citation impact @xmath2 is considered here as the absolute impact of research group while its average value @xmath3 ( calculated per head ) corresponds to specific impact .
thus , `` absolute '' refers to total institutional measurements while `` specific '' means average properties per individual .
we then define _ strength _ as `` volume of quality '' and _ absolute impact _ as `` volume of average impact '' . in this paper , to make a quantitative comparison we use data from the british system of evaluation of research and education ( _ research assessment exercise _ ) and from _ evidence _ , a company within thomson reuters , one of the world s leading providers of scientific information . using biology research groups and departments in the uk as a test case ,
we show that the citation - based measure @xmath3 is not a good proxy for the peer - review measure @xmath1 , in that these two specific measures are rather poorly correlated . however , when scaled up to the actual size of the department @xmath4 , the _ absolute _ citation impact @xmath5 is very strongly correlated with the overall strength @xmath6 as measured by peer review .
( here and below @xmath4 means the number of researchers in group . )
this means that citations , if used in an informed manner , could possibly be used as a proxy for departmental or group _ strength _ particularly for large groups , but should not be as an estimate of research _
quality_. when applied to large research groups or departments in particular , this may offer an alternative or at least complement to peer review .
for the last quarter century , the _ research assessment exercise _ ( rae ) provided the premier yardstick for the measurement of scientific performance of research institutions in united kingdom .
this introduced an explicit and formalized process to assess the quality of research @xcite .
six evaluation exercise have occurred since the first one in 1986 and evaluation frameworks of this type now constitute the single most important event in the uk research calender each five or so years . for the purpose of the 2008 exercise ,
each academic discipline was assigned to one of 67 units of assessment ( uoa ) . in order to receive quality related funding any higher education institution ( hei )
could make submissions to rae in any uoa .
using published criteria , rae experts assess these submissions and generate graded profile for each of them .
these profiles quantify the proportion of a department s or research centre s work which falls into each of five quality bands as follows @xcite : * 4 * : quality that is world - leading in terms of originality , significance and rigour ; * 3 * : quality that is internationally excellent in terms of originality , significance and rigour but which falls short of the highest standards of excellence ; * 2 * : quality that is recognised internationally in terms of originality , significance and rigour ; * 1 * : quality that is recognised nationally in terms of originality , significance and rigour ; * unclassified quality that falls below the standard of nationally recognised work or work which does not meet the published definition of research for the purposes of the assessment . an example of a quality profile is given in table [ tab1 ] .
[ tab1 ] @xmath7an overall quality profile is constructed by summing the three separate weighted components ( `` output '' , `` environment '' and `` esteem '' ) using a special cumulative rounding methodology which avoids unfair consequences that simple rounding can produce @xcite . besides obvious benefits in terms of prestige , marketing and publicity potential , the research quality profiles determined how much funding each university department receives in the years following the exercise
funding is determined by a formula combining the weighted quality scores of individual research group or centres .
the formula is subject to regional and temporal variation , but the one introduced by the higher education funding council for england ( hefce ) immediately subsequent to rae 2008 was @xcite @xmath8 here @xmath9 represents the percentage of a team s research which was rated @xmath10 and @xmath1 may be considered as a single _ specific _ measure of a university s overall research quality in a particular discipline .
the total amount of quality related funding distributed by hefce to a given university post rae is proportional to the overall _ strength _ of the submission @xmath11 where @xmath4 is the number of researchers submitted from a given university .
thus , strength @xmath0 is an _
absolute _ measure of performance . notwithstanding various managerial tactics and manoeuvres , the rae , and similar exercises in other countries ,
is considered to be reasonably reliable since it is based on peer - review evaluation . despite its many drawbacks and limitations ,
peer review is considered by the academic community at large to be the most reliable and trustworthy scheme to evaluate the worth of curiosity - driven research , in particular . on the other hand ,
peer - review - based evaluation is expensive and time - consuming process .
in addition , the very act of measuring the scientific system distorts the very process it purports to measure .
this is goodhart s law , a type of socioeconomic counterpart of the heisenberg uncertainty principle in physics .
therefore , it is reasonable to propose some alternatives to the rae .
evaluations based entirely upon citation counts are some of many possible candidate schemes ( see , for example , @xcite ) .
some authors advocate substitution of the rae by citation counting due to claimed good correlation between the resultant rankings obtained @xcite .
but there are also authors ( e.g. , @xcite ) who point to different weaknesses of citation analysis such as self - citing , bias , technical errors connected with citation data retrieval or different typical rates of citations for various disciplines .
although methods have been , and are being , developed to try to deal with these phenomena , no universally satisfactory automated system has yet emerged .
so , the question arises as to how good citation - based indicators can be as proxies for peer review . since the last uk exercise generated quality measurements in far more detail than previous raes , and since subsequent evaluation frameworks in the uk will amalgamate some of the uoa s used in 2008 into fewer , larger subject groupings , rae 2008 provides a window of great detail in which answers to this question may be sought . having precise numbers assigned to each institution for the 67 disciplines it is possible to perform accurate comparisons between rae peer - review evaluation and citation scores .
thus we arrive at the main questions we wish to address herein : how do rae 2008 scores correlate with citation rates ?
as already noticed before , to get the citation - based measure , we use the data provided by _
this company offers a service analysing research performance tailored to individual client requirements @xcite .
we consider the so - called _ normalised citation impact _
( nci ) @xmath3 used by _ evidence _ as a coefficient of departmental performance in a given discipline . _
evidence _ calculate the nci using data from thomson reuters databases @xcite .
similarly to relative citation rate ( rcr ) ( i.e. , @xcite ) , nci is a relative measure as it is calculated by comparing to a mean or expected citation rate . on the other hand , this is a specific measure of citation impact because it is averaged to be a measure of impact for research group .
it has long been known that citation rates are different for different disciplines .
the main advantage of _ evidence _ calculations is non - trivial normalisation of citation counts between different academic disciplines . to achieve this ,
the total citation count for each paper can be normalised to an average citations per paper for the year of publication and either the field or journal in which the paper was published ( the so - called `` rebasing '' of the citation count ) .
the normalised value is known as the nci @xcite .
only the four papers per individual which were submitted to rae 2008 were taken into account by _ evidence _ in order to calculate the average nci for research groups @xcite .
the rae does not emplot a comparable normalisation process and this is a serious and acknowledged weakness of that peer evaluation exercise @xcite .
thus , the nci may be considered as a _ specific _ measure of the research impact of a department in a given field and we denote it by @xmath3 .
the corresponding _ absolute _ measure of impact ( the total impact of the department or group ) is denoted by @xmath2 .
the relationship between the two is @xmath12 in the following , we compare the _ specific _ indicators of quality and impact @xmath1 and @xmath3 as measures of the average strength and impact of the group or department _ per individual _ contained within it .
we also compare the counterpart _ absolute _ measurements @xmath0 and @xmath2 as measures of the overall strength and total impact of the group as a whole .
we use the biology research sector in the uk as a test case .
we show that while the specific measures @xmath1 and @xmath3 are only weakly correlated , the absolute ones @xmath0 and @xmath2 are strongly aligned .
the analysis thus suggests that citations form a poor basis for measuring institutional research quality but could form a good basis for measuring strength . since quality related funding as defined by hefce is proportional to departmental strength rather than quality
, the citation - based measures may offer a reasonable and cost effective way to decide on funding while ameliorating the negative effects of goodhart s law .
we have to note that although rae assessments as well as the evidence results are used for practical purposes , they are still subject to debate within the scientometric community .
this is an additional reason to compare them .
.the ranking of uk biology departments using the rae 2008 scores @xmath1 and the nci - based scores @xmath3 . [ cols="<,<,<,<,<",options="header " , ] [ tab_coefficients ]
a conclusion of the above analysis is that there is only weak correlation between the specific measure of impact @xmath3 and the rae peer - evaluated measure of quality @xmath1 , and that this is especially weak , or absent for small groups .
possible reasons for this include the facts that ( i ) there are few small and medium groups ( only 5 small groups and only only 7 medium groups ) in biology and ( ii ) small groups , by definition , have relatively few members . both of these factors mean that the statistics for small ( and to a lesser extent medium ) groups are relatively poor in comparison to those for large groups .
since the number of large groups is bigger and also the statistics for large groups are better , more robust results can be obtained for large groups . moreover , in the above analysis , all research groups are treated as contributing the same weight to the analysis .
e.g. , the three researchers from bath spa university contribute to the same extent as the 213.69 biologists from cambridge university ! it is therefore sensible to correct this anomaly by introducing weights . multiplying the average quality of groups by their size renders the specific measures absolute : quality becomes strength and the nci is also scaled up to the volume of the group or department .
the rae , peer - reviewed measures of strength @xmath0 are compared to the absolute , citation - based measures of nci in fig .
[ fig5_rae_vs_evidence_unnormalized ] .
the correlation between the two measures is impressive .
this is reflected in the almost perfect person correlation coefficient @xmath13 ( see table [ tab_coefficients ] ) .
moreover , restricting the analysis only to large groups also reveals an excellent correlation of @xmath14 .
the correlation coefficients for medium groups alone is less good , at @xmath15 and for small groups that figure is @xmath16 .
however , since small and medium research groups tend to have the same linear dependency of quality on quantity @xcite it is more sensible to combine them in the correlation analysis .
indeed , the pearson correlation coefficient for small and medium groups combined is @xmath17 .
thus we can say that the nci forms a good proxy for peer - review estimates of group research strength for large and small / medium research departments .
[ fig5_rae_vs_evidence_unnormalized ] should be compared to fig .
[ fig1_rae_vs_evidence ] .
the replacement of specific measures of quality and impact by their absolute counterparts has had the effect of stretching the corresponding axes by an amount proportional to the quantity of the groups or departments . because of the clear relationship between quality and quantity identified in ref .
@xcite , and a similar relationship between nci impact and quantity observed in ref .
@xcite , this stretching induces the improved correlations observed .
a goal of scientometrics is to develop a method to provide a reliable measurement of scientific excellence using minimal efforts .
however , any attempts to replace current peer - evaluation based systems have to be done carefully and robustly . quality related funding , such as that administered by hefce post rae in the uk is proportional to absolute measures of research strength since financial support for research obviously needs to be weighted according to group size .
therefore , a citation based approach , at least for large groups , may be introduced as less intrusive and more cost effective alternative to national peer review .
this would also have the advantage of ameliorating the distortions to the research system introduced on a national basis every five years or so through goodhart s phenomenon , although one might expect a citation - based system to introduce distortions of its own .
however , we have also seen that the citation metrics used here are not well correlated with peer - review measures of group research excellence .
since the latter are used to rank institutions , it is clear that citation counts are not a good basis on which to make such comparisons .
thus we arrive at the almost paradoxical conclusion that the citation - based metric ( nci ) may be used as an excellent proxy for peer - reviewed measurements of institutional scientific strength but it is only a poor proxy for measures of quality .
since quality related funding is strength based , this may offer a much cheaper alternative to the system currently in use in the uk and some other countries .
the analysis presented here is based on biology research groups in the uk .
further analyses are underway for other disciplines to determine the broader suitability of scientometric measures of specific and absolute research - group properties .
the nuances connected with peculiarities of groups of different sizes should be also studied . since in the uk system ,
a large group with , say 40 members , provides @xmath18 submissions , which may be sufficiently large for statistical fluctuations to be ironed out .
but the rae evaluation of a small group , say with 5 staff members , is based on the statistics of only approximately @xmath19 submissions .
this is far more susceptible to inaccuracies and statistical noise .
so , any automated evaluation of large groups is much more robust than for small groups even if one is only interested in measuring strength rather than quality . a possible way to balance this ,
if introducing citation counts , would be to keep peer review for small and medium sized groups or to require a greater number of outputs per person for such groups to improve statistics .
investigations into such schemes are underway . in any case , it is clear from the current test case that any attempt to automatically evaluate research quality should very carefully take account of group or departmental sizes @xcite .
this work was supported in part by the 7th fp , irses project no .
269139 `` dynamics and cooperative phenomena in complex physical and biological environments '' and irses project
295302 `` statistical physics in diverse realizations '' .
the authors thank jonathan adams from evidence for the data and ihor mryglod for fruitful discussions .
kenna r. , berche b. , critical masses for academic research groups and consequences for higher education research policy and management , higher education management and policy , 2011 , * 23 * , iss . 3 , 121 .
holmes a. , oppenheim c. use of citation analysis to predict the outcome of the 2001 research assessment exercise for unit of assessment ( uoa ) 61 : library and information management .
information research , ( 2001 ) , * 6(2)*. retrieved from http://informationr.net/ir/6-2/paper103.html the overall quality profile , and sub - profiles for research outputs , research environment and esteem indicators for each submission for rae 2008 could be found here : http://www.rae.ac.uk/pubs/2009/pro/quality_profiles.xls | a desirable goal of scientific management is to introduce , if it exists , a simple and reliable way to measure the scientific excellence of publicly - funded research institutions and universities to serve as a basis for their ranking and financing .
while citation - based indicators and metrics are easily accessible , they are far from being universally accepted as way to automate or inform evaluation processes or to replace evaluations based on peer review . here
we consider absolute measurements of research excellence at an amalgamated , institutional level and specific measures of research excellence as performance per head .
using biology research institutions in the uk as a test case , we examine the correlations between peer - review - based and citation - based measures of research excellence on these two scales .
we find that citation - based indicators are very highly correlated with peer - evaluated measures of group strength but are poorly correlated with group quality .
thus , and almost paradoxically , our analysis indicates that citation counts could possibly form a basis for deciding on how to fund research institutions but they should not be used as a basis for ranking them in terms of quality . |
single electron control in semiconductor quantum dots would be valuable for spintronics , quantum information , and spin qubits@xcite .
adiabatic time evolution of degenerate eigenstates of a quantum system provides a means for controlling individual quantum states@xcite .
they exhibit non - abelian gauge structures and often give finite non - abelian berry phases ( they are also called matrix berry phase or holonomic phase ) .
these phases often depend on the geometry of the path traversed in the parameter space of the hamiltonian .
nuclei@xcite , superconducting nanocircuits@xcite , optical systems@xcite , and atomic systems@xcite have such degenerate quantum states .
it has been shown that universal quantum computation is possible by means of non - abelian unitary operations@xcite .
manipulation is expected to be stable since symmetries of the hamiltonian that give rise to degeneracy are not broken during the adiabatic transformations . however , the degree of its stability is under investigation@xcite .
recently several interesting possibilities for electronic manipulation in semiconductors have been proposed .
spin manipulation of quasi - two - dimensional electrons by time - dependent gate voltage is possible through the dresselhaus and rashba spin - orbit coupling mechanisms @xcite .
a geometric spin manipulation technique based on acceptor states in p - type semiconductors with spin - orbit coupling was proposed@xcite .
spin - orbit coupling and a revolving external electric field may generate holonomic qubit operations in cdse nanocrystals @xcite .
holonomic quantum computation using excitons in semiconductor nanostructures has been also proposed@xcite .
a matrix berry phase is experimentally interesting because it represents mixing between degenerate levels .
let us explain briefly the basic ideas using a simple system possessing kramers double degeneracy@xcite .
the hamiltonian depends on some external parameters @xmath0 .
if the system is in a superposition state @xmath1 at time t=0 an adiabatic evolution of the parameters @xmath0 can transform this state into another state @xmath2 after some time @xmath3 . here the orthonormal basis states @xmath4 are the instantaneous eigenstates of the hamiltonian . for
a cyclic change with the period @xmath5 , represented by a closed contour @xmath6 in the parameter space , the states @xmath7 and @xmath8 return to the initial states @xmath9 and @xmath10 , but the coefficients @xmath11 and @xmath12 may not return the initial values .
in such a case a @xmath13 matrix berry phase ( non - abelian berry phase ) @xmath14 is generated @xmath15 this non - abelian geometric phase ( holonomy ) is connected to non - abelian gauge potentials , as we discuss below . during the adiabatic cycle
the degenerate energy @xmath16 varies with time .
but hereafter we will set @xmath17 since it can be easily restored when needed .
the expansion coefficients satisfy the time - dependent schr@xmath18dinger equation @xmath19 the matrix elements @xmath20 are given by @xmath21 , where the sum over @xmath22 in @xmath20 is meant to be the sum over @xmath0 .
@xmath23 are @xmath13 matrices and are called the non - abelian vector potentials .
they are given by @xmath24 the formal solution of the time - dependent schr@xmath18dinger equation gives the matrix berry phase @xmath25 , where @xmath26 represents a path ordering . under a unitary transformation @xmath27
the non - abelian gauge structure emerges @xmath28 here we propose that quantum dots in n - type semiconductors with spin - orbit interaction can have matrix berry phases .
all the discrete energy levels of quantum dots possess a double degeneracy because the rashba and/or dresselhaus spin orbit coupling terms have time reversal symmetry . in ii - vi semiconductors the rashba term
is expected to be larger than the the dresselhaus coupling . in iii - v semiconductors , such as gaas ,
the opposite is true@xcite . in our work both the rashba and dresselhaus terms are included .
the adiabatic transformation can be performed electrically by changing the confinement potentials of the quantum dot , see fig.[fig : setup](d ) .
we find that , although the spin orbit terms formally give rise to a non - abelian structure for the matrix vector potentials , double degeneracy does not necessarily lead to finite non - abelian berry phases .
we show that when the circular symmetry of two - dimensional harmonic potentials is broken matrix berry phases can be produced ( such a distortion is shown in fig.[fig : setup](b ) ) .
we propose that the presence of a matrix berry phase may be detected by measuring the the optical spectrum . in sec .
ii we describe the hamiltonian of the system in detail , and in sec .
iii we discuss a model calculation of the matrix berry phase .
a possible experimental detection of the the matrix berry phase is suggested in sec .
conclusions are given in sec .
an electric field @xmath29 is applied along the z - axis and electrons are confined in a triangular potential @xmath30 , see fig.[fig : setup](a ) . when the the width of the quantum well along the z - axis is sufficiently _ small
_ we may include only the lowest subband state along the z - axis .
we denote this wavefunction by @xmath31 , see fig.[fig : setup](a ) . from the expectation value @xmath32
we estimate the characteristic length scale along the z - axis : @xmath33 .
the hamiltonian in the absence of the spin orbit coupling is @xmath34 .
we take the two - dimensional potential to be @xmath35 , see fig.[fig : setup](b ) .
the strengths of the harmonic potentials are denoted by @xmath36 and @xmath37 .
they may be varied by _ changing gate potentials _ of the quantum dot system .
the characteristic lengths scales along x- and y - axis are @xmath38 .
the potential @xmath39 represents a distortion of the harmonic potentials and its strength @xmath40 may be _ varied electrically_. in a periodic crystal potential of a semiconductor the spin orbit interaction has two contributions .
the rashba spin orbit term is @xmath41 here @xmath42 are pauli spin matrices and @xmath43 are momentum operators ( @xmath44 and similarly with @xmath45 . ) . the constant @xmath46 _ depends on the external electric field _
@xmath29 applied along the z - axis .
the dresselhaus spin orbit term is @xmath47 there is another term of the form @xmath48 in the dresselhaus spin orbit term but it vanishes since the expectation value @xmath49 for the first subband along z - axis .
the constant @xmath50 represents breaking of inversion symmetry by the crystal in zinc blende structures .
the total hamiltonian of an electron in a semiconductor quantum dot is @xmath51 .
an eigenstate of the hamiltonian @xmath52 may be expanded as a linear combination of the eigenstates of @xmath53 @xmath54 the expansion coefficients satisfy a matrix equation @xmath55 . in the basis states @xmath56 the quantum number @xmath57 and @xmath58 label the harmonic oscillator levels along the x - axis ( y - axis ) and the component of electron spin .
the subband wavefunction @xmath31 is suppressed in the notation @xmath59 . in the absence of the zeeman term
the total hamiltonian is invariant under time reversal symmetry : @xmath60 and @xmath61 .
the time reversal operator is @xmath62 , where the operator @xmath6 stands for complex conjugation .
the time reversed state of @xmath63 is @xmath64 note that @xmath65 .
these two states are degenerate and orthonormal .
we have suppressed the bloch wavefunction of the conduction band in applying the time reversal operator since it is unaffected by the operator @xmath66 .
our wavefunctions are all effective mass wavefunctions and only the conduction band bloch wavefunction at @xmath67 is relevant . an adiabatic change implemented by changing the energy parameters @xmath68 and @xmath69 that characterize the rashba term and the distortion potential @xmath70 : @xmath71 where @xmath72 .
the first parameter @xmath68 depends on the electric field through the constant @xmath46 .
the typical value of the energy scale associated with the rashba constant depends on the electric field applied along the z - axis and the semiconductor material : it is of order @xmath73 , where the length scale @xmath74 is the lateral dimension of the quantum dot .
the second adiabatic parameter @xmath69 represents the strength of the distortion potential @xmath75 : the expectation value of the distortion potential is @xmath76 .
its magnitude is of order @xmath77 , depending on the electric field applied along the y - axis and the width of the triangular potential @xmath30 .
we work out a model that can be solved analytically .
this model is simple but much can be learned from it .
let us take @xmath78 ( other values of @xmath36 can also be chosen ) .
then the lowest eigenenergy state of @xmath53 is @xmath79 with the energy @xmath80 and the next lowest eigenenergy state is @xmath81 with the energy @xmath82 .
the typical value of the energy spacing between the quantum dot levels , @xmath83 , is of order @xmath84 .
let us write down the eigenstates of the total hamiltonian as a linear combination of four basis states made out of these states and spin degree of freedom : @xmath85 the @xmath86 truncated hamiltonian matrix is @xmath87 or @xmath88 the matrix elements of the distortion potential are @xmath89 .
note that the distortion potential couples even and odd parity states of the one - dimensional harmonic potential .
other functional form of @xmath70 may be also used to generate the non - abelian berry phase as long as it couples even and odd parity states . the first term in the dresselhaus spin orbit term , eq .
( [ dresselhaus ] ) , is zero since @xmath90 . from the second term of the dresselhaus term we get @xmath91 , where the constant @xmath92 with @xmath93 and @xmath94 .
here we have used the momentum matrix elements @xmath95 .
the magnitude of the energy scale associated with the dresselhaus term is of order @xmath96 , and it can be larger or smaller than the rashba term , depending on the material@xcite .
the function @xmath97 depends on @xmath68 because the electric field @xmath29 enters through @xmath98 .
there is no simple way to calculate @xmath46 because it depends both on the electric field inside the semiconductor heterostructure and on the detailed boundary conditions at the interface . for simplicity
we take @xmath99 and @xmath100 , where @xmath101 and @xmath102 are numerical constants .
more complicated functions , for example , @xmath103 and @xmath104 , could be used , but our calculation indicates that the essential physics does not change .
diagonalization of this @xmath86 hamiltonian matrix gives the eigenenergies @xmath105 .
let us choose the following state @xmath106 and its time reversal state @xmath107 as the basis states in the lowest energy degenerate hilbert space .
in the evaluation of the matrix @xmath23 we use @xmath108 , where @xmath109 .
note that @xmath110 is the two - dimensional harmonic oscillator wavefunction and that it is a real function .
one can show that @xmath111 , where the pair of degenerate states are @xmath112 and @xmath113 .
the orthonormalization @xmath114 gives that the diagonal matrix elements @xmath115 are real and that the off - diagonal elements satisfy @xmath116 .
the adiabatic parameters are @xmath117 , @xmath118 .
we calculate the matrix berry phase for @xmath119 .
the matrix vector potentials with respect to @xmath120 and @xmath121 have the following structures @xmath122 and @xmath123 the functions @xmath124 , @xmath125 , and @xmath126 depend on @xmath68 and @xmath69 and are real . under an adiabatic time evolution a state in this degenerate hilbert space changes as @xmath127 .
suppose initially @xmath128 and @xmath129 , i.e. , @xmath130 .
the cyclic adiabatic path with the period @xmath131 is shown in fig.[fig : setup](d ) and is given by @xmath132 the frequency @xmath133 can be taken to be a fraction of @xmath83 .
we solve eq.([eq : time_schrod ] ) numerically and find that @xmath12 is non - zero .
for the parameters @xmath134 , @xmath135 , @xmath136 , @xmath137 , and @xmath138 we find @xmath139 and @xmath140 .
the matrix berry phase is absent when the distortion potential @xmath70 is zero . when only rashba and/or dresselhaus terms are present in eq.([matrix_hamil ] ) the adiabatic transformation can be performed by varying the parameters @xmath117 and @xmath141 .
in this case _ an orthonormal basis set exists _ in the degenerate hilbert space such that the matrix vector potentials are diagonal and the non - abelian berry phase is zero .
this degenerate basis set , @xmath142 and @xmath143 , has the property that for each @xmath144 either @xmath145 or @xmath146 , where @xmath147 ( see , for example , eqs.([eq : finite_b_1 ] ) and ( [ eq : finite_b_2 ] ) ) .
the off - diagonal matrix elements of the vector potential @xmath148 because either @xmath145 or @xmath146 .
therefore the matrix vector potentials are diagonal . in this case
the matrix berry phase will be absent .
let us construct explicitly @xmath149 and @xmath150 when only the rashba term is present , i.e. , when @xmath151 .
they are given by @xmath152 and @xmath153 with these new eigenstates it is possible to show that the non - abelian gauge potentials are not only diagonal but also zero : @xmath154 and @xmath155 .
let us also test the non - abelian gauge structure given by eq .
( [ eq : vec_pot ] ) .
we make a unitary transformation from ( @xmath149 , @xmath150 ) , given in eqs.([eq : finite_b_1 ] ) and ( [ eq : finite_b_2 ] ) , to a pair of time reversed degenerate eigenstates @xmath156 , @xmath157 and @xmath158 using these new basis states one can show that the non - abelian gauge potentials are @xmath159 and @xmath160 since the old vector potentials are @xmath155 and @xmath154 it follows from eq .
( [ eq : vec_pot ] ) that @xmath161 .
we have explicitly verified that this relation holds by computing the unitary matrix @xmath162 .
we have also verified independently the absence of the matrix berry phase by solving the time - dependent schr@xmath18dinger eq.([eq : time_schrod ] ) with @xmath163 and @xmath164 .
this provides a check on our numerical method of solving the time - dependent schr@xmath18dinger equation .
after an adiabatic cycle the electron acquires a matrix berry phase .
the presence of such a phase may be detected by measuring the strength of dipole optical transitions before and after the cycle . in this section
we calculate the optical strengths using the truncated hamiltonian .
this calculation is not quantitatively accurate but it will give us an estimate of the magnitude of the effect . first we need to prepare physically some particular pair of degenerate states . in the presence of a magnetic field along the z - axis any degenerate pair of states at zero magnetic field
will be split into two states .
we define the lowest energy pair of degenerate eigenstates in the zero magnetic field limit of @xmath165 as @xmath166 and @xmath167 , where @xmath168 and @xmath169 are the split lowest and next lowest energy states in a finite magnetic field , see fig.([fig : mag_field_limit]a ) . in a similar way we define degenerate eigenstates in the zero magnetic field limit of @xmath170 : @xmath171 and @xmath172 , see fig.([fig : mag_field_limit]b ) .
the states @xmath173 and @xmath174 are calculated from the hamiltonian by replacing @xmath175 with @xmath176 and adding the zeeman term ( @xmath177 is the vector potential and @xmath178 ) .
suppose that the electron is in the lowest energy state . in order to measure the matrix berry phase we perform the following set of procedures : 1 .
we apply a magnetic field along the x - axis and take the zero field limit .
the resulting state @xmath179 ( fig.([fig : mag_field_limit]b ) ) can be written as a linear combination of @xmath180 and @xmath181 ( fig.([fig : mag_field_limit]a ) ) : @xmath179=@xmath182 with @xmath183 .
we apply adiabatically a magnetic field along the z - axis .
we increase it to a value @xmath165 .
then the probability for the electron to be in the state @xmath168 is given by @xmath184 .
3 . we then measure the intensity of the dipole transition from the lowest energy state to the third lowest energy state , see fig.([fig : mag_field_limit]c ) .
this intensity is proportional to @xmath184 , i.e , proportional to the probability that @xmath168 is occupied .
the lowest and third lowest energy states can be written as a linear combination of the basis states @xmath185 : both states can be written in the form @xmath186 .
when the photon is polarized along the y - axis only the basis states with different parities are coupled in the dipole approximation , for example , @xmath187 is non - zero .
again we assume that the electron is in the lowest energy state .
we perform the second set of procedures : 1 .
we apply a magnetic field along the x - axis and take the zero field limit .
2 . we then perform an adiabatic cycle following a closed loop in the parameter space of @xmath188 , given by eq.([closed_loop ] ) . after the adiabatic cycle the electron will be in the state @xmath189 with @xmath190 ( the parameters are @xmath191 , @xmath192 , @xmath193 , @xmath194 , and @xmath195 ) .
we apply adiabatically a magnetic field along the z - axis , see fig.([fig : mag_field_limit]c ) .
then the probability for the electron to be in the first lowest energy state , @xmath168 , is @xmath196 while the probability that the electron will be in the second lowest energy state , @xmath169 , is @xmath197 . here
the value of @xmath165 is the same as that of step 2 in the first set of procedures .
we then measure the intensity of the dipole transition from the lowest energy state to the third lowest energy state .
this intensity is proportional to @xmath196 , i.e , proportional to the probability that @xmath168 is occupied .
these two sets of measurements are repeated many times .
then the ratio @xmath198 gives the intensity ratio of the dipole transitions in the first and second sets of procedures ( the dipole matrix elements cancel each other ) .
it is the direct measure of the matrix berry phase .
the analysis of the intensity of optical transitions in zero magnetic field is complicated due to the presence of the degeneracy in the final states of the transition . in a finite magnetic field
this degeneracy is lifted .
each discrete eigenstate of a semiconductor quantum dot with the rashba and/or dresselhaus spin orbit coupling terms possesses a double degeneracy due to time reversal symmetry .
we have investigated the matrix berry phase of such a quantum dot in a simple truncated model hamiltonian that can be solved analytically .
we have found that the double degeneracy does not necessarily lead to a finite non - abelian berry phase .
the addition of a parity breaking distortion potential to the hamiltonian when both the rashba and dresselhaus spin orbit coupling terms are present gives rise to a finite matrix berry phase .
we have proposed that this phase may be detected in the dipole transitions between the ground and first excited states in a magnetic field .
for an accurate modeling of possible experiments the number of basis vectors in the truncated hamiltonian matrix must be chosen sufficiently large .
this number will be determined by the ratios @xmath199 .
calculation of the matrix berry phase in such a case requires a heavy numerical computation .
accurate experimental determination of the functional dependence of the rashba constant on the electric field would be also valuable in determining the magnitude of matrix berry phases .
the quantum dot studied in this work contains a single electron .
it may be worthwhile to investigate the effect of many body interactions .
this work was supported by grant no .
r01 - 2005 - 000 - 10352 - 0 from the basic research program of the korea science and engineering foundation and by quantum functional semiconductor research center ( qsrc ) at dongguk university of the korea science and engineering foundation .
awschalom , d. loss , and n. samarth , semiconductor spintronics and quantum computation ( springer , berlin , 2002 ) . , edited by a. shapere and f. wilczek ( world scientific , singapore , 1989 ) .
f. wilczek and a zee , phys .
lett . * 52 * , 2111 ( 1984 ) ; a. zee , phys . rev.a * 38 * , 1 ( 1988 ) .
j. a. jones , v. vedral , a. ekert and g. castagnoli , nature * 403 * , 869 ( 2000 ) . g. falci , r. fazio , g. palma , j. siewert and v. vedral , nature * 407 * , 355 ( 2000 ) ; m.s .
choi , j. phys . : condens .
matter * 15 * , 7823 ( 2003 ) ; l. faoro , j. siewert , and r. fazio , phys .
lett . * 90 * , 028301 ( 2003 ) .
j. pachos and s. chountasis , phys .
a. * 62 * , 052318 ( 2000 ) . c. a. mead , phys .
lett . * 59 * , 161 ( 1987 ) ; j. segert , j. math .
* 28 * , 2102 ( 1987 ) .
duan , j. i. cirac and p. zoller , science , * 292 * 1965 ( 2001 ) .
r. g. unanyan , b.w .
shore , and k. bergmann , phys .
a * 59 * , 2910 ( 1999 ) .
p. zanardi and m. rasetti , phys . lett . * 264 * , 94 ( 1999 ) .
j. pachos , p. zanardi and m. rasetti , phys . rev . a * 61
* , 010305(r ) ( 2000 ) ; j. pachos and p. zanardi , int . j. mod
b * 15 * , 1257 ( 2001 ) .
d. parodi , m. sassetti , p. solinas , p.zanardi , and n. zanghi , quant - ph/0510056 .
e.i . rashba and al .
l. efros , phys .
lett . * 91 * , 126405 ( 2003 ) .
b. a. bernevig and s .- c .
zhang , phys .
b * 71 * , 035303 ( 2005 ) .
serebrennikov , phys .
b * 70 * , 064422 ( 2004 ) .
p. solinas , p. zanardi , n. zanghi , and f. rossi , phys . rev.a
* 67 * , 062315 ( 2003 ) .
e. rashba cond - mat/0507007 | we propose that n - type semiconductor quantum dots with the rashba and dresselhaus spin orbit interactions may be used for single electron manipulation through adiabatic transformations between degenerate states .
all the energy levels are discrete in quantum dots and possess a double degeneracy due to time reversal symmetry in the presence of the rashba and/or dresselhaus spin orbit coupling terms .
we find that the presence of double degeneracy does not necessarily give rise to a finite non - abelian ( matrix ) berry phase .
we show that a distorted two - dimensional harmonic potential may give rise to non - abelian berry phases .
the presence of the non - abelian berry phase may be tested experimentally by measuring the optical dipole transitions . |
high harmonic generation can be used for attosecond imaging of molecules @xcite .
this is a consequence of the fact that the structure of an aligned molecule can be inferred from quantum - interference patterns , which appear in the high - harmonic spectra
. understood in terms of the three - step model , in which an electron reaches the continuum by tunneling or multiphoton ionization , is accelerated by the field and subsequently recombines with a bound state of its parent molecule , emitting a high - order harmonic photon @xcite , this phenomenon occurs due to the electron returning to different spatially separated centers .
the simplest scenario is high - harmonic generation in a diatomic molecule , for which , in principle , recombination and thus high - harmonic emission can take place from either atomic center .
this is a microscopic equivalent to the double slit experiment doubleslit . how to accurately describe the whole process of high harmonic generation ( hhg ) in molecules is still an open problem .
there are two main theoretical approaches that can be used .
firstly one can attempt a fully numerical solution by using density functional theory , which gives accurate results but inhibits a more complete understanding of the process @xcite .
secondly , one can employ the traditional atomic three step model , within the strong - field approximation @xcite , and calculate the molecular bound - state wavefunctions using a quantum chemistry code @xcite .
in this framework , the bound states of the molecule are mainly incorporated in the ionization and recombination steps , in the form of prefactors .
the latter approach allows a better physical understanding of the problem . the price one pays , however , is that , in order to be able to treat hhg within a semi - analytical framework , several approximations must be made .
it is well known that the harmonic spectrum is proportional to the modulus squared of the fourier transform of the dipole acceleration . by using the ehrenfest theorem
the dipole acceleration can be replaced by the dipole velocity or the dipole length , with no loss of accuracy so long as the integration is over all time and no physical approximations are made upon all wavefunctions involved .
however , when the sfa is used , this is no longer the case , and replacement of the dipole operator gives rise to inconsistencies in the sfa dipole matrix elements , which vary depending on the form of the operator used .
the most well known of these is the lack of translational invariance when using the length form of the dipole operator . in the single - active electron approximation
, it has been shown that the velocity form gives the most reliable results @xcite and is easier to use than the acceleration form . in calculating the sfa
prefactors the simplest approach is to use the highest occupied molecular orbital ( homo ) @xcite .
however , the problems that arise in this case are 1 ) which orbitals have the predominant influence on high harmonic generation and 2 ) how to accurately include the multielectron effects of electron exchange and correlation .
the first problem has to some extent been addressed in ref .
@xcite by noting that the ionization potential of different orbitals varies with the rotation of the molecule and hence the corresponding cutoffs in the high harmonic spectra will be different .
furthermore , the symmetry of the orbital will strongly affect the contribution of specific orbitals to high - order harmonic spectra .
this is particularly true as the alignment angle of the molecule with respect to the laser - field polarization varies . in previous work @xcite
we also investigated high - order harmonic generation beyond the single - active orbital approximation .
in particular , we addressed the quantum interference between the homo and the homo-1 of @xmath1 and the homo and the lowest unoccupied molecular orbital ( lumo ) of @xmath2 .
for the latter case , we have also observed that electron recombination to different orbitals can be mapped into different cutoffs in the hhg spectra .
our investigations , however , did not incorporate electron correlation .
in fact , we have either dealt with simplified multielectron models , in which the time evolution of the optically active electrons has been decoupled , or with a coherent superposition of one - electron states .
the second problem is what we propose to investigate in this paper and in doing so to consider the influence that these multielectron effects have on high harmonic generation . in particular , we wish to incorporate such effects perturbatively around the contributions of the highest occupied molecular orbital .
although correlation in molecular high - order harmonic generation has been addressed in refs .
@xcite , utilizing a multielectron version of the strong - field approximation , such studies have been performed numerically to a great extent . including the correlation analytically in the dipole matrix elements facilitates more insight into the effects that correlation has on the high harmonic spectra . in ref .
santra , many - body perturbation theory was used to include electron correlation as corrections to the standard three step model .
therein , however , only the response of single atoms has been addressed . from the viewpoint of
which form of the dipole operator to use , we are particularly interested in how such corrections are affected by the different formulations , especially in molecular targets .
indeed , even in the single - active electron approximation , the discrepant results obtained for each dipole form have raised considerable debate . by extending the debate to multielectron effects in molecules , more insight into
which operator is most accurate for harmonic generation , and indeed all strong field phenomena , is gained .
the use of molecules is beneficial because the two - center interference pattern gives another indicator of how the formulation is performing . in order to make such an assessment
, we will employ many - body perturbation theory along the lines of ref .
@xcite and the dipole moment in its length and velocity forms . for that purpose , we will consider two isoelectronic molecules : @xmath1 and co@xmath3 the homo , lumo and homo-1 in both molecules have similar geometry . the main difference between them is that @xmath1 is a homonuclear molecule , while co is heteronuclear .
this means that , for the latter molecule , there is an intrinsic static dipole moment .
according to ref .
@xcite , one expects the multielectron corrections to be more prominent in this case . this work is organized as follows . in sec .
[ theory ] , we provide the theoretical framework which we will employ in order to compute the high - order harmonic spectra .
we then calculate the transition amplitudes , in different formulations ( sec . [ sfa ] ) , within the strong field approximation , and calculate the multielectron corrections ( sec .
corrections ) , with particular focus on the form of the dipole operator .
subsequently , in sec .
[ results ] , these expressions are employed to compute the high - harmonic spectra . finally , the paper is summarized in sec .
[ concl ] .
atomic units are used throughout .
the sfa transition amplitude in the particular formulation of ref .
@xcite reads @xmath4+c.c . ,
\label{amplhhg}\end{aligned}\ ] ] with the action @xmath5^{2}d\tau -\epsilon _ { 0}(t - t^{\prime } ) + \omega t \label{actionhhg}\]]and the prefactors @xmath6 and @xmath7 in eqs .
( amplhhg ) and ( [ actionhhg ] ) , @xmath8 , @xmath9 , @xmath10 and @xmath11 denote the dipole operator , the laser - polarization vector , the ionization potential of the molecule in question , and the harmonic frequency , respectively , and @xmath12 gives the bound state with which the electron recombines , or from which it tunnels .
( [ amplhhg ] ) is given in the length - gauge formulation of the sfa .
we will employ this gauge throughout . in the standard velocity - gauge formulation for the molecular sfa
, the two - center interference patterns vanish @xcite .
to first approximation , we assume that there is one active electron , which tunnels from the highest occupied molecular orbital , reaches the continuum and recombines .
many - body corrections due to the presence of the other electrons will only be incorporated in the prefactors .
the above - stated expression describes a physical process in which an electron tunnels from its parent molecule at an instant @xmath13 , and propagates in the continuum from a time @xmath13 to a subsequent time @xmath14 with the intermediate momentum @xmath15 . at @xmath16
the electron recombines with its parent ion , generating a high - energy photon of frequency @xmath17 these steps are implicit in the action ( [ actionhhg ] ) .
they can also be directly identified in the saddle - point equations , obtained from the values of @xmath18 and @xmath15 which render the action stationary .
this implies @xmath19 and @xmath20 these saddle - point equations read @xmath21 ^{2}}{2}+\epsilon _ { 0}=0 , \label{tunnelsame}\]]@xmath22 = % \mathbf{0}. \label{return}\]]and @xmath23 ^{2}}{2}+\epsilon _ { 0}=\omega .
\label{recsame}\]]physically , eq . ( [ tunnelsame ] ) gives the kinetic energy of the electron during tunneling .
one should note that this equation has no real solution , so that the tunneling time @xmath13 is complex .
this is a consequence of the fact that tunneling has no classical counterpart .
( [ return ] ) constrains the momentum of the electron , so that it may return to the geometrical center of its parent molecule . finally , eq . ( [ recsame ] )
gives the conservation of energy upon recombination , in which the kinetic energy of the returning electron is converted in a high - frequency photon .
the stationary - phase method will be employed to compute the spectra in this work . for more details
we refer to @xcite .
we will now briefly discuss the many - body corrections derived in ref .
they have been obtained by starting from a many - body dipole operator in the recombination and ionization dipole matrix elements @xmath24 and @xmath25 , and applying mller plesset @xcite perturbation theory .
@xmath26 where the dipole matrix element associated with the homo is given by @xmath27the first - order corrections read @xmath28and @xmath29}}{\epsilon _
{ i}+\epsilon _ { j}-\epsilon _ { 0}-\epsilon _ { a}}+ \notag \\ & & \sum_{a}\langle \mathbf{k}|\psi _ { a}\rangle \sum_{b}\sum_{i}\mathbf{d}_{ib}% \frac{v_{ab[0j]}}{\epsilon _ { 0}+\epsilon _ { i}-\epsilon _ { a}-\epsilon _ { b } } , \label{corrections2}\end{aligned}\]]where @xmath30@xmath31}=\left\langle \psi _ { i},\psi _ { j}\right\vert v\left\vert \psi
_ { a},\psi _ { 0}\right\rangle -\left\langle \psi _ { j},\psi _ { i}\right\vert v\left\vert \psi _ { a},\psi _ { 0}\right\rangle , \]]and@xmath32}=\left\langle \psi _ { 0},\psi _ { j}\right\vert v\left\vert \psi _ { a},\psi _ { b}\right\rangle -\left\langle \psi _ { j},\psi _ { 0}\right\vert v\left\vert \psi _ { a},\psi _ { b}\right\rangle .\]]in the above - stated equations , @xmath33 denotes the electron - electron correlation , @xmath34 the orbital energies , and the index @xmath35 , with @xmath36 ( rec , ion ) , refers to recombination and ionization .
the indices @xmath37 represent unoccupied orbitals and the indices @xmath38 represent occupied orbitals .
the index 0 is related to the orbital around which the corrections are inserted . in the specific scenario
addressed in this paper , this is the highest occupied molecular orbital ( homo ) .
the indices @xmath39 in the dipole matrix element ( [ dipbound ] ) are general . in the first - order corrections ( [ corrections1 ] )
, they may relate to the expectation value of the dipole operator ( @xmath40 ) , or to the dipole coupling between the homo and an occupied bound state ( @xmath41 and @xmath42 ) . in the framework of the second - order corrections ( [ corrections2 ] )
, they relate to the couplings occupied and unoccupied bound states ( @xmath43 and @xmath44 or @xmath41 and @xmath45 ) .
physically , @xmath46}$ ] and @xmath47}$ ] give the exchange terms , and are dependent on the form of the electron - electron interaction .
in the corrections ( [ corrections1 ] ) and ( [ corrections2 ] ) , the terms @xmath48 will determine the two - center interference conditions , while the remaining terms will mainly act as weights .
these corrections will also behave differently with regard to homonuclear and heteronuclear molecules .
for the former type of molecules , the orbitals exhibit definite parities . hence , the dipole matrix elements @xmath49 will only couple gerade and ungerade states . as a direct consequence ,
the first term in ( [ corrections1 ] ) will vanish .
physically , this is expected , as there is no static dipole moment for a homonuclear molecule .
in contrast , for heteronuclear molecules , the matrix elements @xmath49 will in principle couple all states involved .
the first term in eq .
( [ corrections1 ] ) will be non - vanishing , and , physically , is the contribution of the static dipole moment of the molecule in question .
this term , however , will only lead to quantitative changes in the dipole matrix element .
in fact , an inspection of eqs .
( [ prefsinglee ] ) and ( corrections1 ) shows that it will yield the same two - center interference condition as the one - particle prefactor @xmath50 . in the above - mentioned framework
, all the structure of the molecule is embedded in the recombination prefactor @xmath51 this prefactor can be written in different forms , which will lead to different results . explicitly , the length , velocity and acceleration forms of the dipole operator read @xmath52@xmath53and @xmath54respectively , where the hats denote operators .
the prefactor @xmath55 will mainly influence the overall harmonic intensity , as it is associated with ionization . in this work
, we will consider the length and the velocity forms of the dipole operator . in the single - active - electron framework , the former and the latter lead to the worst and best description of the two - center interference patterns jmocl2007 . in calculating the molecular orbitals we use a linear combination of atomic orbitals ( lcao ) approximation along with the born oppenheimer approximation .
in this framework , the molecular orbital wavefunction is @xmath56where * r * is the internuclear separation , @xmath57 and @xmath58 the orbital and magnetic quantum number respectively .
the coefficients @xmath59 , with @xmath60 form the linear combination of atomic orbitals which are extracted from gamess - uk gamess , and the indices @xmath61 and @xmath62 refer to the left or to the right ion , respectively .
the internuclear axis is taken to be in the @xmath63 direction , and the laser - field polarization is chosen along the radial coordinate . for homonuclear molecules ,
@xmath64 and @xmath65 the parameter @xmath66 can be varied depending on the symmetry . for homonuclear molecules , @xmath67=@xmath68 and @xmath66=@xmath68 + 1 for gerade and ungerade symmetry , respectively .
the wavefunctions themselves are then expanded as gaussian type orbitals with a 6 - 31 basis set , with @xmath69for the @xmath70 @xmath71 and @xmath72 orbitals , @xmath73 and @xmath74 , respectively .
the index @xmath75 is related to the left or right ion , and @xmath76 and @xmath77 form the contraction coefficients and exponents , also obtained from gamess - uk .
note that these coefficients are real .
the momentum - space wavefunctions , which will be used in this work to compute the prefactors , are given by@xmath78c_{\alpha , j}^{(l)}\phi _
{ \alpha , j}^{(l)}(\mathbf{k } ) \\ & & + ( -1)^{l_{\alpha } -m_{\alpha } + \lambda _ { \alpha } } \exp [ -i\mathbf{k\cdot } % \frac{\mathbf{r}}{2}]c_{\alpha , j}^{(r)}\phi _ { \alpha , j}^{(r)}(\mathbf{k } ) , \notag \label{orbitalpspace}\end{aligned}\]]where , similarly to the position - space situation@xmath79with @xmath80.\]]therein , @xmath81 , @xmath82 and @xmath83 for the @xmath84 , @xmath71 and @xmath72 orbitals , respectively .
the return condition ( [ return ] ) guarantees that the momentum @xmath15 and the external field are collinear .
hence , for a linearly polarized field @xmath85 is equal to the alignment angle @xmath86 between the molecule and the field .
the above - stated equations have been derived under the assumption that only @xmath87 and @xmath88 states will be employed in order to build the wavefunctions employed in this work . for more general expressions
@xcite . in fig . 1
, we exhibit the homo for and co and @xmath1 obtained with gamess - uk [ figs . [ fig1].(a ) and
( b ) , respectively ] .
the position - space wavefunctions exhibit a central maxima and two minima located at the ion for both species .
the main difference is that , for co , there is a strong bias towards the carbon atom , while for @xmath1 the wavefunction is symmetric .
this is in agreement with the standard molecular - physics literature .
in contrast , we have found that both momentum - space wavefunctions , given in figs .
[ fig1].(c ) and ( d ) , are symmetric with respect to @xmath89 . in comparison to @xmath90 co
exhibits a much deeper maximum at vanishing momenta .
such maxima and minima may be determined by writing the exponents in eq .
( [ orbitalpspace ] ) in terms of trigonometric functions . in this case , one obtains@xmath91 + i\mathcal{c}_{-}^{(\alpha ) } \sin \left [ \frac{\mathbf{k\cdot r}}{2}\right ] , \]]with@xmath92calling @xmath93,$]we find @xmath94 .
\label{pspaceinterf}\]]eq .
( [ pspaceinterf ] ) exhibit minima for @xmath95 . note , however , that the coefficients @xmath96 defined in eq .
( [ coeffinterf ] ) depend on the wavefunctions at the left and right ions . since these wavefunctions themselves depend on the momentum @xmath97 , one expects the two - center patterns to be blurred for heteronuclear molecules . in contrast , for homonuclear molecules , @xmath98 and @xmath99 .
this implies that the momentum dependence in the argument @xmath100 cancels out and that the interference condition in refs .
@xcite is recovered . in this case , sharp interference fringes are expected to be present . the above - stated condition does not only lead to interference fringes in the bound - state momentum wavefunctions but also in the high - harmonic spectra .
this is due to the fact that the dipole matrix elements depend on the wavefunctions ( [ orbitalpspace ] ) . in this latter case , @xmath101 in eqs .
( [ orbitalpspace])-([pspaceinterf ] ) . and @xmath1 [ panels ( b ) and ( b ) ] , in the position and momentum space ( upper and lower panels , respectively).,width=359 ] in this section
, we will provide explicit expressions for the dipole matrix elements , depending on whether the dipole operator is given in its length or velocity form .
we will restrict ourselves to the first - order corrections to the single - active - electron prefactor , as they are expected to be dominant .
we will commence with the single - active - electron dipole prefactor ( prefsinglee ) , which is employed in the standard strong - field approximation . in the length form ,
this prefactor reads @xmath102where @xmath103 is given by eq .
( [ orbitalpspace ] ) . if explicitly written in terms of the wavefunctions centered at the left and right ions , @xmath104\partial
_ { \mathbf{k}}\phi _ { \alpha , j}^{(l)}(\mathbf{k } ) \notag \\ & & + ( -1)^{l_{\alpha } -m_{\alpha } + \lambda _ { \alpha } } c_{\alpha , j}^{(r)}\exp [ -i\mathbf{k}\cdot \frac{\mathbf{r}}{2}]\partial
_ { \mathbf{k}}\phi _ { \alpha , j}^{(r)}(\mathbf{k } ) \notag \\ & & + \frac{\mathbf{r}}{2}\phi _ { j}(\mathbf{k } ) , \label{alength}\end{aligned}\ ] ] with @xmath105c_{\alpha , j}^{(l)}\phi _ { \alpha , j}^{(l)}(\mathbf{k } ) \\ & & -(-1)^{l_{\alpha } -m_{\alpha } + \lambda _ { \alpha } } c_{\alpha , j}^{(r)}\exp [ -i\mathbf{k\cdot } \frac{\mathbf{r}}{2}]\phi _ { \alpha , j}^{(r)}(\mathbf{k } ) . \notag\end{aligned}\]]the
first two terms in eq .
( [ alength ] ) are mainly oscillating , and give the two - center interference maxima and minima . the third term in eq .
( [ alength ] ) increases with the internuclear distance , and has the main effect of blurring such patterns . in fact
, the existence of such term has caused considerable debate and has been attributed to the lack of orthogonality between the continuum and the bound - state wavefunctions in the sfa @xcite .
similar terms are also present in the first - order corrections , to be discussed later . in the velocity form , the single - electron dipole matrix element reads @xmath106\phi _ { \alpha , j}^{(l)}(\mathbf{k } ) \\ & & + ( -1)^{l_{\alpha } -m_{\alpha } + \lambda _ { \alpha } } c_{\alpha , j}^{(r)}% \mathbf{k}\exp [ -i\mathbf{k\cdot } \frac{\mathbf{r}}{2}]\phi _ { \alpha , j}^{(r)}(\mathbf{k } ) .
\notag\end{aligned}\]]in contrast to the length - form scenario , there is no term linearly growing with the internuclear separation . indeed ,
previous studies for simplified models for homonuclear molecules indicate that this is the preferred form of the dipole operator , in order to obtain the correct interference patterns @xcite .
we will now write down the explicit forms of eq .
( [ corrections1 ] ) for molecular orbitals in the lcao approximation .
according to ref .
santra , this term is responsible for the dominant corrections to the single - electron prefactor , and depends on the momentum space wavefunction ( [ orbitalpspace ] ) , and on the matrix element ( [ dipbound ] ) . the former is independent on the form of the dipole operator , while the latter is explicitly given by the sum @xmath108where@xmath109 ^{\ast } c_{\beta , \mu } ^{(\xi ) } \phi _ { \alpha , \nu } ^{\ast ( \xi ^{\prime } ) } ( \mathbf{r}+(-1)^{a}\mathbf{r}% /2 ) \notag \\ & & \times d(\mathbf{r})\phi _ { \beta , \mu } ^{(\xi ) } ( \mathbf{r}+(-1)^{b}% \mathbf{r}/2 ) .
\label{integralslcao}\end{aligned}\]]in eq .
( [ dipolebound ] ) and ( [ integralslcao ] ) , the indices @xmath110 are related to the right or left ions , the exponents @xmath37 are 1 and 0 for @xmath111 and @xmath112 respectively , and @xmath113 denotes the dipole operator in the position representation . in its length and velocity forms , @xmath114 and @xmath115.the coefficients @xmath116 are such that @xmath117 and @xmath118 in eq .
( [ integralslcao ] ) , one may recognize two types of contributions .
the direct integrals @xmath119 ^{\ast } c_{\beta , \mu } ^{(\xi ) } \phi _ { \alpha , \nu } ^{\ast ( \xi ) } ( \mathbf{r}\pm \mathbf{r}/2)d(\mathbf{r})\phi _ { \beta , \mu } ^{(\xi ) } ( % \mathbf{r}\pm \mathbf{r}/2 ) , \label{direct}\]]which contains one - electron wavefunctions centered at the same ion , and the so - called overlap " integrals @xmath120 ^{\ast } c_{\beta , \mu } ^{(\xi ) } \phi _ { \alpha , \nu } ^{\ast ( \xi ^{\prime } ) } ( \mathbf{r}\pm \mathbf{r}/2)d(% \mathbf{r})\phi _ { \beta , \mu } ^{(\xi ) } ( \mathbf{r}\mp \mathbf{r}/2 ) , \label{overlap}\]]for @xmath121 with one - electron wavefunctions localized at different ions .
for the moment ignoring the overlap integrals , @xmath122 ^{\ast } c_{\beta , \mu } ^{(l)}\mathcal{i}_{l , l}^{\alpha , \beta } \label{dipmolecule } \\ & & + ( -1)^{l_{\beta } + \lambda _ { \beta } + l_{\alpha } + \lambda _ { \alpha } } \left [ c_{\alpha , \nu } ^{(r)}\right ] ^{\ast } c_{\beta , \mu } ^{(r)}\mathcal{i}% _ { r , r}^{\alpha , \beta } .
\notag\end{aligned}\]]the explicit expressions for the overlap integrals are provided in appendix 1 .
we will commence by providing expressions for ( [ dipolebound ] ) using the velocity form .
for that purpose , one must solve the integrals ( integralslcao ) . depending on whether the dipole operator couples only @xmath123 or @xmath124 orbitals , or a @xmath84 orbital with a @xmath124 orbital , these integrals will be different .
if only @xmath84 orbitals are involved,@xmath125with @xmath126 or @xmath127 and@xmath128 ( \zeta _ { j,\nu } ^{(\xi ) } + \zeta _ { j^{\prime } , \mu } ^{(\xi ^{\prime } ) } ) ^{-1/2-l/2}\gamma \left [ \frac{1+l}{2}\right ] .
\label{f}\]]the same expression is encountered if only @xmath71 or @xmath72 are coupled , i.e. , @xmath129 the above - stated expression is only non - vanishing if @xmath130 is even . in our framework , this implies that only @xmath87 and @xmath88 states are mixed .
if , on the other hand , a @xmath84 orbital is coupled to a @xmath71 or a @xmath72 orbital , this integral reads @xmath131with @xmath132 thereby , @xmath133@xmath134and @xmath135 is defined according to eq .
( [ f ] ) .
( [ integralsigmpi ] ) also holds for the dipole matric elements coupling the @xmath71 and @xmath72 orbitals , i ..
e , @xmath136 from eqs .
( [ pisigm1 ] ) and ( [ pisigm2 ] ) we see that @xmath137 if @xmath138 and @xmath139 and that @xmath140 @xmath141if @xmath142 and @xmath143 hence , once more , only @xmath87 and @xmath88 states are coupled .
all these contributions are then included in the corrections ( corrections1 ) . in the specific case of a homonuclear molecule , an inspection of eqs .
( [ dipmolecule])-([pisigm2 ] ) shows that only orbitals of different parities are coupled .
in fact , if @xmath144 and @xmath145 , eq .
( [ dipmolecule ] ) is only nonvanishing if @xmath146 is even .
since , as discussed above , @xmath147 are only nonvanishing if @xmath57 and @xmath148 have values corresponding to different parities , this must also hold for @xmath149 and @xmath66 .
in contrast , for heteronuclear molecules , these coefficients are different and in principle all terms contribute .
we will now state the length - form dipole matrix element .
we will mainly consider the integrals @xmath150 if only @xmath123 orbitals are coupled , these integrals read@xmath151 , \notag \label{lengthsigmsigm}\end{aligned}\]]the same expression holds for transitions involving only the @xmath71 or the @xmath72 orbitals , i.e. , @xmath152 .
if the dipole operator couples @xmath84 with @xmath124 orbitals , such integrals are given by @xmath153 , \notag \label{lengthsigmpi}\end{aligned}\]]with @xmath132 eq .
( [ lengthsigmpi ] ) also holds for couplings between different @xmath124 orbitals , i.e. , @xmath154 in the above - stated expressions , one may identify two distinct behaviors . the terms @xmath155 @xmath156 and @xmath157 are non - vanishing only if @xmath57 and @xmath148 have values corresponding to different parities , i.e. , in our framework , it couples @xmath87 and @xmath88 states .
the terms @xmath158 and @xmath159 on the other hand , couple states of the same parity ( either @xmath87 or @xmath88 ) . for homonuclear molecules ,
these terms will vanish , as the dipole operator will only couple states of different parities , i.e. , @xmath160 must be even in eq .
( [ dipmolecule ] ) .
for heteronuclear molecules , in contrast , they may in principle be present as the coefficients at each center of the molecule are different .
we will now discuss some of the subtleties that arise in the evaluation of our multielectron corrections .
firstly , the overlap integrals , in eq .
( overlap ) are often presumed to be very small in comparison to their direct counterpart , but this is not always the case .
this can be seen by using the gaussian theorem , @xmath161 where @xmath162 @xmath163 here @xmath164 and @xmath165 correspond to our exponents in eq .
( [ contraction ] ) . at first sight
it would appear that these contributions are much smaller than those from the direct integrals ( [ direct ] ) .
this does not hold , however , for exponents that are less than unity , of which there are many for various different atomic orbitals .
it is therefore not astute to ignore the overlap integrals in this instance .
in fact , we have verified that in most cases the overlap integrals ( overlap ) are comparable to the integrals ( [ direct ] ) at the same center .
more importantly , we have also found that the velocity form gives zero static dipole moment for heteronuclear molecules when using a gaussian basis set .
this is an unphysical result , which has been verified for the direct integrals by using the properties of gamma functions to analyze the term in brackets in eq .
( [ velocitysigmasigma ] ) .
explicitly , it can be shown that this term is proportional to @xmath166 $ ] . since only @xmath87 and
@xmath88 states can be coupled , each pair @xmath167 and @xmath168 will lead to symmetric contributions which will cancel out .
we have also verified numerically that the overlap integrals are vanishingly small in this case .
apart from that , inclusion of the overlap integrals gives rise to terms that depend on r , and thus removes the translational invariance from the velocity form .
this problem is also encountered in the single - active electron matrix element when using the length form .
in contrast to the one - electron length - form scenario , however , these terms will vanish at large internuclear distances due to the gaussian exponentials ( [ gaussianoverl ] ) .
( see a discussion in @xcite for their counterpart in the single - active electron framework ) in contrast , the length form predicts non - zero static dipole moment in heteronuclear molecules , which makes sense physically .
however , for the heteronuclear case the r - dependent terms that arise in the direct integrals couple states of the same parity as seen in eq .
( [ lengthsigmsigm ] ) and eq .
( lengthsigmpi ) .
this should not occur for an odd operator such as the dipole , which is an unphysical result .
this is another reason to suggest that these terms should be neglected . in the results that follows ,
this term will be neglected unless otherwise stated .
below we compare the harmonic spectra predicted by different forms of the dipole operator .
we will place particular emphasis on the comparison between homonuclear and heteronuclear molecules .
the former and the latter will be exemplified by @xmath1 and co , respectively .
we will consider the full , three - dimensional dynamics of the problem .
for that purpose , one must integrate over the azimuthal coordinate , @xmath169 , such that , @xmath170where @xmath171 is the overall transition amplitude .
this allows one to consider the degeneracy of the @xmath124 orbitals @xcite .
the corrections are incorporated in the ionization and recombination prefactors .
the former gives the particular weight for tunneling to take place , and the latter gives the shape of the high - order harmonic spectra and the two - center fringes .
the explicit expressions for the dipole matrix elements are provided in appendix 2 .
subject to a linearly polarized laser field of frequency @xmath172 a.u . and intensity @xmath173 , as a function of the alignment angle @xmath174 between the molecule and the field .
in the figure we consider the corrections in both the ionization and the recombination prefactors . panels ( a ) and ( c ) have been computed employing the length form , while panels ( b ) and ( d ) have been calculated using the velocity form . the upper panels display the @xmath175 spectrum in which the first - order corrections have been incorporated . in the lower panels ,
we depict the yield obtained with the first order corrections in the length and the velocity form .
the bound - state energy of the homo and the equilibrium internuclear distance have been taken from gamess - uk .
explicitly , @xmath176 a.u and @xmath177 a.u .
the contour plots have been normalized with regard to the maximum yield in each panel.,width=359 ] we will now start by discussing the high - harmonic spectra of diatomic nitrogen ( @xmath0 ) computed including the single - atom prefactor and the first - order corrections . in fig .
[ n2ionrec ] , we display such spectra , in the velocity and length form ( panels ( a ) and ( b ) , respectively ) , together with the corrections alone ( panels ( c ) and ( d ) , respectively ) . in the length - form prefactor @xmath178
, we omitted the term growing linearly with the internuclear distance @xmath179 . in the corrections , we included both the direct and the overlap integrals . in the figure , we notice a clear two - center minimum , which is due to the interference of high - harmonic emission at the two spatially separated centers in the molecule .
this minimum varies with the alignment angle @xmath180 of the molecule with respect to the laser - field polarization . in the upper panels , its energy position is characteristic of the @xmath181 orbital @xcite .
a comparison of panels ( a ) and ( b ) shows that the two - center minimum is slightly displaced , depending on the form of the dipole operator .
for instance , for vanishing alignment angle , in the velocity form this minimum is near @xmath182 , while in the length form it is at a slightly higher energy ( @xmath183 ) .
this is due to the fact that the @xmath184 mixing , which determines the energy position of this minimum , is different for each case .
the velocity form slightly favors the contributions from the @xmath88 states , compared to the length form . for a discussion of the @xmath184 mixing in a slightly different context , namely of different basis sets , and its influence on the high - order harmonic spectra ,
see @xcite . here ,
however , this bias is introduced by the different forms of the dipole operator .
the remaining panels exhibit the spectra obtained employing the first - order corrections only .
such contributions exhibit a well - defined minimum , whose position is at a quite different energy from that observed in the overall prefactor .
we have verified that this pattern can be related to the contributions of the @xmath185 orbitals .
the contributions of the @xmath186 orbitals are strongly suppressed in this parameter region due to to the presence of a nodal plane along the internuclear axis .
quantitatively , the yield in panels ( c ) and ( d ) is several orders of magnitude smaller than those in the upper panels .
this shows that , in the present framework , it suffices to consider the zeroth order contributions .
this behavior occurs both for the length and the velocity form of the dipole operator , and is due to the fact that the corrections are considerably smaller both in the ionization and recombination prefactor .
the above - mentioned suppression of ionization for the @xmath187 orbitals also contributes to such results .
the influence of the ionization prefactor can be seen in fig .
[ cutsover ] .
therein , we present spectra computed for a fixed alignment angle assuming @xmath188 to be constant and equal to unity .
physically , this implies that any influence of the orbital symmetry on tunneling ionization , such as the presence or absence of nodal planes , has been removed , both for the zeroth - order prefactor and the corrections . for comparison
, we are also providing the full spectra , with both the ionization and recombination prefactors .
the figure shows that , while the overall spectra changes relatively little , or may even increase if this prefactor is removed ( see orange and black lines in fig .
[ cutsover].(a ) ) , the corrections increase in several orders of magnitude .
this can be attributed to the following . in the parameter range of interest , the zeroth order ionization prefactor , which is the main influence in the overall spectra , is of the order of unity or even slightly larger for a @xmath181 orbital .
this is due to the absence of nodal planes for either parallel or perpendicular alignment in this particular case .
in contrast , in the corrections , mainly @xmath185 or @xmath186 orbitals are involved . the former orbitals exhibit nodal planes for alignment angles @xmath189 and the latter for @xmath190 .
this will cause an overall suppression in tunnel ionization for a wide range of angles , when such orbitals are summed over .
furthermore , the dipole matrix elements @xmath191 in the spectra are quite small and will also contribute to the above - mentioned suppression . in a driving field with the same parameters as in the previous figure , and an alignment angle @xmath192 .
the figure shows the overall spectra together with the first - order corrections taking into account only the direct or overlap integrals .
we consider both the ionization and recombination prefactors , or the recombination prefactor only ( orange and red lines in the figure ) .
panels ( a ) and ( b ) give the velocity and length forms of the dipole operator , respectively.,width=340 ] another noteworthy feature also presented in fig .
[ cutsover ] is the influence of the so - called overlap integrals , in which the dipole operator couples wavefunctions at different centers in the molecule , in the multielectron corrections . in general
, these integrals are assumed to be very small in comparison with the direct integrals , in which wavefunctions at the same ion are taken .
hence , in many cases the overlap integrals are neglected .
we have verified , however , that this depends strongly on the form of the dipole operator .
for the velocity - form dipole , the spectra obtained from the overlap integrals are comparable to those obtained from the direct integrals only [ fig . [ cutsover].(a ) ] .
hence , both contributions must be included .
in contrast , if the length form is considered , the contributions from the overlap integrals are considerably smaller than their direct - integral counterparts [ fig . [ cutsover].(b ) ] .
therefore , the former may be neglected .
next , we will discuss the spectra encountered for co. the overall spectra are depicted in fig .
[ figco ] , together with the first - order corrections ( upper and lower panels , respectively ) . as an overall feature , the spectra exhibit no perceivable two - center interference minimum , as co is a heteronuclear molecule .
this means that , upon recombination , there will be more high - order harmonic emission from a specific spatial region , where the orbital with which the electron recombines , i.e. , the homo , is more localized .
this occurs both for the velocity and length forms ( figs .
[ figco].(a ) and ( b ) , respectively ) .
one notices , however , that the decrease in the overall yield as the molecular alignment angle approaches @xmath189 remains , due to the fact that the homo is a @xmath193 orbital , and thus highly localized along the internuclear axis . as the laser - field polarization direction moves away from this axis , recombination , and hence high - order harmonic emission becomes less and less probable .
due , however , to the heteronuclear nature of this molecule , this suppression occurs at a different angle compared to @xmath0 .
once more , we find that the contributions of the first - order multielectron corrections to the spectra are several orders of magnitude smaller than the zeroth - order terms .
these contributions are depicted in fig .
[ figco].(c ) and ( d ) , for the length and velocity form , respectively . for such corrections , the interference minimum due to high - harmonic emission at different centers has been washed out .
this minimum was very clear for @xmath0 ( see figs .
[ n2ionrec].(c ) and ( d ) for comparison ) .
there is , however , a very pronounced minimum near @xmath194 for the velocity - form corrections , which is independent of the harmonic frequency ( see fig .
[ figco].(d ) ) . this minimum is due to the geometry of the three lower - lying @xmath193 orbitals , which influence the corrections .
this suggests that the imprints on the high - order harmonic spectra due to purely geometric features , such as nodal planes or minima in the orbitals , are more robust than those related to quantum interference between harmonics emitted as spatially separated centers .
the length - form corrections , in contrast , do not exhibit any such feature .
in fact , we have verified that they are dominated by the static dipole moment , which behaves in the same way as the homo .
below , in fig .
[ cutsco ] , we perform a more detailed analysis of the above - mentioned issue . for that purpose , we consider the contributions of individual orbitals to the corrections , and their influence on the harmonic spectra , for fixed harmonic order .
we consider a mid - plateau harmonic , namely @xmath195 .
these results are displayed for both @xmath0 and co , as functions of the alignment angle ( upper and lower panels in the figure , respectively ) . as an overall feature
, we observe that , for a homonuclear molecule , all contributions are symmetric upon @xmath196 ( see panels(a ) and ( b ) ) .
physically , this is expected , as such molecules exhibit inversion symmetry and a rotation of @xmath187 should not alter the physics of the problem .
clearly , this no longer holds for a heteronuclear molecule . in this latter case , however , the contribution of @xmath197 to the yield is the mirror image of @xmath198 .
this is clear , as in this case the right and the left ions are reversed .
harmonic of a field with the same parameters as in the previous figures , as functions of the alignment angle @xmath180 of the molecule relative to the laser - field polarization .
the thick black lines in the panel give the overall contributions of all @xmath193 orbitals and the orange stars the contributions of the @xmath187 orbitals , respectively .
panels ( a ) and ( b ) refer to @xmath0 , and panels ( c ) and ( d ) correspond to co , respectively .
the left and right panels display the results using the length and velocity form of the dipole operator , respectively .
the minima due to the two - center interference are indicated by green arrows in the figure .
, width=359 ] the individual contributions provide useful information about the geometry of the orbitals involved in the corrections .
this can be seen in all cases , regardless of whether the molecule is heteronuclear or homonuclear .
for instance , the contributions from the @xmath187 orbitals , shown as the orange symbols in the figure , decay very abruptly for @xmath199 .
this is due to the fact that such orbitals exhibit nodal planes in the case of @xmath0 and a feature similar to nodal planes in the case of co. the main difference between the homonuclear and the heteronuclear case is the two - center interference minimum , which is absent in the heteronuclear case .
furthermore , the shapes of such distributions do not change with the form of the dipole operator .
any changes are mostly quantitative ( for a comparison , see panels ( a ) and ( b ) for @xmath0 and panels ( c ) and ( d ) for co )
. the contributions from the @xmath193 orbitals , on the other hand , can behave in very different ways depending on whether the velocity or the length form is taken , in the heteronuclear case , such as co. for @xmath0 , in contrast , different forms of the dipole operator will mainly lead to a different weighting of the contributions from individual orbitals to the overall corrections .
for instance , in the length - form case , the contributions of the @xmath200 orbital are prevalent , while if the velocity form is taken , the contributions from both @xmath185 orbitals are comparable .
another interesting feature is that both forms lead to a sharp harmonic suppression near @xmath201 .
this is due to the fact that the nodal planes of the @xmath202 and @xmath200 orbitals are parallel to the laser - field polarization for such angles . in co , however , we notice two interesting features .
firstly , the nodes disappear if the length form is taken .
this is partly due to the static dipole moment , which is only present in the length form and leads to corrections which behave like the homo .
this implies that the overall dipole will decrease for @xmath203 as shown in the figure .
in contrast , in the velocity form , as this term is absent , there are minima near @xmath192 .
these minima come from the geometry of @xmath204 and @xmath205 , which are the heteronuclear counterparts of @xmath202 and @xmath200 , leading to a strong suppression in the yield close to @xmath201 .
they are , however , slightly shifted from these values due to the heteronuclear character of the molecule .
the remaining @xmath193 orbitals shift these minima even further , to @xmath206 .
in conclusion , our results show that an adequate choice of the dipole operator in order to model high - order harmonic generation is still an open question . in the single - active - electron approximation ,
enough evidence has been provided in @xcite that the velocity form yields the best agreement with the double - slit physical picture .
if one goes beyond this approximation , however , the velocity form leads to a vanishing static dipole moment for heteronuclear molecules .
this may be a particularity of the basis set involved .
another issue in the velocity form are the so - called overlap integrals , in which the dipole operator couples wavefunctions localized at different ions in the molecule . throughout the literature
, it has been argued that such integrals are small and may be neglected without loss of information .
our results , however , indicate that , depending on the basis set used , their outcome may be comparable to that of the direct integrals , in which wavefunctions at the same ion are coupled . as the internuclear distance increases ,
however , the overlap integrals become vanishing small . in contrast
, the length - form of the dipole operator leads to a non - vanishing static dipole moment for a heteronuclear molecule , and very small overlap integrals , in comparison to their direct counterparts .
the main problem , however , is that such a formulation leads to terms in the dipole - matrix elements with increase indefinitely with the internuclear separation .
they also couple states with the same parity , which is unphysical for an odd operator such as the dipole .
such terms have been also found in the single - active electron context @xcite .
hence , it seems that the velocity form still gives better results .
furthermore , suppression of the harmonic spectra caused by the presence of nodes in the orbitals is far more robust than the maxima and minima coming from two - center interference .
interestingly , for heteronuclear molecules , despite interference minima being washed out , the high - harmonic suppression caused by the geometry of the orbitals is still present in the spectra .
however , compared to their homonuclear counterpart , this suppression occurs at different alignment angles .
finally , in order to incorporate multielectron effects in sfa - like models in an appropriate way , one must consider the dynamics of the target .
concrete examples are excitation , relaxation , electron orbits starting or finishing at different molecular orbitals , or the motion of the bound electrons .
this paper has shown that , if multielectron effects are considered statically by employing many - body perturbation theory around the single - electron strong - field approximation , they lead to corrections which are several orders of magnitude smaller than the zeroth - order spectrum .
this holds for both homonuclear and heteronuclear molecules .
in fact , recent results in which excitation or relaxation have been incorporated numerically exhibited larger corrections @xcite .
* acknowledgements : * we profited from several discussions with h. j. j. van dam , p. j. durham , p. sherwood and in particular j. tennyson .
this work has been funded by the uk epsrc and by the daresbury laboratory .
in this appendix , we are providing the generalized expressions for the overlap integrals ( [ overlap ] ) considering the dipole operator in its velocity and length forms .
we will commence by the velocity - form expressions .
if only @xmath84 orbitals are coupled , one may show that @xmath207 , \notag \label{explicitoverlap}\end{aligned}\]]where @xmath208 ^{l_{\beta } } % \left [ u\mp \frac{\rho _ { j , j^{\prime } , \nu , \mu } ^{(\xi ^{\prime } ) } } { 2}r% \right ] ^{l\alpha } \notag \\ & & \times \exp [ -(\zeta _ { j,\nu } ^{(\xi ) } + \zeta _ { j^{\prime } , \mu } ^{(\xi ^{\prime } ) } ) u^{2}]du,\end{aligned}\]]with@xmath209@xmath210and @xmath211 for @xmath212 if , on the other hand , we consider a transition from a @xmath84 to a @xmath213 or @xmath72 orbital , @xmath214 \right ] , \end{aligned}\ ] ] where @xmath215 the function @xmath216 is defined in eq .
( [ f ] ) .
if only @xmath71 or @xmath72 orbitals are coupled , the overlap integrals read@xmath217 \right ] , \end{aligned}\ ] ] with @xmath218 or @xmath219 finally , if we consider the overlap integral coupling different @xmath124 orbitals , we will find@xmath220with @xmath221and @xmath222 in eq .
( overlaplength ) , @xmath223 the function @xmath224 is defined in eq . ( [ pisigm1 ] ) . for the length form of the dipole operator ,
if only transitions involving @xmath123 orbitals are considered , we find that@xmath225where @xmath226 ^{l_{\beta } } \hspace*{-0.15cm}\left [ u\mp \frac{\rho _ { j , j^{\prime } , \nu , \mu } ^{(\xi ^{\prime } ) } } { 2}r\right ] ^{l\alpha } \\ \hspace*{-0.15cm}&\hspace*{-0.15cm}&\hspace*{-0.15cm}\times \left [ u\mp \frac{\zeta _ { j,\nu } ^{(\xi ) } -\zeta _ { j^{\prime } , \mu } ^{(\xi ^{\prime } ) } } { % 2(\zeta _ { j,\nu } ^{(\xi ) } + \zeta _ { j^{\prime } , \mu } ^{(\xi ^{\prime } ) } ) } r% \right ] \exp [ -(\zeta _ { j,\nu } ^{(\xi ) } + \zeta _ { j^{\prime } , \mu } ^{(\xi ^{\prime } ) } ) u^{2}]du .
\notag\end{aligned}\]]for transitions from a @xmath84 to a @xmath71 or @xmath72 orbital , the overlap integral reads@xmath227\mathcal{j}(0,l_{\beta } ) + % \mathcal{f}[l_{\alpha } ] \mathcal{k}(0,l_{\beta } ) \right\ } , \notag\end{aligned}\]]with @xmath218 or @xmath219 finally , if only @xmath124 orbitals are involved , one may distinguish two types of overlap integrals . either the dipole matrix element couples the same initial and final orbital , @xmath71 or @xmath228or the orbitals @xmath229 and @xmath230 @xmath231 the former case is given by @xmath232% \mathcal{j}(0,0)+\mathcal{f}[l_{\alpha } + l_{\beta } ] \mathcal{k}(0,0)\right\ } , \notag\end{aligned}\]]while the latter reads@xmath233with @xmath234 @xmath235 specifically for the basis set employed in this work , we find@xmath238 @xmath239 @xmath240 , \ ] ] @xmath241\ ] ] and@xmath242 .
\notag\end{aligned}\]]for the integrals @xmath243 and @xmath244 and @xmath245 @xmath246 , \ ] ] @xmath247\ ] ] and @xmath248\]]for the integrals @xmath249 .
the above - stated expressions show that the integrals @xmath250 and @xmath249 will lead to a polynomial dependence on @xmath251 .
this dependence will be damped by the gaussian factor in eq .
( [ explicitoverlap ] ) , so that the overlap integrals are expected to decrease with the internuclear distance .
in this appendix , we provide the expressions obtained for the dipole matrix elements , if the azimuthal angular variable is integrated over .
let us start with @xmath1 . in this case
, the homo is a @xmath252 orbital .
the orbitals that will contribute to the corrections are @xmath253 , @xmath254 , @xmath255 and 1@xmath256 .
the remaining , @xmath257 orbitals do not contribute to the corrections for symmetry reasons .
we will be able then to write for the first - order dipole matrix element , where @xmath259 and @xmath260 in the above - stated equations , @xmath261 and @xmath132 with respect to the dependence on the azimuthal coordinate @xmath262 , eq .
( 1stordern2 ) can be written as @xmath263 \notag \\ & & + \mathcal{d}_{\pi _ { y}}(k,\tau , \theta _
{ k})\sin [ \phi _ { k}],\end{aligned}\]]where @xmath264 indicates the part of the prefactors which do not depend on the azimuthal coordinate @xmath262 .
the modulus square of the product @xmath265 ^{\ast } d^{(1)}(\mathbf{k}+\mathbf{a}(t^{\prime } ) , $ ] when integrated over the azimuthal coordinate , leads to the effective prefactor @xmath266where the constants arise from integrals of the form @xmath267\cos ^{n}[\phi _ { k}]d\phi _ { k},$ ] where @xmath268 are even integer numbers .
the explicit expressions for the functions @xmath269 read @xmath270 for co , there will be two main differences .
firstly , a static dipole moment is expected .
secondly , due to the fact that the orbitals do not exhibit a well - defined parity , more states are coupled by the dipole operator through the matrix elements @xmath275 . in fact , apart from the homo , there will now be four contributing occupied @xmath193 states and the two degenerate @xmath187 states corresponding to the homo-1 .
the explicit expressions for @xmath276 will change to @xmath277 with @xmath278 this will lead to changes in @xmath279 . furthermore , in eq .
( [ dipolesigmapi ] ) @xmath181 should now be replaced by @xmath280 and @xmath281 by @xmath282 .
there will be , however , no change in the structure of the subsequent equations or in the weights obtained upon angular integration .
m. lein , n. hay , r. velotta , j. p. marangos , and p. l. knight , phys .
* 88 * , 183903 ( 2002 ) ; phys .
a * 66 * , 023805 ( 2002 ) ; m. spanner , o. smirnova , p. b. corkum and m. y. ivanov , j. phys .
b * * 37 * * , l243 ( 2004 ) .
x. chu and s.i .
chu , phys .
a * 64 * , 063404 ( 2001 ) ; dmitry a. telnov and shih - i chu , phys .
a * 80 * , 043412 ( 2009 ) ; xi chu and shih - i chu ; phys .
a * 70 * , 061402 ( 2004 ) ; sang - kil son and shih - i chu , phys .
a * 80 * , 011403 ( 2009 ) .
gamess - uk is a package of ab initio programs .
see : `` http://www.cfs.dl.ac.uk/gamess-uk/index.shtml '' , m.f .
guest , i. j. bush , h.j.j .
van dam , p. sherwood , j.m.h .
thomas , j.h .
van lenthe , r.w.a havenith , j. kendrick , mol . phys . * 103 * , 719 ( 2005 ) .
s. patchkovskii , z. zhao , t. brabec and d. m. villeneuve , phys .
. lett . * 97 * , 123003 ( 2006 ) ; s. patchkovskii , z. zhao , t. brabec , and d.m .
villeneuve , j. chem .
phys . * 126 * , 114306 ( 2007 ) ; o. smirnova , s. patchkovskii , y. mairesse , n. dudovich , d. villeneuve , p. corkum , and m. yu .
ivanov , phys .
* 102 * , 063601 ( 2009 ) . | we make a detailed assessment of which form of the dipole operator to use in calculating high order harmonic generation within the framework of the strong field approximation , and look specifically at the role the form plays in the inclusion of multielectron effects perturbatively with regard to the contributions of the highest occupied molecular orbital .
we focus on how these corrections affect the high - order harmonic spectra from aligned homonuclear and heteronuclear molecules , exemplified by @xmath0 and co , respectively , which are isoelectronic .
we find that the velocity form incorrectly finds zero static dipole moment in heteronuclear molecules .
in contrast , the length form of the dipole operator leads to the physically expected non - vanishing expectation value for the dipole operator in this case .
furthermore , the so called `` overlap '' integrals , in which the dipole matrix element is computed using wavefunctions at different centers in the molecule , are prominent in the first - order multielectron corrections for the velocity form , and should not be ignored .
finally , inclusion of the multielectron corrections has very little effect on the spectrum .
this suggests that relaxation , excitation and the dynamic motion of the core are important in order to describe multielectron effects in molecular high - order high harmonic generation . |
many species of animal live in group and are capable of making decisions while maintaining group coherence @xcite .
animal living in groups are constrained by what they can perceive thus most of the time they can not have a global view of the group and their environment .
yet those species are capable of remarkable group behaviours .
the question arises as to understand what kind of mechanism allows the group to perform so well given their individual limitations .
several forms of social structures exists that go from all individuals apparently having the same weight in the group to various forms of leadership and social hierarchy . + among the many group - living species , the chicken ( _ gallus gallus domesticus _ ) is an interesting animal model : the chickens are social animals and chicks present a social attachment to their siblings and to their mother .
indeed , in precocial birds such as chickens , the formation of social attachments occurs in the first days after hatching .
this process is used as a model for the study of learning @xcite because the chicks environment and conditions prior to hatching can be controlled and standardized .
hence , the period immediately after hatching is likely to be particularly useful for studying social motivation and learning . +
this social structure allows to address the question of the role of leadership in group decision - making .
chicks are attracted to each other and at the same time they tend to remain in a relatively close vicinity to their mother that they can also follow as a group .
the question arises as to what makes the group coherent and what is the importance of the leader , the hen , in group decision making and group coherence .
the development of filial behaviour can be separated in terms of two interacting processes : filial motivation and filial imprinting .
filial motivation causes chicks to approach and follow conspicuous objects with certain general properties like colours , shapes , sizes , and movement patterns . in a natural environment , chicks are attracted toward the hen and the other chicks of the group .
filial imprinting is a learning process through which the chick comes to restrict this behaviour to focus to particular stimuli @xcite .
as chicks become progressively more familiar with an object , they tend to approach this object more @xcite .
thus , filial motivation determines the propensity to make social attachments , while imprinting establishes preference .
+ in ethological studies one of the long - standing interests is to understand social communication , relationships and structures . until recently ,
to study these mechanisms researchers had used simple specially designed mock - ups whose behaviour can be controlled so as to trigger a response of the animals . but nowadays availability of low - cost miniaturized computer chips , motors and sensors allowed these artificial models to become sophisticated and reliable robotic devices that can be used to test hypotheses @xcite .
for instance , robots were used to study male territorial instinct in dart - poison frogs @xcite and mate selection in tungara frogs @xcite , to test ideas about nest mate recognition in brush turkeys @xcite , and the predator avoidance by ground squirrels @xcite .
a number of recent works in ethology have successfully used robots to investigate individual and collective animal behaviours , in particular by creating mixed robot - animals societies : robots were mixed with chicks @xcite , cockroaches @xcite , fruit flies @xcite , killifish @xcite and zebrafish @xcite .
+ here , we address these questions by using a robot ( cf . fig .
[ fig1 ] ) that will become a surrogate mother to individual chicks . to build a social attachment between the chicks and the robot we use the filial imprinting mechanism .
after hatching a critical period exists where the chick will learn to build social attachment to specific objects .
classical ethology has shown that this special form of learning works for any object remaining close to the chicks and presenting specific colour , movement and sound patterns . in this study
, we develop an imprinting methodology involving autonomous robots to study individual behaviour of free moving animals .
thanks to this methodology we are able to quantify the individual behaviours and social responses to the robot .
we also develop automated experimental methodologies that allow to make fast and efficient data analysis , and compare to existing ones @xcite .
a workflow of this methodology is found in fig .
the analysis is based on high throughput data allowing fine quantification of individual behavioural traits .
thanks to this method we can make quantitative assessment of individual variability of a large number of individuals .
+ omnidirectional camera that allows the robot to perceive visually its surrounding .
the small rectangular holes allow infra - red sensors to detect objects or animals in close vicinity.,width=313 ] we investigate whether the imprinting quality of chicks can be influenced by social bonding . after the imprinting process , the chicks are randomly put in groups of 6 individuals for 20 days .
we then perform a new quantification of individual behaviour traits , and compare them to original results .
additional experiments include experiments during the period of time the chicks are put into groups , where entire groups of chicks interact with a robot , but the associated results are not investigated in this article . here
, we quantify the individual characteristics for 205 animals .
the results show a large variability of behavioural patterns .
one of the main variability is related to the success of imprinting .
three types of chicks are observed : imprinted , indifferent and avoiders . among the successfully imprinted chicks we still observe a large variability of behavioural patterns .
finally we show that the use of autonomous robots and automated ethograms allow detailed quantitative studies of animal behavioural patterns as it has also been shown with drosophila flies and bees @xcite .
the first step in the behavioural analysis is to classify the chicks according to the success of the filial imprinting on the robot .
filial imprinting corresponds to the propensity of the chicks to be attracted by the robot and to follow it as if it was the hen mother .
we observe two different kind of non - imprinted chicks : indifferent chicks that ignore the presence and movement of the robot , and avoiders that run along the walls , possibly running away from the robot .
imprinting success is assessed by looking at the behavioural response of the chicks towards the robot . + figure [ fig3 ] shows both the speed of three different chicks depending of their spatial position in the arena , and the robot trajectories .
the figure shows the three typical behaviour responses of a chick to the robot : a chick can be imprinted ( panel a ) , indifferent ( panel b ) , or avoider ( panel c ) .
the distributions of chicks speed , their distance toward the robot , and their walking behaviour ( whether they are walking or stopping ) , for 30 individuals , is shown in fig .
[ fig4 ] . imprinted and avoider chicks tend to alternate a walking behaviour with small stops .
indifferent chicks stay in place in the arena , with very few apparent walking behaviour .
imprinted chicks tend to be close to the robot .
indifferent and avoider chicks tend to be distant from the robot .
the three typical behaviour responses of a chick to the robot correspond to different distributions of chick speeds , as shown in fig .
[ fig5 ] : imprinted chicks move with variable speed , indifferent chicks do not move much , avoider chicks tend to move at higher speed than imprinted chicks .
+ the classification of the behavioural response of each chick toward the robot can be done by a human observer looking at all the videos . however , this represents a tedious and long qualitative work that would greatly benefit from automation .
only @xmath0 of all experiments ( 64 individuals ) is classified by a human observer .
the rest of the experiments is classified automatically , using an algorithm to classify the data automatically extracted from the videos by image analysis .
+ because it is possible for a human observer to classify with good accuracy the type of chick , we select a supervised learning method .
we use a linear support vector machine ( svm ) classifier , an algorithm often used to solve problems in classification , regression , and novelty detection @xcite .
this method is first trained on the human - classified dataset , separating experiments into three classes : imprinted , avoider , indifferent .
the classification process used the following features : _ mean distance between a robot and a chick _ , _ mean speed of a chick _ , and its _ standard deviation_. the results of the classification are presented in fig . [ fig6 ] .
panel a shows the training dataset composed of @xmath0 of the all data ( 64 individuals ) .
the green points correspond to the imprinted chicks , the blue points to the indifferent chicks , and red points to the avoider chicks .
the experiments have a duration of one hour , and were repeated for @xmath1 different chicks .
panel b of fig .
[ fig6 ] shows the results of the trained classifier on the whole data set ( 205 individuals ) .
the classifier present an accuracy of 98.05% ( 201/205 ) .
the crossed points corresponds to misclassification ( 4 individuals ) .
these misclassification are the consequence of mixed behavioural pattern . even for human observers ,
these individuals are difficult to categorize because , during the test , they present characteristics of imprinting or avoider behavioural traits . + as a result of individual tests we observed @xmath2 ( 113/205 ) imprinted chicks , @xmath3 ( 74/205 ) indifferent and @xmath4 ( 18/205 ) of avoider chicks .
) , and the last line is the distribution of the distance ( @xmath5 ) between the robot and the individual . only 30 individuals are considered .
all experiments have a duration of 1 hour .
the bars are drawn in the experimental chronological order from top left to right.,width=302 ] of all experiments : 64 individuals , priorly classed by human observation ) .
the green points corresponds to chicks that are attracted and follow the robot ( imprinted ) , the red points to chicks that avoid the robot ( avoider ) and the blue to chicks that do no react to the presence of the robot ( indifferent ) .
panel b shows how the trained classifier divides the plane into three regions and shows the result of the classification on the whole data set ( 205 individuals ) .
the green area corresponds imprinted chicks , the pink area to avoider chicks and the blue to chicks indifferent chicks .
crossed points ( 4 individuals ) correspond to misclassification after checking the whole dataset by human inspection .
misclassification happens when the chicks show a mixed behavioural type that is also difficult to interpret by a human observer .
, width=264 ] ethograms are typically used to describe animal behaviour .
they are made usually by human experts consistently documenting behaviour from observations .
the two main approaches are focal sampling ( in which the human expert make systematic observation of the behaviour of an individual it is a precise but time - expensive method ) and scan sampling ( in which the observer only uses observation recorded at regular intervals to classify behaviours it is less time - consuming compared to focal sampling , but also less accurate ) .
ethograms done by human observation are time consuming and tedious if one wants to obtain quantitative results .
we present a combined methodology that performs detailed quantitative studies of single individual repeated for a large number of animals ( 205 individuals ) .
+ recently , a number of ethological studies have used similar methodologies to automatically generate ethograms of animal behaviour from behavioural data @xcite .
this process of automatic behaviours detection is call `` automated ethology '' , and produces `` automated ethograms '' .
it is usually done in two steps .
the first step is called `` segmentation '' : the trajectories are split into segments .
each segment is labelled independently from the others , using either classification or clustering algorithms . in the second step , a label ( or `` class '' )
is assigned to each segment , describing which kind of behaviour the chick is exhibiting as the corresponding time .
two kind of algorithms can be used for labelling : classification algorithms and clustering algorithms . for each segment
, these algorithms use a set of features ( statistical measures ) of the observed individual to find a corresponding label . before the labelling process ,
classification algorithms need to be trained with a training set of segments already labelled by a human .
this training set is used as reference ( or _ a priori _ knowledge ) to guide the labelling process .
clustering algorithm do not need to be trained beforehand , and only uses the similarity of the segments features to group them together into clusters .
each cluster correspond to a different label .
+ the segmentation strategy can impact the reliability of the labelling process of trajectories compared to a labelling done by humans .
this is especially the case when segments incorporate periods of time when the chicks exhibit several different behaviour .
on the other hand , when segments are too small , it can add noise into the features used for labelling . in an optimal segmentation
, segments end when a chick changes its behaviour . a typical segmentation method used in the literature
is the @xmath6-steps window segmentation : trajectories containing @xmath7 time - steps are split into @xmath8 segments each with a size of @xmath6 time - steps .
an extreme case is the @xmath9-step window segmentation , where every time - step is a new segment .
note that the choice of @xmath6 can influence the reliability of a segmentation : if @xmath6 is too big , segments can include period of times where the chicks exhibit several different behaviours
. on the other hand , @xmath6-step window segmentation with a low @xmath6 can be more sensitive to noise .
+ we introduce a more robust segmentation method , the threshold crossing method .
we make two hypothesises , supported by human observations : first , that chicks reduce their speed below a given threshold when they change behaviour ; second , that chicks accelerate when they begin to exhibit a given behaviour , and decelerate when they change behaviour .
this allow us to introduce an original segmentation method , that we call threshold crossing. this segmentation method is based on the speed and acceleration of a chick ( fig .
[ fig7 ] ) : time - steps when the speed or acceleration of a chicks are below predefined thresholds are considered as stops .
+ ) the animal is considered as not moving .
to further segment intervals during which the animal is moving , the criteria is based on the acceleration .
first we define a deceleration threshold ( @xmath10 ) below which we consider that the animal is going to change type of movement . after being below the threshold , when the acceleration equals to 0
, the trajectory is segmented .
then a new segment is started where the animal has changed its movement type . in the shown example of experimental data ,
doted line ( a ) marks the transition from stop to movement , ( b ) from movement to stop , ( c ) from stop to movement , ( d ) change of movement type , ( e ) from movement to stop , ( f ) change of movement type and ( g ) from movement to stop.,width=313 ] the segments obtained are then labelled using state - of - the - art algorithms of the literature , either using classification algorithms : decision trees ( dt ) , decision trees with random forest ( dt / forest ) , k - nearest neighbours ( k - nn ) , support vector machines ( svm ) ; or a clustering algorithm : k - means .
these methods are described in detail in sec .
[ sec : behaviouralmeasurments ] and [ sec : classificationandclusteringoftrajectories ] . for each segment , the classification and clustering algorithms use the following features ( statistical measures ) to find a label ( corresponding to the behaviour of the chick ) : the mean speed of the chick , the acceleration of the chick , the distance between the chick and the robot , the distance between the chick and the wall of the arena , and the distance traveled by the chick during the experiment .
+ based on human observation ,
we separated chick behaviours into seven classes : stops : : the chick stops moving joins : : the chick , initially distant from the robot , joins the robot ( with a higher speed than when following it ) follows : : the chick is close to the robot and follows it in front : : the chick goes in front of the robot loops : : the chick goes around the robot and make a full loop around it bumped : : the robot pushes gently the chick out of its way other : : the chick has a behaviour that is not related to the robot note that the robot is programmed to constantly move , even if a chick is on its trajectory .
this set of classes correspond to the labels used by the classification algorithms during the labelling process .
+ we validated the automated labelling of chicks trajectories by computing the error compared to human - labeled trajectories : the collection of human - labeled behavioural sequences was split into two parts : the first part was used during the training step ( learning base ) , and the second part was used to perform this validation . as the observed error of classification is very low ( @xmath11 ) , we can say that the automated labelling is very close to the labelling done by humans .
+ l l l | l | l | l | l | l | l | & & & + & & & & & + & & & mape & std.dev . & mape & std.dev . & mape & std.dev .
+ & & dt @xcite & @xmath12 & @xmath11 & @xmath13 & @xmath11 & @xmath14 & @xmath15 + & & dt / forest @xcite & @xmath16 & @xmath17 & @xmath16 & @xmath18 & @xmath19 & @xmath20 + & & knn @xcite & @xmath21 & @xmath22 & @xmath23 & @xmath24 & @xmath25 & @xmath26 + & & svm @xcite & @xmath20 & @xmath27 & @xmath28 & @xmath28 & @xmath29 & @xmath30 + & & k - means @xcite & @xmath31 & @xmath32 & @xmath33 & @xmath34 & @xmath35 & @xmath36 + we measured the reliability of the different methods of segmentation and labelling presented compared to human - labeled trajectories .
in addition to the threshold crossing segmentation method , we also tested the @xmath6-step window segmentation for several values of @xmath6 : @xmath37 and @xmath38 .
these results are compiled in table .
[ table1 ] using mean absolute percentage error ( mape ) values .
the mape values are computed using the formula : @xmath39 with @xmath40 the human - labelled reference values and @xmath41 the values obtained by classification methods .
the best - performing classification method is the decision trees with random forests method , when using a threshold crossing segmentation or a 1-window segmentation .
methods using the @xmath42-step window segmentation method have a smaller reliability than methods using the @xmath9-step window segmentation : the segments obtained by @xmath42-step window segmentation can incorporate periods of time when the chick exhibit several different behaviours .
it should also be noted that chicks behavioural traces often include shorts stops of the chicks when they change behaviour .
these stops can have a duration inferior to @xmath42 time steps , and may impact negatively the performance of the @xmath42-step window segmentation .
the k - means clustering algorithm give promising results , but is outperformed by the tested classification algorithms .
it is not surprising , as this is a clustering algorithm : it is not trained beforehand , and can not use any _ a - priori _ information ( except the number of labels ) .
+ figure [ fig8 ] shows an automated ethogram of an individual chick using the best - performing segmentation and classification methods : `` threshold crossing '' and decision trees with random forest .
+ figure [ fig9 ] shows the distribution of durations of each behaviour for all imprinted chicks ( 113 individuals ) , all indifferent chicks ( 74 individuals ) and all avoider chicks ( 18 individuals ) , over three series of experiments .
statistical analysis ( kruskal - wallis one - way analysis of variance @xcite ) shows that the three datasets are significantly different ( p=@xmath43 ) .
+ figure [ fig10 ] presents an example of a trajectory of an imprinted chick .
this trajectory is segmented and coloured according to the observed behaviour of the chick .
the chick tends to stop when it is changing direction , or when it is changing it s behaviour .
+ automated ethograms generated by our methodology can take two forms .
individual ethograms , in the form of behavioural sequences ( fig . [ fig8 ] ) ,
describe the evolution of the behaviour of a single chick ( in one given experiment ) with respect to time .
the global dynamics of chicks behaviour for all imprinted chicks can be described using the transition frequencies between the seven behavioural patterns considered ( table [ table2 ] ) .
transition frequencies can be used to build global ethograms , in the form of finite state machines , or state transition diagrams , as shown in fig .
[ fig11 ] ( self - transitions are not taken into account ) . according to these figures ,
chicks tend to stop when they are changing their behaviour . &
_ loops _ & _ is bumped _ & _ other _
+ _ stops _ & & @xmath44 & @xmath45 & @xmath46 & @xmath47 & @xmath48 & @xmath49 + _ joins _ & @xmath50 & & @xmath51 & @xmath52 & @xmath53 & @xmath54 & @xmath47 + _ follows _ &
@xmath55 & @xmath56 & & @xmath57 & @xmath48 & @xmath58 & @xmath56 + _ in front _ & @xmath59 & @xmath54 & @xmath60 & & @xmath56 & @xmath47 & @xmath56 + _ loops _ & @xmath61 & @xmath62 & @xmath63 & @xmath54 & & @xmath54 & @xmath56 + _ is bumped _ & @xmath64 & @xmath54 & @xmath65 & @xmath54 & @xmath54 & & @xmath56 + _ other _ & @xmath66 & @xmath67 & @xmath56 & @xmath56 & @xmath56 & @xmath56 & + [ table2 ] ) .
the thickness of a transition arrow is proportional to its occurrence probability .
only transitions with an occurrence probability higher than 0.05 are represented .
self - transitions are not taken into account.,width=264 ] in this section , we show that is it possible to quantify the imprinting quality of the chicks toward the robot , by using the individual ethograms described in sec . [
sec : automatedethograms ] .
only data from imprinted and indifferent chicks are considered , as avoiders chicks avoid any imprinting stimuli toward the robot : the imprinting process fails and the imprinting quality is null .
+ the global distribution of behaviours for every chick can be found in fig .
[ fig12 ] .
our objective is to define an imprinting index , corresponding to the measure of individual attachment of a chick toward the robot .
we use information from the distribution of behaviour of each chicks to compute this index .
as we want to define a one - dimensional index , this information needs to be compressed . + , taking into account the most important behaviours . a pca analysis show that @xmath68 of the variance in the data is explained by the first principal component .
then , we build an imprinting index on the base of a linear transformation of the vector .
this imprinting index allows to qualitatively sort each individual on an imprinting scale .
only data from imprinted and indifferent chicks are used.,width=313 ] we introduce a new representation of the behaviour of each chicks , similar to the one used in [ fig12 ] and previous sections , but with a reduced dimensionality ( @xmath69 instead of @xmath70 ) : for each chick , a vector of three values @xmath71 represents respectively the portion of time the chick ( 1 ) rests , ( 2 ) runs toward the robot and ( 3 ) stays close to the robot .
these values are presented in fig .
[ fig12 ] . in this representation ,
the `` bumped '' behaviour is ignored because it only occurs rarely .
the `` other '' behaviour is ignored because it does not correspond to a clearly identified behaviour .
+ to obtain a representation of chick behaviour with a further reduced dimensionality , we performed principal component analysis ( pca ) on the dataset composed of @xmath72 vectors for all chicks : @xmath68 of the variance in this dataset is explained by the first principal component .
this shows that the first principal component is sufficient to quantify the behaviour of a chick .
the first principal component is computed as follow : @xmath73 , where @xmath74,@xmath75 and @xmath76 .
we define an imprinting index , corresponding to the measure of individual attachment of a chick toward the robot , by a linear transformation of the first principal component : @xmath77 where @xmath78 is chosen to separate imprinted and non - imprinted ( indifferent and avoiders ) sets .
+ this imprinting index can be used to sort all chicks by the `` imprinting force '' of a chick toward the robot . + a quantitative analysis of the imprinting level of imprinted and indifferent chicks ( 187 individuals ) was performed ( as described in sec .
[ sec : classificationres ] ) , with results shown in fig .
[ fig13 ] .
it shows that the majority of indifferent chicks have an imprinting index close to @xmath79 , with small imprinting variability : it can be explained by the fact that indifferent chicks are immobile during most of the experiments .
imprinted chicks have a larger imprinting variability than indifferent chicks .
most imprinted chicks have an imprinting index between @xmath80 and @xmath81 .
+ we investigate the influence of social bonding on the imprinting quality of chicks toward the robot .
after the imprinting process , @xmath82 randomly selected chicks are put in @xmath83 groups of @xmath84 individuals for @xmath85 days .
each of these group contains a random number of imprinted , indifferent and avoiders chicks .
then , we assess if the chicks have changed their individual behaviour toward the robot by performing another set of experiments , with the same approach described in previous sections .
these experiments were only performed on @xmath86 individuals in @xmath70 groups .
we produce imprinting indexes using the method described in [ sec : indivvar ] . to study the change in imprinting quality of chicks put in groups compared to chicks alone , we introduce the measure : @xmath87 where @xmath88 and @xmath89 are the imprinting index of a chick respectively before and after being put into groups . the distribution of the @xmath90 measure is shown in fig .
[ fig14 ] .
the mean value of all considered @xmath90 is @xmath91 : we observe a general decline of individual attachment toward the robot .
it can possibly be explained by the aging of the animals .
+ , thus we observe a general decline in the individual attachment towards the robot , possibly due to the aging of the animals .
note that this distribution is not normal , even if it looks the part ( the kolmogorov - smirnoff test fails when compared to the normal distribution , with @xmath92).,width=313 ] to take into account the composition of the groups in our analysis , we introduce the _ group imprinting index _ as an average of imprinting indices of group members in the first individual test : @xmath93 where @xmath94 is the imprinting index of the @xmath95-th chick in a group before being put into a group .
we compute @xmath96 as a difference between the group average and the individual imprinting index : @xmath97 .
chicks for which this value is positive shows an imprinting quality below the average in their group , and chicks for which this value is negative show more attachment toward the robot compared to the average of their group .
the link between the @xmath98 and @xmath90 is shown in fig .
[ fig15 ] .
the represented distributions are significantly positively correlated ( student s t test for a transformation of the pearson s correlation @xmath99 , with a correlation coefficient of @xmath100 ) .
we observe that , after being put in groups , chicks have an imprinting quality toward the robot that is closer to the average of their group . a chick that originally was
not very interested in the robot develops a stronger attachment after being placed in the group of individuals strongly imprinted toward the robot . on the other hand , a chick that originally was very attached to the robot
become less interested in the robot after being placed in the group of individuals that have a low interest toward the robot .
+ chicks in @xmath70 groups of @xmath84 chicks presenting various level of imprinting .
after being put in groups , chicks have an imprinting quality toward the robot that is closer to the average of their group .
a chick that originally was not very interested in the robot develops a stronger attachment after being placed in the group of individuals strongly imprinted toward the robot . a chick that originally was very attached to the robot
become less interested in the robot after being placed in the group of individuals that have a low interest toward the robot.,width=313 ]
* ethics statement * animal experiments were performed in accordance with the recommendations and guidelines of the local competent authority and ethic committee .
the robot does not look like a chicken , however chicks developed a strong social attachment to it thanks to the filial imprinting mechanism @xcite .
chicks prefer imprinting object with spotted pattern , dots or strikes @xcite .
moving objects are more attractive @xcite , as visual imprinting would occur on the association between object form and movement @xcite .
chicks are attracted by colours that are easy to discriminate from the colours of the ground .
red - orange or blue revealed to be the most attractive colours @xcite . not only the nature but the characteristics of the colours have an influence on recognition and imprinting @xcite .
chick can learn or be imprinted on static coloured two - dimensional stimuli @xcite .
contrasted patterns attract more the chick attention and are more effective for accurate memory @xcite .
+ auditory communication is of importance in poultry @xcite and imprinting also depends on auditory stimuli @xcite .
early exposure to auditory stimulus induces preferences in experiments .
several vocalizations and their roles still need to be investigated in imprinting process but it is known that early exposure to a sound influence the later preferences .
it seems that both auditory and visual learning are enhanced during the critical period , which is consistent with the interpretation that imprinting is a within - event learning occurring when elements and their representations are linked @xcite .
+ the main features of the robot namely movement , colour pattern and sound emission have been designed taking into account these results . moreover ,
each of them can be switched on and off and programmed independently of each other allowing the study of the multi - modality of factors favoring filial imprinting .
the fact that the robot is fully autonomous and can move in the same environment as the chicks allows studies where the animals can move freely and in groups , thus allowing to study the link between individual and collective behaviour .
+ the mobile robot used in this study is a track - type mobile robot presented in fig .
[ fig1 ] @xcite .
we named it the _ poulbot_. the robot is a configuration of the marxbot robot a modular research robot developed at epfl @xcite .
the poulbot consists of the following modules : a _ base module _ providing energy , means for locomotion , a gyroscope , an accelerometer , 24 short range infrared proximity sensors around the robot and a speaker , a _ colour pattern module _ allowing a robot to carry a specific pattern , an extra _ ring of proximity sensors _ placed above the colour pattern module and a _ main computer board _ with a 533mhz freescale i.mx31 and an omnidirectional camera , the main board also provides wi - fi and bluetooth wireless connectivity . on the top of the robot we fixed three colour markers used to track the robot position and orientation with a camera fixed above the experimental setup ( see below ) .
every module of the robot consists of one or several microchip dspic33 microcontrollers that drive the sensors and/or actuators of the module .
the microcontrollers are connected with each other and with the main computer of the robot through a controller - area network ( can ) bus .
the robot can be extended with the sound acquisition module and with the pecking module .
+ the control system of our robot is a hierarchical behaviour based controller @xcite .
the robot is equipped with a set of primitive behaviours tightly bonded with the sensors and actuators of the robot .
each primitive behaviour serves to achieve a particular goal or to perform a specific activity ( e.g. , _ wall - following _ or _ obstacle - avoidance _ behaviours ) . the primitive behaviours are combined together to form higher level composite behaviours for specific experiments .
behaviour - based systems work well in dynamic environments , in cases when fast reaction and high adaptability are important .
these characteristics make the behaviour based approach a natural choice when designing a control system for a robot that interacts with animals .
we performed experiments with chicks , _ gallus gallus domesticus _ , from hatching up to @xmath69 week - old .
as breed , we chose the egg layer white leghorn that is common in scientific studies .
the incubators were disposed together in a @xmath101 isolated rearing room that was maintained in the dark and at a constant temperature of @xmath102 during incubation .
eggs were incubated at the classical temperature of @xmath103 with a relative humidity of @xmath104 .
after hatching , chicks were identified with numbered plastic rings their hatching time was recorded .
they were left to rest and dry in the incubators .
when dry , they were kept in the dark in individual boxes until the imprinting process .
the light was left off to reduce imprinting of first hatched chicks on siblings .
the presence of humans was reduced to the minimum necessary for handling . during hatching ,
the frequency of visit by experimentalists was of @xmath9 hour .
+ during incubation and up to the imprinting process , two speakers and a mp3 device played a calibrated sound sound that the robots also played during imprinting and tests .
the sound was played from egg state to the end of the imprinting sessions .
this calibrated sound had a frequency of @xmath105 and with a beep duration of @xmath106 and interval between two beeps of @xmath107 .
these features have been chosen in correspondence to the range of chick audition @xcite .
we chose the upper threshold of the range and a sound designed to be neutral since it is totally artificial , which means that it is not related to a natural chick or hen calls , such as alarm calls or clucking , and have thus no biological sense .
this was made for auditory imprinting not oriented towards a special message signal .
+ after the imprinting process , chicks are kept in groups of @xmath84 in breeding nurseries and given food and water _
ad libitum_. the temperature of the breeding room was reduced during the three weeks from @xmath108 to @xmath109 according to breeding standards .
we ran open - field tests , where animals and robots were released on a flat arena surrounded by walls and their behaviour was studied .
the experimental arena is a flat square of @xmath69 m by @xmath69 m , surrounded by a wooden wall of @xmath110 cm in height .
a common daylight lamp is a source of a strong infrared ( ir ) emission that affects the ir sensors of the robot . to resolve this issue , we used lamps with reduced infrared emission fq49w/965 by osram .
twelve lamps were uniformly fixed on the ceiling to provide lighting conditions as homogeneous as possible .
a scout sca1000 - 30gc colour camera by basler vision technologies with a cs - mount t3z2910cs varifocal lens by computar was fixed above the setup for tracking and recording tasks .
the camera has a resolution of @xmath111 pixels and a maximum frame rate of @xmath112 frames per second .
it uses a gigabit ethernet connection to transfer video to the pc .
the experimental pc has an eight - core intel core@xmath1132 quad processor and 2 gb of ram .
it runs the monitoring and recording software presented in the next section and the gui module of the robot control system .
it also serves as a temporary storage of the experimental data recorded during the day , before the data is transferred to external hard drives .
the temperature in the experimental room during the whole experimentation period was kept constant .
the experimental facilities are described in fig .
[ fig16 ] .
the imprinting set - up is shown in fig .
[ fig17 ] .
+ the imprinting procedure precedes every series of experiments . depending on birth rates
, we tried to compose imprinting batches with chicks born within at most two hour intervals .
this was made to have a sufficient number of chicks per imprinting session . to perform
the imprinting we build an arena composed of two rows of boxes on the sides .
four boxes were built in mdf wood with a plexiglas side facing the wall to let chicks see the robot that traveled back and forth in front of them .
wooden boxes were @xmath114x@xmath115x@xmath116 mm .
each box had ten compartments of @xmath115x@xmath115 mm to house one chick at a time . by putting one chick per box
the setup allowed the simultaneous imprinting of 40 chicks .
four robots moved along the walls in front of each box at a close distance of about @xmath117 mm , forward and backward , at a speed of @xmath110 mm / s , one in front of each box .
the robots emitted a calibrate sounds that was the same as during incubation and hatching ( see above ) and displayed a specific colour pattern .
this individual and in - line imprinting process was designed to reduce imprinting among siblings .
we performed the imprinting sessions according to the optimal period for chicken filial imprinting the the first exposure to the robots was 9 hours after hatching @xcite . the first exposure
was programmed 9 hours after the birth of the chicks .
three sessions of one hour in the presence of the robot were done interrupted in - between by one hour resting time in the breeding room .
rotation were done to imprint all chicks of a hatching batch ( 100 individuals at most ) . +
shortly after an end of the imprinting procedure we tested all chicks : every chick was left for a half an hour with a robot _ wandering _ on the experimental arena and chick behaviour was observed .
the robot emits a calibrated beeping sound and moves straight at a constant speed of @xmath118 mm / s .
when the robot reaches an edge of the set - up it bounces back with a random angle .
the speed of the robot was selected according to observation of the mean chick speed during previous calibration experiments .
each test lasted for 30 minutes and for data analysis the first 2 minutes are skipped and the next 25 minutes are kept .
this avoids artifacts due to animal introduction and extraction from the experimental setup .
all tests were recorded and stored as digital videos for further processing .
the chicks were tested , one after the other , in the sequence corresponding to their age , older first , younger last . for classifying chicks imprinting success , as described in sec .
[ sec : classificationres ] , we used a linear support vector machine ( svm ) classifier , a supervised learning method that is often used to solve problems in classification , regression , and novelty detection @xcite .
+ classification using linear svm is performed by solving the problem of finding a hyperplane partitioning the training dataset . during training ,
the examples of separate classes are separated by a gap that is as wide as possible .
we make the hypothesis that our exploration space can be linearly separated , so using linear svm ( instead of the non - linear svm using kernel function @xcite ) is sufficient .
svm is fundamentally a two - class classifier , but a number of methods exist to handle a higher number of classes @xcite .
+ the libsvm ( version 3.0 ) @xcite was used to perform the classification .
+ we used the following statistics on chicks and robot movements as relevant features ( explanatory variables ) for the classification process : _ mean distance between a robot and a chick _
, _ mean speed of a chick _ , and its _
standard deviation_. we reduced the dimensionality of the related feature vector from three to two by using principal components analysis ( pca ) @xcite .
the resulting dataset was used to train a three - class svm classifier .
+ the training takes about 6 milliseconds ( with 64 individuals ) and the classification takes about 7 milliseconds ( for 205 individuals ) with matlab r2011b running on an ordinary laptop pc ( lenovo thinkpad t60p , linux ubuntu 12.04 ) .
moreover , for individual tests the extraction of the chick and robot positions is done automatically in real time during the recording of the test .
thus the classifier allows to categorize , in real time , rapidly , large batches of experiments saving a huge amount of experimental time .
the recorded experimental videos were analyzed during the experimental run using a swistrack that is an open source multi - platform tracking software @xcite .
the robot position and orientation are defined by three colour markers on top of it .
thus the detection of robots and chicks is colour based .
the coordinates of the robots and the animals extracted from the video were mapped to the real - world coordinates ( in mm ) by using the calibration routine based on the well known tsai s calibration technique @xcite . in order to remove high frequency noise introduced into trajectories by the detection errors ,
every trajectory is filtered by using the savitzky - golay smoothing filter @xcite . +
this trajectories are then segmented .
the segmentation process identify segments of the trajectories ( i.e. : group of continuous time - steps in the chicks behavioural traces ) when the chick exhibit a particular behaviour .
this is a first step toward labelling the trajectories with the observed behaviour of the chicks toward the robot . in an optimal segmentation
, segments should end when a chick changes into behaviour . a typical segmentation method used in the literature
is the @xmath6-steps window segmentation : trajectories containing @xmath7 time - steps are split into @xmath8 segments each with a size of @xmath6 time - steps .
an extreme case is the @xmath9-step window segmentation , where every time - step is a new segment .
note that the choice of @xmath6 can influence the reliability of a segmentation : if @xmath6 is too big , segments can include period of times where the chicks exhibit several different behaviours
. on the other hand , @xmath6-step window segmentation with a low @xmath6 can be more sensitive to noise .
we introduce a more robust segmentation method , the threshold crossing method .
we make two hypothesises , supported by human observations : first , that chicks reduce their speed below a given threshold when they change behaviour ; second , that chicks accelerate when they begin to exhibit a given behaviour , and decelerate when they change behaviour .
this allow us to introduce an original segmentation method , that we call threshold crossing. this segmentation method is based on the speed and acceleration of a chick ( fig .
[ fig7 ] ) : time - steps when the speed or acceleration of a chicks are below predefined thresholds are considered as stops . to find the quantitative representation of these behaviours in the parameter space we use either classification algorithms or clustering algorithms . to prepare the training set we manually labeled a number of segments of trajectories , obtained using the segmentation method described in [ sec : behaviouralmeasurments ] .
while classification algorithms are trained on human - labelled data , clustering algorithms identify behaviours without any training .
either techniques will label observation data into 7 classes of behaviours : _
`` stops '' , `` joins '' , `` follows '' , `` in front '' , `` loops'',``bumped '' _ and _ `` other '' _ , which reflect the activity of the chick , and its relation to the movements of the robot .
the python library scikit - learn was used to perform all classification and clustering operations . + ( 1 ) decision trees ( @xcite , * dt * in table [ table1 ] ) .
decision trees correspond to a recursive partitioning of the training set into several homogeneous subsets .
it has a flowchart structure where nodes and branches represent conditions ( expressing how to partition the training set ) , and leafs nodes represent class labels .
decision trees are non - parameter and nonlinear , and so can be used in cases when little is known a - priori about the relationship between the variables .
they are also simple to interpret and verify by a human .
the classification could incur a loss due to overfitting , a pruning mechanism @xcite was used to solve this issue .
+ ( 2 ) random forests ( * dt / forest * in table [ table1 ] ) .
it is a variation of the decision tree method : one can use an ensemble of trees , that is known as a random forest @xcite , instead of a single tree , to increase the accuracy of classification .
+ ( 3 ) a k - nearest neighbours classifier , with an euclidean distance metric @xcite .
this algorithm makes use of a measure of similarity between all the training records and the features of the object to be classified , then identify the most similar @xmath95 neighbours of all objects , and determine each object class by an object by a majority vote of its neighbours ( i.e. assign the class which is the most frequent among the @xmath95 training records nearest to that object ) .
note that knn classifiers are known to be sensitives to noisy , irrelevant features , or by the difference of scaling of the features @xcite .
that may explain the classification results ( table [ table1 ] ) , where knn was out - performed by other algorithms .
+ ( 4 ) linear multi - class svm classifier , as described in section [ sec : classificationmethod ] .
+ ( 5 ) k - means @xcite .
this clustering algorithm partitions the observations into @xmath119 clusters , in which each observation belongs to the cluster with the nearest mean .
the results are akin to partitioning the data space into voronoi cells .
convergence speed can be greatly affected by the choice of initial values .
we use the popular k - means++ @xcite algorithm for choosing initial values .
this study validates the use of imprinting to ensure a social bond between robots and animals in order to perform controlled experiments with calibrated stimuli and sustained social interactions .
when the robots are accepted by the animals , they can be programmed to keep sustained social interactions with them in long duration trials ( our experiments lasted 30 minutes ) .
as the robots are autonomous and work in close loop of interactions with the animals , they allow to avoid any human interference during the trials .
the whole set - up presented here works in absence of humans in the experimental room and without supervision .
the robots can be programmed to produce calibrated and repetitive multi - modal stimuli including visual ( colour patterns ) , sound ( prerecorded natural cues or totally artificial ) and motion patterns .
+ moreover , we build a framework that allows automation of high throughput data analysis . from the positions of animals and their dynamics in time we construct quantitative automated ethograms . + we test state - of - the - art algorithms to automate data analysis and show that supervised method based on decision tree with forest gives the best results .
we also show that simple unsupervised methods such as k - means already gives interesting results . here , the algorithms allow to combine the knowledge of human experts with fast computational methods .
it is then possible to analyze the experimental data of 205 animals in a few minutes on modern computers .
+ this paper has two contributions .
first , we introduce a methodology that allows the automated generation of quantified ethograms both on the individual level ( represented by sequences of labelled behaviours ) and on the group level ( represented by finite state machines ) .
second , we investigate the individual variability of chicks in term of imprinting , using a classification algorithm . results also confirm that social bonding have an influence on the imprinting quality . we observe that , after being put in groups , chicks have an imprinting quality toward the robot that is closer to the average of their group .
a chick that originally was not very interested in the robot develops a stronger attachment after being placed in the group of individuals strongly imprinted toward the robot . on the other hand , a chick that originally was very attached to the robot
become less interested in the robot after being placed in the group of individuals that have a low interest toward the robot .
+ this methodology is a first step towards mathematical modelling based on quantified ethograms , and this study is a first step toward making the link between inter - individual variability and collective behaviour . in our methodology , we first study and quantify individually each chick .
then , the chicks are put in different grouping configuration , according to their inter - individual variability .
an interesting and on - going perspective is to make a group behaviour analysis taking in account the chicks individual behavioural traits , and the group composition .
this work is supported by the swiss national science foundation grant no .
112150 `` mixed society of robots and vertebrates '' , eu - ict project `` assisi - bf '' , no . 601074 .
a gribovskiy , j halloy , jl deneubourg , h bleuler , and f mondada . towards mixed societies of chickens and robots . in _ proceedings of the ieee / rsj 2010 international conference on intelligent robots and systems ( iros 2010 ) _ , 2010 .
g sempo , s depickre , jm am , c detrain , j halloy , and jl deneubourg .
integration of an autonomous artificial agent in an insect society : experimental validation . in _ from animals to animats 9 _ , pages 703712 .
springer , 2006 .
j halloy , g sempo , g caprari , c rivault , m asadpour , f tche , i said , v durier , s canonge , jm am , c detrain , n correll , a martinoli , f mondada , r siegwart , and jl deneubourg . social integration of robots into groups of cockroaches to control self - organized choices . , 318(5853):11551158 , 2007 .
p phamduy , g polverino , rc fuller , and m porfiri .
fish and robot dancing together : bluefin killifish females respond differently to the courtship of a robot with varying color morphs .
, 9(3):036021036021 , 2014 .
m bonani , v longchamp , s magnenat , p rtornaz , d burnier , g roulet , h bleuler , and f mondada .
the marxbot , a miniature mobile robot opening new perspectives for the collective - robotic research . in _ to appear in proceedings of the 2010 ieee / rsj international conference on intelligent robots and systems _ , 2010 .
d arthur and s vassilvitskii .
k - means++ : the advantages of careful seeding . in _ proceedings of the eighteenth annual acm - siam symposium on discrete algorithms _ , pages 10271035 .
society for industrial and applied mathematics , 2007 . | inter - individual variability has various impacts in animal social behaviour .
this implies that not only collective behaviours have to be studied but also the behavioural variability of each member composing the groups . to understand those effects on group behaviour
, we develop a quantitative methodology based on automated ethograms and autonomous robots to study the inter - individual variability among social animals .
we choose chicks of _ gallus gallus domesticus _ as a classic social animal model system for their suitability in laboratory and controlled experimentation . moreover , even domesticated chicken present social structures implying forms or leadership and filial imprinting .
we develop an imprinting methodology on autonomous robots to study individual and social behaviour of free moving animals .
this allows to quantify the behaviours of large number of animals .
we develop an automated experimental methodology that allows to make relatively fast controlled experiments and efficient data analysis .
our analysis are based on high - throughput data allowing a fine quantification of individual behavioural traits .
we quantify the efficiency of various state - of - the - art algorithms to automate data analysis and produce automated ethograms .
we show that the use of robots allows to provide controlled and quantified stimuli to the animals in absence of human intervention .
we quantify the individual behaviour of 205 chicks obtained from hatching after synchronized fecundation .
our results show a high variability of individual behaviours and of imprinting quality and success .
three classes of chicks are observed : imprinted with variability , indifferent to the imprinting stimuli ( i.e the robots ) and chicks that avoid the imprinting stimuli .
our study shows that the concomitant use of autonomous robots and automated ethograms allows detailed and quantitative analysis of behavioural patterns of animals in controlled laboratory experiments . |
extragalactic jets are a common feature of radio - loud active galaxies @xcite .
these galaxies presumably host a supermassive black hole in their nuclei ( @xmath0 ) , and the accretion disks of matter that form around them originate the observed strong optical / uv emission . around one per cent of active galaxies are radio - loud , with the radio emission being emitted from jets .
current models suggest that the strong magnetic fields in these systems are able to extract energy from the black - hole or the accretion disk @xcite .
the particles get collimated and accelerated by the magnetic fields , acquiring terminal speeds close to the speed of light @xcite .
jets propagate through the host galaxy and even outside , bringing large amounts of particles and energy from the central regions of the host galaxies to regions hundreds of kiloparsecs away ( @xmath1 during @xmath2 for powerful jets ) .
this process completely alters the gas distribution and evolution of the host galaxies ( typically massive ellipticals at the centers of clusters ) , as a direct consequence of the crossing of a supersonic flow .
thus , it is important to understand the nature and properties of jets not only because they are macroscopic laboratories of plasma physics but also to derive accurate information about the processes that determine their interaction with the ambient and quantify the role of relativistic outflows in galaxy and cluster evolution .
jets emit electromagnetic radiation throughout the whole electromagnetic spectrum , primarily via synchrotron and inverse compton radiation .
the parsec - scale structure of jets in active galactic nuclei ( agn ) is mainly observed in the radio band , using the vlbi technique .
the nature and properties of the emitting region as related to the flow are still scarcely known .
jets can be treated as flows because the larmor radius of particles is very small when compared to the spatial scales of the problem studied @xcite , thus allowing to consider the whole system as a fluid .
this interpretation has different implications : such systems , which are composed of magnetic fields and particles propagating along the jet channel are expected to host the growth of different hydrodynamical and/or magnetic instabilities .
the instabilities are waves triggered by any external or internal perturbation that grow in amplitude with distance , as they are advected with the flow ( see reviews in refs .
the jet acts in a similar way to waveguides , hosting the waves that bounce against its `` walls '' . in the case of the unstable modes the reflection coefficient of
the wave is larger than one @xcite .
the energy gain comes from the kinetic energy of the flow , which is correspondingly decelerated .
if instabilities do really grow in jets , we would expect to observe wave - like structures .
different works have related observed structures to instabilities in extragalactic jets , at parsec - scales ( knots , bendings , helices ; see e.g. , @xcite ) .
current issues that are open in jet physics include an aspect as basic as the nature of the observed radio jets in active galaxies .
in particular it is not clear whether the vlbi observations of parsec - scale jets reveal wave - like structures that develop and propagate along the jet , or they trace the whole jet flow . whether the jet flow is energetically dominated by non - radiating thermal particles or it is magnetically dominated up to kiloparsec scales , as expected to be the case at the formation region , is also a matter of debate . in reference
@xcite , the authors used observations of the jet in s5 0836 + 710 at different frequencies and epochs and showed that the ridge line of this jet behaves as expected if they are interpreted as pressure waves .
the jet in bl lac also shows wave - like structures propagating along the jet and these are identified by the authors as alfvn waves in a magnetically dominated jet @xcite . in this contribution
i review results obtained for the quasar s5 0836 + 710 using vlbi observations and put them in context with other recent works .
the luminous quasar s5 0836 + 710 is at a redshift @xmath3 and hosts a powerful radio jet extending up to kiloparsec scales @xcite . at this redshift , @xmath4 ( see , e.g. , mojave database ) .
vlbi monitoring of the source showed kink structures @xcite and yielded estimates of the bulk lorentz factor @xmath5 and the viewing angle @xmath6 of the flow at milliarcsecond scales @xcite .
the jet was observed at 1.6 and 5 ghz with vsop ( vlbi space observatory program , a japanese - led space vlbi mission ) , and oscillations of the ridge - line were also observed @xcite .
these oscillations were interpreted as produced by kelvin - helmholtz ( kh ) instability modes @xcite .
the source has also been monitored by the mojave program @xcite .
finally , it has been suggested that the lack of a collimated jet structure and hot - spot at the arc - second scales is due to jet disruption , as indicated by the growth in the amplitude of the helical structure @xcite . the jet in s5 0836 + 710 as obtained with global vlbi and vlba ( very long baseline array ) at 8 and 15 ghz in 1997 and 2002 , respectively .
the ridge - line points are indicated by red circles in each image @xcite.,title="fig:",scaledwidth=45.0% ] the jet in s5 0836 + 710 as obtained with global vlbi and vlba ( very long baseline array ) at 8 and 15 ghz in 1997 and 2002 , respectively .
the ridge - line points are indicated by red circles in each image @xcite.,title="fig:",scaledwidth=45.0% ]
in reference @xcite the authors presented a number of observations of the jet in s5 0836 + 710 obtained at different epochs and different frequencies ranging from 1.6 ghz to 43 ghz , using various ground arrays .
the position of the ridge - line along the jet was defined as the peak of emission at a given distance to the radio - core .
figure 1 shows two example images of the jet at 8 and 15 ghz together with the points that form the ridge - lines ( red dots ) .
the combination of all the maxima obtained along the peak gives the resulting ridge - line .
the jet in s5 0836 + 710 shows a structure compatible with a projected helix .
this work showed that the observing array and epoch ( through the _ uv_-coverage ) of a given observation does not have any effect on the large - scale structures observed : the large - scale structures of the ridge - lines of the jet at a given epoch do coincide independently from the observing frequency and array .
this result was interpreted as evidence of the physical nature revealed by ridge - lines .
small - scale oscillations may , on the contrary , be produced by observational artifacts .
figure 2 shows the coincidence of the ridge - lines for a given epoch .
the empirical relationship between the ridge - line and the conditions in the jet that produce more emission was established by the highest - resolution images at 15 ghz ( from mojave data , @xcite ) , which showed that the peak is not necessarily located at the geometrical center of the brightness distribution . on the contrary
, the peak of emission is located at the expected positions if it is associated to the pressure asymmetry driving the helical pattern .
therefore , it was concluded that the peak in emission should correspond to a pressure maximum across the jet cross - section . in references
@xcite , cohen and collaborators presented a similar study for the jet in bl lac with observations at 15 ghz and were able to measure the propagation velocity of the ridge - line , which was interpreted as an alfvn wave , under the assumption that the energy flux of the jet is dominated by the poynting flux .
in addition , the mojave collaboration has shown that the bright features in jets ( typically associated to traveling shock - waves ) appear moving in different directions from the radio - core with time and that a single observation at a given epoch would not show the complete width of the jet @xcite .
the obvious conclusion is that the jet flow fills a wider region than that revealed at a single epoch .
a remarkable example of the structure described in this section , albeit at kiloparsec scales , is the _ chandra _ image of the jet in the radio - galaxy ngc 315 @xcite . summarizing the collected information
, there is evidence that the emission maxima along radio jets do not correspond to observational artifacts .
furthermore , the jet flows at parsec - scales are wider than the region shown in radio observations .
we do not have any conclusive evidence , however , about the exact nature of the observed wave - structures or how the interaction with the propagating perturbation influences them .
ridge - lines over plot of the ridge - lines obtained at different frequencies in 1997 @xcite . ]
the parsec - scale jets seem to show only a portion of their complete cross - section at ghz frequencies .
the following discussion from @xcite claims that there are two possible explanations for one side of the jet being brighter than the other : _ if the jet is transversally resolved , the maximum in the emission , or the position of the ridge - line can be due to two main factors : 1 ) differential doppler boosting caused by velocity gradients in the jet , or 2 ) an increase in the local pressure and number of emitting particles .
if the position of this maximum is due to differential doppler boosting , changes in the flow direction would have an influence on the local emission from the different regions .
this should result , depending on the internal velocity gradient , in irregular structures , contrary to the results obtained .
in addition , relatively large changes in speed and/or flow direction are required in order to get significant differential doppler boosting within the jet .
thus , the most reasonable hypothesis is that the ridge - line corresponds to a pressure maximum .
_ it is possibly important to note that emission asymmetries can also be produced by a helical field .
however , this option requires that the magnetic field changes intensity along the jet resulting in a periodical oscillation of the bright region with distance . and
this , in turn , brings us again to factor 2 in the previous paragraph : a perturbation in the jet that propagates ( rotating ) downstream .
how large the emissivity difference has to be for our detectors not to reveal the weakly emitting side of the jet remains to be studied .
it has been suggested , for the case of bl lac @xcite , that the observed pattern corresponds to a pulse traveling along the jet magnetic field as an alfvn wave . on the other hand , in the case of 0836 + 710 ,
it was pointed out that the longest helix should correspond to a periodical perturbation producing a wave - pattern @xcite .
jet disruption at larger scales was reported later @xcite , and it was interpreted as further evidence of the helical wave coupling to a disruptive kh mode , growing in amplitude and eventually disrupting the flow .
these two views are , nevertheless , compatible : certain jets showing large - scale helical structures can be subject to precession motion at the jet base and , at the same time , irregularities at the formation region can be produced by changes in the accretion rates onto the central compact object ( e.g. , @xcite ) or any other kind of stand - alone perturbation , e.g. , perturbations triggered by asymmetries in the jet itself or in the ambient medium , lateral winds or entrainment of clouds that rotate around the galactic centre from a side of the jet , for instance .
these processes are , however , extremely difficult to detect and isolate .
as we stated before , in the case of bl lac the structures observed to propagate downstream were interpreted as an alfvn wave in a magnetically dominated jet @xcite . the strongest evidence to favor
this hypothesis is the well - ordered distribution of polarization vectors .
however , this can not be a conclusive evidence of a strong magnetic field , taking into account that if the flow of particles is ordered ( e.g. , collimated by external gas pressure ) and dominates the dynamics , similar distributions could be observed under the appropriate circumstances .
in addition , different authors have discussed that during the jet acceleration process the magnetic field is the only available energy source for the jet particles to be accelerated ( see , e.g , @xcite ) .
therefore , stronger evidence is required to clarify this aspect .
bright features are routinely observed traveling along parsec - scale jets ( e.g. , @xcite ) .
these are typically associated to inhomogeneities at injection that trigger shock waves in the jet because they are supersonic .
both these perturbations and helical structures may coexist in jets , as shown by numerical simulations and recent results from the blazar cta 102 ( e.g. , @xcite ) . on the one hand
, jet precession at the jet base can trigger the large - scale morphology in jets as s5 0836 + 710 @xcite , or a perturbation of the magnetic field could propagate downstream in the way that has been reported in bl lac @xcite . on the other hand , perturbations in the mass - energy flux can produce the observed radio - components .
these processes can be related to physical processes as different as precession of the rotation axis of the central region and accretion flux , so they could both happen .
the jet in s5 0836 + 710 showed transversal motion of the ridge - line on the plane of the sky at 15 ghz compatible with an oscillatory behavior , along the first 3 mas of the radio - jet ( see figure 3 ) @xcite .
the displacements measured were smaller than the errors in the determination of the positions and the velocities measured were superluminal , which can not be explained by projection effects ( the motions are measured on the plane of the sky ) .
this effect has also been observed in the motion of radio - components in sources like nrao 150 @xcite and oj 287 @xcite . in particular , for the case of s5 0836 + 710 , it was suggested that the radio - core could be oscillating and that simply by forcing it to fall at the same position introduced artificial displacements when measuring the ridge - line displacements .
it was shown that even a very small oscillation ( beyond the available resolution ) could be responsible for the apparent superluminal motions .
moreover , the oscillation of the radio - core position had been proven for another , closer source ( m 81 , @xcite ) using the phase - reference technique .
the same idea has been recently applied to the superluminal transversal motion of radio - components in nrao 150 @xcite .
measured transversal velocity from two epochs at 15 ghz in mas / yr @xcite . ]
the success of the first observation of the quasar s5 0836 + 710 using the russian space antenna _ radioastron _ in combination with the available on earth arrays ( global vlbi ) will allow for unprecedented resolution in the determination of the ridge - line location at the observing frequencies : 1.6 ghz and 5 ghz so far .
in addition , these observations will be compared with the images obtained in 1997 with vsop ( vlbi space observatory program ) using the previously available space - antenna ( the japanese _ halca _ ) @xcite .
the aim is to try to measure the transversal velocity of the ridge - line using the displacement between those well resolved images .
this measure can help to characterise the structure in detail and obtain relevant information about the jet physics .
relativistic outflows in active galaxies represent one of the most extreme scenarios that we know in the universe .
they are macroscopic laboratories of plasma physics in the relativistic regime and could play a fundamental role in the evolution of their host galaxy ( typically a massive galaxy at the center of clusters ) and its environment . among the open issues in this field we find the relation between the observed radio jets and the bulk ( thermal ) flow . in this contribution ,
i have summarized the evidence published in recent papers about the association of the brightness ridge - line in radio - jets with wave - like structures .
there are , however , different models that could explain them . on the one hand , the helical pattern given by the ridge - line can be interpreted as a wave produced by perturbations at the or close to the formation regions that couple to an unstable kh mode and grow in amplitude as they propagate downstream . under this interpretation , the ridge - line would correspond to the region with maximum pressure ( implying a larger concentration of particles and magnetic field ) and would oscillate with time across the jet .
on the other hand , the motion of the bright ridge - line can be interpreted as an alfvn wave produced by a perturbation of the magnetic field in a magnetically dominated flow .
in addition , the radio components often associated to differential injections of plasma in the jet produce shocks that travel through the jet and may change the brightness configuration of the jet at the affected regions .
altogether , jets are an extremely complex scenario that requires further modeling and observing efforts .
this complexity requires accurate observations .
thus , only by means of high resolution observations using the space antenna _ radioastron _ will we be able to make a step further in the understanding of the physics involved in the processes described in this contribution .
the first data obtained from these observations is being currently analysed .
i obtained financial support by the spanish `` ministerio de ciencia e innovacin '' ( micinn ) grants aya2010 - 21322-c03 - 01 and aya2010 - 21097-c03 - 01 through an extension of these grants that had expired nine months before the meeting .
no new call had been open on time and during 2014 we experienced significant lack of support .
i want to thank y.y .
kovalev , a.p .
lobanov , p.e .
hardee and i.agudo for their contribution to this project .
i acknowledge desire igual - chicn and aina perucho - igual for their support . | extragalactic jets are a common feature of radio - loud active galaxies .
the nature of the observed jets in relation to the bulk flow is still unclear . in particular
it is not clear whether the observations of parsec - scale jets using the very long baseline interferometric technique ( vlbi ) reveal wave - like structures that develop and propagate along the jet , or trace the jet flow itself . in this contribution
i review the evidence collected during the last years showing that the ridge - lines of helical radio - jets do not correspond to observational artifacts .
this conclusion was reached by studying a number of vlbi observations of the radio jet in the quasar s5 0836 + 710 at different frequencies and epochs .
the ridge - line of the emission in the jet coincides at all frequencies within the errors .
moreover , small differences between the ridge - lines as observed at different epochs reveal wave - like motion transversal to the jet propagation axis .
i also discuss similar results , albeit with different interpretations , obtained by other authors .
the current challenge is to measure the propagation velocities of these waves and to try to characterise them in terms of simple perturbations or kelvin - helmholtz instability , which would help understanding the physical conditions of the flow where the waves develop .
this problem can only be tackled by high - resolution observations such as those that can be achieved by the space radio - antenna _ |
the perceptron which was first analyzed with statistical mechanics techniques in the seminal paper of gardner @xcite is by now a well - known and standard model in theoretical studies and practical applications in connection with learning and generalization @xcite .
a number of extensions of the perceptron model have been formulated , including many - state and graded - response perceptrons ( e.g. , @xcite ) . here
we present some new extensions allowing for so - called coloured or ashkin - teller type neurons , i.e. , different types of binary neurons at each site possibly having different functions .
the idea of looking at such a model is based upon our recent work on ashkin - teller recurrent neural networks @xcite .
there we showed that for this model with two types of binary neurons interacting through a four - neuron term and equipped with a hebb learning rule , both the thermodynamic and dynamic properties suggest that such a model can be more efficient than a sum of two hopfield models .
for example , the quality of pattern retrieval is enhanced through a larger overlap at higher temperatures and the maximal capacity is increased . for more details and an underlying neurobiological motivation for the introduction of different types of neurons
we refer to @xcite . in the light of these results
an interesting question is whether such a coloured perceptron can still be more efficient than the standard perceptron . in other words
, can it have a larger maximal capacity than the one of a standard perceptron , which is known @xcite to be @xmath0 ( for random uncorrelated patterns ) .
it has been suggested that this number is characteristic for all binary networks independent of the multiplicity of the neuron interactions .
thereby , the capacity is defined as the thermodynamic limit of the ratio of the total number of bits per ( input ) neuron to be stored and the total number of couplings per ( output ) neuron @xcite .
we remark that `` input '' and `` output '' refer specifically to the perceptron case . in the sequel
the maximal capacity of coloured perceptron models is studied using the gardner approach @xcite .
first - step replica - symmetry - breaking effects are evaluated and the analytic results are compared with extensive numerical simulations using various learning algorithms . the rest of this paper is organized as follows . in sec .
ii we introduce two ashkin - teller type perceptron models .
section iii contains the replica theory and determines the maximal capacity by calculating the available volume in the space of couplings both in the replica - symmetric ( sec .
iiia ) and the first - step replica - symmetry - breaking approximation ( sec .
section iv describes the results of numerical simulations with algorithms obtained by generalizing various algorithms for simple perceptrons . in sec .
v we present our conclusions . finally , two appendices contain some technical details of the derivations .
let us first formulate the coloured perceptron models .
we consider @xmath1 input patterns @xmath2 consisting out of two different types of patterns @xmath3 and @xmath4 , and a corresponding set of outputs @xmath5 which are determined by @xmath6 where @xmath7 ( @xmath8 ) are the local fields acting on the patterns @xmath9 , @xmath10 and their product @xmath11 respectively @xmath12 both types of input patterns and their corresponding outputs are supposed to be independent identically distributed random variables ( iidrv ) taking the values @xmath13 or @xmath14 with probability @xmath15 .
the set of three equations ( [ e.s1])-([e.s3 ] ) defines a mapping of the inputs @xmath16 onto the corresponding outputs @xmath17 .
we call it model i. we remark that the specific form of the equations ( [ e.s1])-([e.s3 ] ) is related to the transition probabilities for a spin - flip in the dynamics @xcite . a second model , denoted by ii , is defined by considering only the two equations ( [ e.s1 ] ) and ( [ e.s2 ] ) . when @xmath18 and @xmath19 then the relations ( [ e.s1])-([e.s2 ] ) are satisfied by two ( out of the four possible ) values of the output @xmath20 , otherwise model ii gives the same output as model i. in other words , due to the presence of the @xmath21 and @xmath22 in the gain functions , model ii contains more freedom and , strictly speaking , it is not a mapping
. the sequential dynamics of these two models has been studied in the case of low loading with the hebb rule and shown to lead to the same equilibrium behaviour @xcite . however , this is not guaranteed here since we are concerned with optimal couplings maximizing the loading capacity .
at this point we remark that when all @xmath23 are equal to zero we find back two independent standard binary perceptron models . in the sequel
we take the couplings to satisfy the spherical constraint @xmath24 .
the coloured perceptron is trained to store correctly @xmath25 patterns with @xmath26 the loading capacity .
the factor @xmath27 follows naturally from the definition of capacity given in the introduction .
a pattern is stored correctly when the so - called aligning field @xcite is bigger than a certain constant @xmath28 whereby the latter indicates the stability .
it is a measure for the size of the basin of attraction of that pattern .
specifically we require that @xmath29 with @xmath30 denoting the configurations in the space of interactions . for @xmath31
all patterns that satisfy equations ( [ e.s1])-([e.s3 ] ) also satisfy ( [ eq : afieldxi])-([eq : afieldxieta ] ) .
we remark that for model ii the last inequality is superfluous . ' '' '' ' '' '' the aim is then to determine the maximal value of the loading @xmath26 for which couplings satisfying ( [ eq : afieldxi])-([eq : afieldxieta ] ) can still be found .
in particular , the question whether this model can be more efficient than the existing two - state models is relevant . following refs .
@xcite we formulate the problem as an energy minimization in the space of couplings with the formal energy function defined as @xmath32\ , .
\label{eq : energyi}\end{aligned}\ ] ] we remark that for model ii the third @xmath33-factor is absent .
the quantity above counts the number of weakly embedded patterns , i.e. , the patterns with stability less than @xmath34 .
therefore , the minimal energy gives the minimal number of patterns that are stored incorrectly .
this number is zero below a maximal storage capacity @xmath35 .
the basic quantity to start from is the partition function @xmath36 \rangle_{\{j\}}\\ \langle ... \rangle_{\{j\ } } & = & \int \prod_i dj_i \prod _ r \delta \left(\sum_i ( j_i^{(r)})^2 - n \right) ... ~ , \end{aligned}\ ] ] with @xmath37 the inverse temperature . as usual
it is @xmath38 which is assumed to be a self - averaging extensive quantity @xcite .
the related free energy per site @xmath39 is equal , in the limit @xmath40 , to @xmath41\rangle_{\{j\ } } } { nz(\beta ) } \ , , \label{eq : flimit}\ ] ] which is the minimal fraction of wrong patterns ( recall eq .
( [ eq : energyi ] ) ) . in order to perform the average over the disorder in the input patterns @xmath42 and the corresponding outputs @xmath43 we employ the replica method .
the calculations proceed in a standard way although the technical details are much more complex . introducing the order parameters @xmath44 , with @xmath8 and @xmath45 we write following @xcite @xmath46 \int\prod_{r ' } { \rm d}x^{r'\gamma } \exp\left\{\sum_{r ' } ix^{r'\gamma}\lambda^{r'\gamma } \right.\right.\right . \nonumber\\ & - & \frac{1}{2}\left[\left(x^{1\gamma}+x^{3\gamma}\right)^2 + \left(x^{2\gamma}+x^{3\gamma}\right)^2 + \left(x^{1\gamma}+x^{2\gamma}\right)^2\right ] \nonumber\\ & -&\sum_{\tau>\gamma}\left[\left(x^{1\gamma}+x^{3\gamma}\right ) \left(x^{1\tau}+x^{3\tau}\right)q_{\gamma\tau}^{(1 ) }
+ \left(x^{2\gamma}+x^{3\gamma}\right ) \left(x^{2\tau}+x^{3\tau}\right)q_{\gamma\tau}^{(2)}\right .
+ \left.\left.\left.\left.\left(x^{1\gamma}+x^{2\gamma}\right ) \left(x^{1\tau}+x^{2\tau}\right)q_{\gamma\tau}^{(3 ) } \right ] \rule{0cm}{0.5cm}\right\ } \rule{0cm}{0.6cm}\right ) \rule{0cm}{0.7cm}\right\}\nonumber
\\ g_1&=&\ln\left\{\int\prod_{r,\gamma}\left({\rm d } j^{(r)\gamma}\right ) \exp\left[i\sum_{r,\gamma}\epsilon_\gamma^r((j^{(r)\gamma})^2 - 1)- i\sum_{r,\gamma,\tau >
\gamma}\phi_{\gamma\tau}^r j^{(r)\gamma}j^{(r)\tau } \right]\right\},\nonumber\end{aligned}\ ] ] 2 where @xmath47 denotes the average over the patterns , @xmath48 for model i and @xmath49 for model ii . because of the latter we remark that for model ii the formula for @xmath50 can be simplified : the integrals with respect to @xmath51 and @xmath52 are not present and thus @xmath52 , @xmath53 and @xmath51 have to be set to zero .
because of this simplification we only outline explicitly the calculations for model ii in the sequel .
the corresponding formulas for model i can be found in appendix b. we continue by making the replica - symmetric ( rs ) anzatz @xmath54 .
moreover , for convenience , we set @xmath55 . the latter is justified for model i because of the symmetry present in this model .
furthermore , since we are going to take all @xmath56 in the gardner - derrida analysis anyway , we keep this equality also for model ii .
taking then the limits @xmath40 , @xmath57 and @xmath58 we arrive , in the case of model ii , at @xmath59 with @xmath60 where @xmath61 , @xmath62 is a modified gaussian measure , @xmath63 and @xmath64 takes those values that minimize @xmath65 , the available ' '' '' ' '' '' volume in the space of couplings . for the corresponding expression in the case of model
i we refer to appendix b. taking @xmath66 and supposing that the maximal capacity , @xmath67 , is signaled by the gardner - like criterion @xmath68 we obtain @xmath69 this maximal capacity as a function of @xmath70 is shown for both models in figs .
[ res1 ] and [ res2 ] as a full line . for model
i we obtain , e.g. , @xmath71 , a value that is smaller than the gardner capacity for the simple perceptron . for model ii however , we get the interesting result that @xmath72 .
it is straightforward to show geometrically that learning almost antiparallel patterns , i.e. , patterns satisfying @xmath73 results in a splitting of the space of couplings into disconnected regions .
this suggests that rs is broken and , consequently , the results for @xmath74 found in sec .
iiia are only upperbounds for the true capacity .
therefore , we want to improve the rs results by applying the first step of parisi s replica - symmetry - breaking ( rsb ) scheme ( e.g. , @xcite ) .
so , we assume that the @xmath75 in equation ( [ zpn ] ) have the following matrix block structure @xmath76 where @xmath77 is the size of the matrix @xmath75 , @xmath78 is the number of diagonal blocks and int(@xmath79 ) denotes the integer part of @xmath79 . for model
ii we take @xmath80 reflecting the symmetry of this model . for model
i we repeat that all @xmath81 s can be taken equal .
we then consider the limits @xmath82 and @xmath58 in such a way that @xmath83 , with @xmath84 , remains finite . after a tedious calculation
we arrive at the following expression for the rsb1 maximal capacity for model ii @xmath85 2 with @xmath86 and d@xmath87 a gaussian measure .
the explicit form of the function @xmath88 can be found in appendix a. an analogous form for model i is written down in appendix b. the results are presented in figs .
[ res1 ] and [ res2 ] as full lines .
as expected they lie below the rs results confirming the breaking of rs , e.g. , @xmath89 for model i and @xmath90 for model ii .
we remark that the breaking for model ii is stronger than for model i , the reason being that model ii allows more freedom as explained in the introduction .
finally , on the basis of results in the literature for the simple perceptron @xcite , @xcite we expect that the rsb1 results are very close to the exact ones .
this is further examined by performing numerical simulations as described in the following section .
the idea of these simulations is to train the network with a certain learning algorithm in order to learn as many random patterns as possible .
the main technical difficulties are to find an efficient algorithm and prove its convergence .
we have tried to generalize various algorithms proposed for simple perceptrons @xcite .
the most effective ones appeared to be some particular generalization of the adaptive gardner algorithm @xcite and the adatron algorithm @xcite . in the sequel
we only report on the results obtained with these two algorithms .
we remark that we have chosen @xmath91 in all simulations .
one of the algorithms that has demonstrated its efficiency and for which convergence has been shown in the case of the standard perceptron is given in ref .
it is an adaptive version of the original algorithm proposed by gardner @xcite . using heuristic arguments presented in @xcite we have constructed for the coloured perceptron model ii the following analogous learning rule @xmath92 \,.\end{aligned}\ ] ] ' '' '' ' '' '' the form of the algorithm for model
i is a bit different and given in appendix b. this algorithm should be carried out sequentially over the patterns and sequentially or parallel over the couplings as long as one of the arguments of the @xmath33 functions is positive .
it appears to have the characteristics of the most efficient , non - linear algorithm discussed in @xcite . using this learning rule
we have trained networks of sizes @xmath93 sites ( depending on the value of @xmath70 ) in order to store perfectly as many randomly chosen patterns as possible . for each value of @xmath70
we have calculated the maximal capacity for different @xmath94 and extrapolated the results to @xmath95 .
results for a given value of @xmath70 and @xmath94 are averages over 1000 samples . as shown in figs .
[ res1 ] and [ res2 ] this algorithm performs especially well for small values of @xmath70 for both the models i and ii .
the second algorithm we report on is the adatron algorithm @xcite which works in a different way . instead of searching the maximal capacity for a given stability it tries to find the maximal stability for a given capacity .
the derivation of this algorithm and a proof of its convergence are based upon the assumption that the problem can be formulated as a quadratic optimization with linear constraints @xcite .
such a formulation can not be given for the coloured perceptron model , because the three different types of couplings have to be normalized independently and because the stability conditions ( [ eq : afieldxi])-([eq : afieldeta ] ) are more complex .
hence , a straightforward generalization similar to the one for the potts model @xcite is not possible .
below we describe a learning rule that tries to incorporate the ideas of the adatron approach .
we assume that the couplings can be written in the form ( cfr.,@xcite and references therein ) @xmath96 where @xmath97 ( @xmath8 ) are the so called embedding strengths of pattern @xmath98 .
then , in the case of model ii the couplings are updated by modifying @xmath97 with the following increments @xmath99 this is done sequentially over the patterns .
we remark that again the algorithm for model i is somewhat different ( see appendix b ) . for each value of the capacity we have considered system sizes @xmath100 and extrapolated the results to @xmath95 .
the best results were obtained for a learning rate @xmath101 .
results for each size are averages over 1000 samples . for small values of the capacity
the algorithm gives better results , both in the case of models i and ii than the first algorithm we have discussed , as shown in figs .
[ res1 ] and [ res2 ] . for larger values of the capacity , however , it performs worse .
the results for the adatron algorithm are displayed only in the region where they are better than the results for the gardner algorithm .
we remark that the numerical simulations with the different algorithms give different results and that we have not shown their convergence analytically such that , in principle , the values for @xmath102 obtained here are lower bounds . looking at figs .
[ res1 ] and [ res2 ] in more detail we see that for the whole range of @xmath70 the values of the maximal capacity in model ii are larger than those of a standard binary perceptron . for @xmath103 ,
e.g. , the simulations give @xmath104 , which is bigger than the maximal capacity of the binary perceptron model @xcite and the binary many - neuron interaction model @xcite , both of which have @xmath0 . for model
i the maximal capacity at @xmath103 found by simulations is @xmath105 .
in this work we have calculated the maximal capacity per number of couplings for two coloured perceptron models . compared with the standard perceptron these models have two neuronal variables per site and a local field that contains higher order neuron terms .
the ' '' '' ' '' '' + + method used is a generalization of the gardner approach and both the rs and rsb1 results have been discussed .
we expect that the latter give very close upperbounds for the exact values .
extensive numerical simulations have been performed for finite systems and extrapolated to @xmath95 . the adaptive gardner algorithm and the adatron algorithm
give the best , but different results .
hence , the results of the simulations can be considered only as lower bounds for the exact maximal capacity .
additional work looking for improved algorithms would be welcome .
comparing both the rsb1 results and the results from numerical simulations we conclude that they are in good agreement . for bigger values of @xmath70
they even completely coincide . for model
i we find that at @xmath103 the maximal capacity satisfies @xmath106 .
this suggests that it is equal to the maximal capacity of the @xmath107-potts perceptron , i.e. , @xmath108 ( after appropriate rescaling of the latter @xcite ) .
this would parallel the situation for hebb learning @xcite . for model
ii we have for @xmath103 that @xmath109 , which is larger than the maximal capacity of the standard binary perceptron .
this is due to the fact that model ii is not a strict mapping such that it allows for more freedom in the determination of the couplings .
the authors would like to thank m. bouten and j. van mourik for critical discussions .
the function @xmath88 in formula ( [ alpharsb1 ] ) reads @xmath110 \nonumber\\ & + & \frac{1}{2c_1}e^{\varepsilon_2 } \int_{-\infty}^{\frac{c1}{c}(u_1-\delta_2 ) } { \rm d}s\left[1+{\rm erf}\left(\sqrt{\frac{3r}{2c^2 } } \left(-\frac{x_2}{\sqrt{3r}}+\delta_2 + \frac{c}{c_1}s\right)\right)\right ] \nonumber\\ & + & \frac{1}{2c_2}e^{\phi_2 } \int_{-\infty}^{-\frac{c2}{c}(u_1-\gamma_2 ) } { \rm d}s\left[1+{\rm erf}\left(\sqrt{\frac{3r}{2c^2 } } \left(\frac{x_2}{\sqrt{3r}}-\gamma_2 + \frac{c}{c_2}s\right)\right)\right ] \nonumber\\ & + & \frac{1}{2c_2}e^{\phi_3}\int_{-\infty}^{-\frac{c2}{c}(u_1-\gamma_3 ) } { \rm d}s\left[1+{\rm erf}\left(\sqrt{\frac{3r}{2c^2 } } \left(-\frac{x_3}{\sqrt{3r}}-\gamma_3 + \frac{c}{c_2}s\right)\right)\right ] \nonumber\\ & + & \frac{1}{2c'}e^{d_1}\int_{-\infty}^{-\frac{u_1}{c ' } } { \rm d}s\left[{\rm erf}\left(\sqrt{\frac{3r}{2 } } \left(\frac{x_3}{\sqrt{3r}}-b_1-\frac{1}{c'}s\right)\right ) + { \rm erf}\left(\sqrt{\frac{3r}{2}}\left(-\frac{x_2}{\sqrt{3r}}-b_1- \frac{1}{c'}s\right)\right)\right ] \nonumber\\ & + & \frac{1}{2}\int_{-\infty}^{u_1 } { \rm d}s\left[{\rm erf}\left(\sqrt{\frac{3r}{2 } } \left(\frac{x_2}{\sqrt{3r}}-s\right)\right ) + { \rm erf}\left(-\sqrt{\frac{3r}{2}}\left(\frac{x_3}{\sqrt{3r } } + s\right)\right)\right]\nonumber\end{aligned}\ ] ] 2 with d@xmath111 a gaussian measure and @xmath112 ' '' '' ' '' '' @xmath113
for model i the calculations are very similar . some resulting expressions , however , have a somewhat different structure . for completeness we write down these expressions here . finally , the learning algorithms for model i differ in the way that the couplings @xmath122 and @xmath123 are updated .
we have for the adaptive gardner algorithm @xmath124 \nonumber\\ j_i^{(2)}&\rightarrow&j_i^{(2)}+\eta_0^{\mu}\eta_i^{\mu}\frac{1}{2 } \left[\left(\kappa_\eta-\lambda_\eta^\mu\right)\right .
\theta\left(\kappa_\eta-\lambda_\eta^\mu\right ) \nonumber\\ & + & \left(\kappa_{\xi\eta}-\lambda_{\xi\eta}^\mu\right)\left .
\theta\left(\kappa_{\xi\eta}-\lambda_{\xi\eta}^\mu\right)\right ] \nonumber\end{aligned}\ ] ] instead of ( [ j1per ] ) and ( [ j2per ] ) and for the adatron algorithm we take @xmath125 instead of ( [ x1ada ] ) and ( [ x2ada ] ) . 99 e. gardner j. , phys .
a * 21 * , 257 ( 1988 ) .
j. hertz , a. krogh and r. g. palmer , _ introduction to the theory of neural computation _
( addison - wesley , redwood city 1991 ) .
b. mller , j. reinhardt and m. t. strickland , _ neural networks : an introduction _
( springer , berlin 1995 ) .
m. opper and w. kinzel , _ models of neural networks iii _ ed .
e. domany , j. l. van hemmen and k. schulten , 151 ( springer , new - york 1996 ) .
d. saad , ed._on - line learning in neural networks _
( cambridge university press 1998 ) .
j. p. nadal and a. rau , j. phys .
i france * 1 * , 1109 ( 1991 ) . f. gerl and u. krey , j. phys .
a * 27 * , 7353 ( 1994 ) .
d. a. khring , j. phys .
france * 51 * , 145 ( 1990 ) .
s. mertens , h. m. khler and s. bs , j. phys .
a * 24 * , 4941 ( 1991 ) .
d. boll , p. dupont and j. van mourik , europhys . lett . * 15 * , 893 ( 1991 ) .
d. boll , r. khn and j. van mourik , j. phys .
a * 26 * , 3149 ( 1993 ) .
d. boll and p. kozowski , j. phys .
a * 31 * , 6319 ( 1998 ) .
d. boll and p. kozowski , j. phys .
a * 32 * , 8577 ( 1999 ) .
e. gardner and b. derrida , j. phys . a * 21 * , 271 ( 1988 ) .
w. whyte and d. sherrington , j. phys .
a * 29 * , 3063 ( 1996 ) .
d. boll and r. erichsen , jr phys .
e * 59 * , 3386 ( 1999 ) .
w. krauth and m. mzard , j. phys .
a * 20 * , l745 ( 1987 ) .
l. f. abbott and t. b. kepler , j. phys .
a * 22 * , l711 ( 1989 ) .
j. k. anlauf and m. biehl , europhys
. lett . * 10 * , 687 ( 1989 ) .
j. imhoff , j. phys .
a * 28 * , 2173 ( 1995 ) .
m. mzard , g. parisi and m. a. virasoro , _ spin glass theory and beyond _
( world scientific , singapore 1987 ) .
p. kozowski , phd thesis k.u.leuven , belgium , in preparation . | ashkin - teller type perceptron models are introduced .
their maximal capacity per number of couplings is calculated within a first - step replica - symmetry - breaking gardner approach .
the results are compared with extensive numerical simulations using several algorithms . 2 |
there is plenty evidence for magnetic fields on the scale of the earth , the solar system , the interstellar medium , galaxies and clusters of galaxies [ for the extragalactic magnetic field see review by kronberg ( 1994 , k94 ) and references therein ] .
we still do nt know however whether any larger scale magnetic fields exist .
there are various upper limits on the rms value of the cosmological magnetic field , but the upper limits are still too high to exclude any relevance to magnetic fields observed on smaller scales .
cosmological magnetic fields might help to resolve the bothersome question of the origin of the @xmath17 g magnetic field which exists on galactic scales or fields of similar size seen on the scales of galaxy clusters . a primordial ( pre galaxy formation )
magnetic field amplified by the gravitational collapse , and then by processes like differential rotation and dynamo amplification ( parker 1979 ) , can serve to seed these observed smaller scale magnetic fields .
the most stringent upper limits to date for magnetic fields on cosmological scales come from three different sources : 1 . big bang baryonic nucleosynthesis ( bbn ) .
+ magnetic fields that existed during the bbn epoch would affect the expansion rate , the reaction rates , the electron density ( and possibly the space - time geometry ) . by taking all these effects into account in the calculation of the element abundances , and then comparing the results with the observed abundances one can set limits on the magnetic field amplitude .
the limits for homogeneous magnetic fields on scales @xmath18 ( distances are quoted in comoving coordinates ) and up to the bbn horizon size ( @xmath19 ) are @xmath20 g in terms of today s values ( grasso & rubinstein 1995 , 1996 ; cheng , olinto , schramm , & truran 1996 ) .
the relevance of these limits for the maximum value of magnetic field seeds on the subgalactic scale is obvious , but in order for these limits to be relevant to the intergalactic magnetic field , further assumptions about the super - horizon magnetic field power spectrum and the magnetic field generation epoch should be made .
2 . the cmb radiation . + the presence of magnetic fields during the time of decoupling causes anisotropy , different expansion rates in different directions ( zeldovich & novikov 1975 ) . masden ( 1989 ) provides a useful formula for the connection between the measured temperature fluctuations on a given scale at @xmath21 , the decoupling time , and the limits on the equivalent scale today . for a magnetic field energy density much smaller than the matter energy density ,
the limit is @xmath22 for typical values of @xmath23 ( the hubble constant in units of @xmath5 km s@xmath6 mpc@xmath6 ) , temperature fluctuations of @xmath24 , @xmath25 and @xmath26 the limit becomes @xmath27 g .
recently , barrow , frreira , & silk ( 1997 ) used the cobe 4-year data to constrain the magnetic field on the horizon scale by @xmath28 ng .
3 . multipole expansion of @xmath0 observations .
+ passing through a magnetoionic medium , polarized light undergoes rotation of the polarization vector ( faraday rotation , for definition see [ sec : corr ] ) .
the rotation amount is proportional to the dot product of the magnetic field and the light propagation direction , and to the distance the light travels through the medium . if any cosmological magnetic field exists , the farther away the source , the more the polarized light component is rotated .
we refer to this as the monopole " term .
this term exists even if the field has no preferred direction on the horizon scale , namely even without breaking the isotropy hypothesis .
+ if there is a preferred direction to the field on a cosmological scale , a dipole will show up in the @xmath0 measurements across the sky , due to the different sign of the dot product ( and thus the different rotation direction ) .
so far attempts have been made only to identify a monopole or a dipole in the magnetic field for scales of @xmath29 ( kronberg 1976 ) and @xmath30 ( valle 1990 ) .
the search for a dipole signature in the faraday rotation values for a sample of extra - galactic sources ( qso ) yields limits of @xmath31 ng depending on the assumed cosmology ( kronberg & simard - normandin 1976 , kronberg 1976 ) .
valle ( 1990 ) tried to estimate the limits on the dipole and the monopole terms and concluded that an all - prevailing ( up to @xmath32 ) field must be smaller than @xmath33 $ ] ng . other potential methods of detecting a possible cosmological magnetic field include distortion of the acoustic ( doppler " ) peak in the cmb power spectrum ( adams , danielsson , grasso , & rubinstein 1996 ) and faraday rotation of the cmb radiation polarized components ( loeb & kosowsky 1996 ) .
these methods where not designed to probe any magnetic field generated after recombination .
their implementation is still pending on the upcoming measurements of the cmb fluctuations and polarization .
a method that does probe magnetic fields generated in the post recombination era , relies on cosmic ray ( cr ) detection .
the effect of magnetic fields on high energy crs ( @xmath34 ev ) is twofold .
it will first alter the energy distribution of crs ( lee , olinto , & sigl 1995 , waxman & miralda - escud 1996 ) and then , if the cr direction is identified and attributed to a known region or source , the magnetic field either smears the directionality ( for field coherent scales much smaller than the distance to the source ) or deflects the cr direction from aligning with the source direction .
current limits from this last effect are either very weak or non - existing
. theories for primordial magnetic fields in the framework of structure evolution paradigms have been suggested by a number of authors .
magnetic fields on cosmological scales emerge in the framework of these theories in one of three epochs : the inflation era , the plasma era , and the post - recombination era . in the inflation era ,
magnetic fields form due to quantum mechanical processes ( turner & widrow 1988 ; quashnock , loeb , & spergel 1989 , vachaspati 1991 , ratra 1992a , 1992b , dolgov & silk 1993 ) or possibly magneto - hydrodynamics ( brandenburg , enquist , & olesen 1996 ) . during the plasma era , prior to recombination , magnetic fields form due to vorticity caused by the mass difference between the electron and the proton ( harrison 1973 , but see rees 1987 ) , or where magnetic field fluctuation survival ( no vorticity assumed ) depends on the fluctuation scale ( tajima 1992 ) .
other authors have pointed out that even in the post recombination era ( but before galaxy formation ) magnetic fields can be generated by tidal torques ( zweibel 1988 ) , or by cosmic strings wakes ( ostriker & thompson 1987 , thompson 1990 ) .
magnetic field generated in the process of structure formation itself , due to falling matter ( pudritz & silk 1989 ) , or star burst regions , are less likely to feed back into the intergalactic space and to affect the cosmological magnetic field .
nevertheless , proposals in this spirit were also considered by taking into account wind driven plasma which fills intergalactic space with magnetic fields .
whatever the origin of cosmological magnetic fields , each suggested theory , specifies the magnetic field amplitude and power spectrum shape .
thus , each of the theories can in principle become refutable if we could measure , or put limits on , the normalized power spectrum of the cosmological magnetic field .
this challenge and the realization that with existing or upcoming data , it will be possible to measure the magnetic field power spectrum has led us to develop a method for doing so .
we argue that much better estimates ( or upper limits ) for the cosmological magnetic field can be calculated by using the rm correlation matrix .
all previous attempts to limit the cosmological magnetic field dipole on very large scales ( @xmath35 ) , suffer from two drawbacks . from a theoretical point of view
it is difficult to reconcile the necessary anisotropy such a dipole imposes with other evidence which suggest global isotropy . from a practical point of view
, this test is confined to one scale , and does nt allow measurements over a whole range . only by extending the measurement to a whole range , can one estimate the full power spectrum ( ps ) . although contribution from a random walk " can be useful even if only a monopole is calculated ( see below )
, the information is partial , inferior to the full derivation of the magnetic field by @xmath0 correlations , and sensitive to evolutionary effects .
why is the correlation approach more advantageous than other type of statistics and in particular why is it better than the monopole approach ? in the monopole approach we are looking for a correlation between rm values and the source redshift .
let s pretend at first that all measurements are exact and there are no noise sources ( we ll notice in [ sec : calc ] that these two assumptions are far too optimistic ) .
consider a scale @xmath36 over which the magnetic field is coherent with an rms value of @xmath10 .
a line - of - sight to a source at a distance @xmath37 will cross @xmath38 such regions . there is no correlation in the magnetic field orientation between the regions , and on average each one of them contributes @xmath39 .
the overall contribution due to @xmath40 random walk steps sums up to @xmath41 .
it is thus clear that the larger @xmath37 , the bigger @xmath42 we expect for an ensemble of sources located at @xmath43 .
now let s turn to the correlation approach .
two lines - of - sight to two adjacent sources ( separation angle @xmath44 and assume the same redshift for the two ) are observed for correlation between the @xmath0 values of the polarized light emanating from them .
if @xmath45 the two light rays undergo _ the same _ rotation in each patch of length @xmath36 ( for small @xmath44 we neglect the difference in the @xmath46 term between the magnetic field and the line - of - sight ) .
the ensemble average of the correlation term would then be @xmath47 , with the same proportionality constant as before . in the limit of @xmath48 ,
the two expressions coincide , but for any value of @xmath49 , the signal from the correlation calculation is amplified by the factor @xmath50 . for example , if @xmath51 @xmath52 and @xmath53 the signal is amplified by more than an order of magnitude .
moreover , the correlation provides information about the power spectrum ( different @xmath44 and @xmath37 values ) of the magnetic field that otherwise we would nt be able to obtain .
the correlation also makes it easyer to separate the noise from the signal , unlike the monopole approach .
we will return to this approximation later on , under more realistic considerations ( [ subsec : ston ] ) , when we attempt to evaluate the number of pairs needed in order to establish a certain signal - to - noise ratio at a given scale ( @xmath44 ) .
for now , this simple model demonstrates the advantages of this paper s approach . in order to relate actual data of rm measurements to the magnetic field ,
we begin by considering the connection between the two . in section [ sec : corr ] we introduce the magnetic field correlation tensor and its connection to the rm correlation . in section [ sec : calc ] we make use of these definitions and describe the calculation procedure for the @xmath0 correlation from the raw data to the constraints on the magnetic field power spectrum . a key issue is the estimate of noise from non - cosmological contributions to the @xmath0 .
the impatient reader can turn immediately to [ subsec : ston ] to get a rough estimate for the expected signal - to - noise ratio .
we demonstrate the procedure in the following section ( 4 ) , where we simulate a few realistic examples of cosmological magnetic fields , and exploit them to get rm measurements from which we derive back an estimate for the original power spectrum .
the simulations also allow us to perform a realistic error analysis . in section 5
we discuss prospects for applying the procedure to real data and conclude with our results .
linearly polarized electro - magnetic radiation of frequency @xmath54 traveling a distance @xmath55 through a non - relativistic plasma medium with the magnetic field @xmath56 , is rotated by the angle @xmath57 . in other words ,
the polarization angle is changed by ( lang 1978 eqs .
1 - 268 @xmath58 1 - 270 ) @xmath59 where @xmath60 , @xmath61 , and @xmath62 are the electron charge , mass , and number density respectively ( we use units of @xmath63 ) .
the rotated polarization vector itself is written @xmath64 a pseudo - vector degenerated in rotation of @xmath65 by @xmath66 .
it is related to the stokes parameters via @xmath67 that in turn are expressed by the electric field components perpendicular to the wave propagation direction @xmath68 @xmath69 @xmath70 @xmath71 where time average is explicitly assumed . in the cosmological context ,
some of the quantities are functions of time ( or equivalently redshift or distance ) .
we assume fully ionized gas ( ionization fraction @xmath72 ) and thus the average number density of free electrons is @xmath73 the baryon number density at a specific location is @xmath74 , with @xmath75 the dimensionless density fluctuation .
number densities scale like @xmath76 , where @xmath77 is the scale factor .
the frequency scales like @xmath78 , and under the assumption of flux conservation , @xmath79 scales like @xmath80 .
the distance unit @xmath55 " in a robertson - walker metric is given by @xmath81 expressed via @xmath82 the conformal time . since @xmath83 for the redshift @xmath84 , the overall rotation of the polarized radiation of a source at the redshift @xmath84 , the direction @xmath85 as observed from @xmath86 , ( we set @xmath87 for simplicity ) and at frequency @xmath88 , is given by @xmath89 ( 1+z')^3 { \vec b_0}({\hat q}r ' ) \cdot { \hat q }
{ { \rm d } r ' \over { \sqrt { 1 - kr^{'2 } } } } = { \rm ( { \hat q } , z ) \over \nu_0 ^ 2 } \ , .\ ] ] today s values are all denoted by the subscript 0 " and @xmath90 .
the time dependence of @xmath91 is taken into account later on ( eq . [ eq : ups_xi ] ) .
the two point correlation function of the rm is defined by @xmath92 see nissen & thielheim ( 1975 ) for a similar definition . throughout the paper , @xmath93 is the notation for an ensemble average .
we use the notation @xmath94 and obtain @xmath95(1+z'_1)^3 \int_0^{r_2(z_2 ) } { { \rm d } r'_2
\over { \sqrt { 1 - kr_2^{'2 } } } } \left[1+\delta(\hat{q}_2r_2')\right ] ( 1+z'_2)^3\ ] ] @xmath96 where summation over the spatial components @xmath97 is assumed . the integral ( [ eq : ups_1 ] ) is made out of four terms .
two of the terms ( those involve only one power of @xmath98 ) vanish because @xmath99 and we assume vanishing correlation between the density fluctuations and magnetic fluctuations [ @xmath100 due to the vector nature of @xmath101 and the scalar @xmath98 .
the two remaining terms are @xmath102 @xmath103 and @xmath104 @xmath105 where @xmath106 is the baryonic matter correlation function that we identify with the matter correlation function .
today s value of it is @xmath107 and the missing powers of @xmath108 in eq .
( [ eq : ups_xi ] ) account for the linear evolution of this correlation .
the rm correlation function is the sum @xmath109 .
the last integral of eqs .
( [ eq : ups_1 ] , [ eq : ups_0 ] , [ eq : ups_xi ] ) is the correlation tensor of the magnetic field defined as @xmath110 . in the cosmological case ,
we assume isotropy on scales much smaller than the horizon and confine @xmath111 to be a function of @xmath112 only . being a divergence - free field the magnetic field correlation can be written ( monin & yaglom 1975 ) as a combination of parallel ( @xmath113 ) and perpendicular ( @xmath114 ) functions .
the parallelism and orthogonality are given with respect to the connecting vector : @xmath115 @xmath116{r_i r_j \over r^2 } + c_{\perp}(r ) \delta^k_{ij}\ , , \ ] ] @xmath117 @xmath118{k_j k_l \over k^2 } + c_{\perp}(k ) \delta^k_{jl}\ , , \label{eq : c_pre_par}\ ] ] where @xmath119 is the kronecker @xmath98-function .
in the spectral domain we define the function @xmath120 and express both correlation functions by the function @xmath121 , namely @xmath122 with @xmath123 denoting the spherical bessel function of the @xmath124 order . when people refer to the magnetic power spectrum " they usually mean @xmath125 which
should further be multiplied by @xmath126 in order to get @xmath121 .
we shall hereafter comply with this notation and refer to @xmath125 as the three - dimensional magnetic power spectrum " , @xmath127 .
figure [ fig : b_corr ] show the two functions @xmath128 and @xmath129 in units of @xmath130 for three values of the power index @xmath131 [ @xmath132 . each function
is shown with a minimal gaussian smoothing of @xmath133 . naturally , smaller power indices mean larger correlation length . -1.2truecm
5truecm = 2.7 in -2truecm 0.5truecm = 2.7 in 2truecm = 2.7 in -0truecm going back to the expression ( [ eq : ups_1 ] ) for the rm correlation we can rewrite it as @xmath134 @xmath135({\hat { q}_1 \cdot \hat r ' ) ( \hat{q}_2 \cdot \hat r ' ) } + c_{\perp}(r'){\hat q}_1 \cdot { \hat q}_2 \right\ } \,.\ ] ] as expected , the rm correlation depends only on the relative distance , @xmath136 , and the angle . for comoving distances between two sources at @xmath137 and @xmath138 separated by the angle @xmath44 on the sky we use weinberg s ( 1972 , eq .
14.2.7 ) coordinate transformation , implemented by osmer ( 1981 ) for the @xmath139 case ( @xmath140 the deceleration parameter ) and generalized by boyle ( 1986 ) for other @xmath140 values . for the flat universe case ( @xmath141 ) the comoving relative distance reduces to the simple form @xmath142^{1/2}\ , .\
] ] the distance to a source at the redshift @xmath84 ( no cosmological constant ) is given by ( kolb & turner 1990 , eq .
3.112 ) @xmath143 and can be easily generalized for other cases . in order to estimate the importance of @xmath144 [ eq .
( [ eq : ups_xi ] ) ] we adopt the apm power spectrum ( baugh & efstathiou 1993 ; tadros & efstathiou 1995 ) with the fitting formula @xmath145 the fitting parameters @xmath146 , @xmath147 , @xmath148 , and gaussian smoothing of @xmath133 .
we then derive the real space correlation function by the fourier transform of the smoothed power spectrum . in the range @xmath149 the result
is very similar to the derived apm real space correlation function ( no explicit smoothing ) @xmath150 ( baugh 1995 ) . for this choice
we get ( regardless of the @xmath127 form , and for @xmath151 ) , @xmath152 .
we hence identify @xmath153 hereafter .
we relate the magnetic field normalization to the power spectrum by @xmath154 where @xmath155 the factor @xmath156 is due to the definition of @xmath127 [ as @xmath129 ] , and @xmath158 is the gaussian ( top - hat ) window function of radius @xmath159 , in @xmath160 space , @xmath161 \ , ; \quad w_{th } = { 3 \over ( kr)^3 } [ \sin(kr ) - kr\cos(kr ) ] \,.\ ] ] a good benchmark to use for the normalization range is the cmb limit due to its simplicity and the availability of @xmath162 measurements on many scales ( either today or in the near future ) .
we therefore work in the range of @xmath163 ng throughout this paper .
the observational limits on the magnetic field as deduced from the @xmath0 dipole [ @xmath164 or @xmath165 are thus very different from the limits imposed by cmb fluctuations on @xmath166 scale .
we note that by eq .
( [ eq : b_norm ] ) one can not infer limits on the magnetic field magnitude from one scale to another .
the limits must involve the power spectrum shape .
since no estimate for the latter exists , and since theoretical predictions for it range from power index of @xmath167 to @xmath168 ( ratra 1992a ) , limits on one particular scale hardly limit other scales . -1.5truecm
5truecm = 2.7 in -3truecm 0.5truecm = 2.7 in 2truecm = 2.7 in -1truecm figure [ fig : rm_corr ] show the values for the rm correlation function for pairs of sources at the same redshift , and three different spectral indices .
the normalization of the magnetic field in each case is @xmath169 ng ( @xmath170 ) .
this scale resembles the mean separation between rich clusters .
we notice that for close pairs , high redshift sources are preferable in order to get high signal , while a transition typically occurs at a fraction of a degree where it becomes preferable to use lower redshift sources .
this transition is due to the uncorrelated magnetic fields experienced by high redshift sources with large separation .
only when the two lines - of - sight approach the correlation length , does contribution to the rm correlation becomes substantial .
the different , uncorrelated magnetic fields up to the correlation length , play the role of uncorrelated noise , and dominate the correlated signal from the low redshift segment of the line - of - sight . on the scale @xmath171 ,
the rm correlation signal still stands on @xmath172 depending on the power spectrum .
the raw data consist of @xmath173 extra galactic objects for which polarization measurements are available in @xmath174 wavelengths .
broten , macleod , & valle ( 1988 ) , and oren & wolfe ( 1995 ) have emphasized the need for a careful selection of wavelengths for each observed source . for the current application there are additional special requirements that will become clear when we get to the discussion ( [ sec : discuss ] ) .
+ each of such @xmath174 measurements consists of two types of data : the degree of polarization , and the polarization position angle [ @xmath175 and @xmath176 of eq .
( [ eq : p_vec ] ) ] .
two functions are then constructed which give the dependence of these on wavelength .
the @xmath0 is computed from the latter by using eq .
( [ eq : p_vec ] ) where @xmath177 .
each measured @xmath178 value has an error , @xmath179 , that emerges from the measurement error ( instrumental , ionosphere model , the earth magnetic field model etc . ) and translates to the error in the fit from which the @xmath0 is derived . for typical sources with polarization degree of @xmath180 ( namely @xmath181 of polarized component at a few ghz frequency ) ,
the measurement error in each datum is of the order of a few degrees ( kato , tabara , inoue & aizu 1987 ; simard - normandin , kronberg , button , 1980 ( skb ) ; oren & wolfe 1995 ) . for realistic @xmath182 , this instrumental error translates to a typical error in the @xmath183 fit of @xmath184 and is dwarfed by other noise terms in the procedure ( [ subsec : gal_mask ] , [ subsec : i+f ] ) .
we therefore set @xmath185 from now on .
the measured rms are the sum of a few contributions , and can not be directly plugged into the expression ( [ eq : ups_2 ] ) to evaluate @xmath186 .
+ to begin with , there exists the internal @xmath187 for every one of the @xmath173 extragalactic sources .
then there is the integrated @xmath188 which is presumably due to the cosmological magnetic field ( and the free electrons ) along the line - of - sight to the source .
there may also exist contribution @xmath189 from intervening systems along the line - of - sight ( foreground screen ) either next to the source itself ( the host galaxy of a quasar ) , lyman-@xmath190 systems , galaxies , or clusters of galaxies . before this combined signal gets to the detector it still has to go through the galactic magnetic field , where it is rotated once more by the amount @xmath191 . the final measurement
, @xmath178 is thus the linear sum @xmath192 where @xmath193 is the redshift - vector [ actually translated to a distance vector ( 2 ) ] for the @xmath124 source . since we are interested in the cosmological contribution to the @xmath0 [ the second term in eq .
( [ eq : lin_sum ] ) ] , the first step should involve cleaning " the measured signal of all irrelevant extra contributions .
we shall attempt to perform this cleaning in a statistical way .
there are two alternatives to assess the galactic contribution to the measured @xmath0 .
the two ways differ by the population sample used for the assessment . if an independent population at the outskirts of the galaxy exists , for which @xmath0 can be measured , this population can be used to map out the galactic @xmath0 .
we shall hereafter use the term `` mask source population '' for the set of objects by which we map the galactic contribution to the rm .
an estimate for the thickness of the galactic magnetic layer is @xmath163 kpc ( simard - normandin & kronberg 1980 ) .
that means that apart from the galactic center direction , sources of distances that satisfy @xmath194 kpc ( @xmath195 kpc for @xmath196 ) are located beyond the reynolds layer and fully probe the galactic contribution to the rm .
we hence consider two possibilities for the mask source population .
one natural candidate for this role is the pulsar population , for which @xmath0s are available . in order to make use of the pulsars we need to find a subset of them that reside in the outer part of the milky - way magnetic layer . + a class of pulsars that is especially appropriate for the task of probing the galactic rm contribution is the millisecond pulsars .
this population of old pulsars is believed to reside far out of the galactic plane .
the number density of millisecond pulsars as implied by a number of surveys at high galactic latitude ( @xmath197 sensitivity at @xmath198 ghz ) ranges between @xmath199 deg@xmath200 , @xmath201 pulsars over the sky ( foster , cadwell , wolszczan , & anderson 1995 ; camilo , nice , & taylor 1996 ) .
the alternative mask population for assessment of the galactic @xmath0 contribution is a sub - set of the closest extragalactic sources .
the disadvantage of using this population is the need for a large enough source number at low enough redshift , to allow differentiation between the cosmological contribution and the galactic one .
exploiting extragalactic sources for the resolution of the galactic contribution affects only slightly the amount of cosmological contribution correlation , as long as the nearby sources are taken within @xmath202 , since the bulk of the sources is at @xmath203 ( [ sec : testing ] ) . in order to use either one of the galactic mask populations ,
its number density should exceed a certain minimal number density .
we now attempt to assess this minimal number . +
the galactic @xmath0 contribution is a two - dimensional field with @xmath0 values .
the data on the other hand are made of distinctive sources located at discrete directions .
the lines - of - sight to the galactic mask population , do not necessarily coincide with the line - of - sight direction towards the extragalactic sources . the way to circumvent this difficulty is by introducing a _ smoothed _ version of the galactic @xmath0 map , namely the field @xmath204 .
the two fields are connected via the smoothing angle @xmath205 , and by a specific choice of a gaussian smoothing for @xmath206 measurements of the galactic @xmath207 located at the directions @xmath208 @xmath209 \over \sum\limits_{i=1}^{n_g } \exp \left[- { ( { \hat z}_i - \hat{z})^2 \over 2 \theta_s^2 } \right ] } \,.\ ] ] when implemented using real data , the smoothing involves weights by measurement errors as well . for the purpose of this paper , we ignore these weights to avoid unnecessary complication . using a smoothed field for the galactic @xmath0 contribution , introduces random noise , @xmath210 , that can be directly evaluated by the very same @xmath206 sample via @xmath211 4truecm 0.5truecm 1.0truecm 4truecm the noise can be a function of position , it is difficult however to account for a position - dependent noise with low number of @xmath206 .
there are a few estimates in the literature for the noise level , @xmath210 , on different smoothing scales .
+ an upper limit for @xmath210 is given in minter & spangler ( 1996 ) , who calculate the observed `` structure function '' , @xmath212 which is the same as eq .
( [ eq : gal_res ] ) but with an arbitrary lag @xmath44 and the unsmoothed values of the rm .
they conclude that @xmath213rad@xmath214m@xmath215 , and drops steeper for @xmath216 .
the value of the @xmath212 is an upper limit for @xmath217 as calculated here , due to the smoothing we apply ( and was not applied in the @xmath212 calculation ) .
simonetti , cordes , & spangler ( 1984 ) calculate @xmath218 rad@xmath219m@xmath215 , depending on the galactic latitude , for much larger angular scales of @xmath220 ! ( linear scale ) .
+ simonetti & cordes ( 1986 ) corroborate these results and consider even larger angular scales to obtain typically @xmath221rad@xmath219m@xmath215 on scales less than @xmath222 .
+ oren & wolfe ( 1995 ) calculate a very similar quantity to @xmath210 ( with a varying top hat window instead of a fixed gaussian ) , and obtain a typical variance of @xmath223 rad@xmath219m@xmath215 on @xmath224 scales .
+ all of these measurements suggest that for a mask population with number density of about one source per @xmath225deg.@xmath219 , the galactic contribution to the rm ( @xmath226 ) can be resolved with a @xmath227 accuracy of @xmath228 .
that means that an isotropic coverage of about @xmath229 mask sources on the sky is enough for achieving this noise level .
+ damage control " for ignoring the position dependence can be devised by comparing the clean " rm field dipole and quadrupole to the galactic direction of the two .
these first two multipole should be the most prominent signature of the galactic rm mask .
we go back to the suggested mask populations and examine whether they are able to fulfill the requirement of the minimal number density . + the largest pulsar catalog to date [ the electronic version of taylor , manchester , & lyne ( 1993 ) ] consists of @xmath230 pulsars out of which @xmath231 have rm measurements .
this catalog was not compiled with a specific emphasis for millisecond pulsars , and therefore only @xmath232 pulsars with rm values lie in galactic latitude @xmath233 . the softer condition @xmath234 ( considering an rm layer " of @xmath235 kpc ) is fulfilled for @xmath236 pulsars with both rm registered value and estimated distance @xmath37 ( taylor & cordes 1993 ) .
+ however , these @xmath236 pulsars are still not distributed isotropically across the sky , and tend to concentrate near the galactic plane . a way to avoid this anisotropy is to compile a sample of millisecond pulsars , that lie far away from the galactic plane .
this population has a number density that exceeds the necessary @xmath237 deg.@xmath200 by at least a factor of two and it may be as high as seven times denser than the minimal value .
this population can be detected for rm ( fernando camilo , private communication , 1997 ) ms , and five for which @xmath238 ms . the smallest polarized flux at @xmath239ghz for a pulsar with a registered rm is @xmath240mjy ] and ensures the @xmath241 noise level .
this kind of sample is ideal for the current paper analysis , as it does not contain any cosmological contribution or internal rm contribution .
the main disadvantage of it is that it does not exist yet . in considering the extragalactic mask population ; for the @xmath242 sources needed to resolve the galactic contribution , it is sufficient to have a sample of @xmath243mjy sensitivity ( at a few ghz ) up to @xmath244 [ see loan , wall , & lahav ( 1997 ) for the number density of radio sources with flux @xmath245 mjy , and dunlop & peacock ( 1990 ) for the combined flat / steep spectrum luminosity function for extrapolation to the @xmath243mjy limit ] . for this paper we follow oren & wolfe ( 1995 ) and use an extragalactic sample of rm measurements as an upper limit for the noise that appears by the galactic rm removal procedure .
the sample we use consists of @xmath246 sources as listed by skb .
the skb catalog allows us to realistically mimic the galactic mask , and is an overestimate since we ascribe all the rm of the sources to the galactic contribution .
moreover , unlike oren & wolfe we do not exclude outliers " .
this use aims to provide a realistic model of the galactic mask contribution , and allows us to demonstrate the method power . for our purpose
it should not be taken literally as the true galactic mask contribution ( even though oren & wolfe did consider it that way ) because no segregation by redshift was applied to choose mask sources from the skb list .
we neither attempted to evaluate the internal rm ( by means explained in
[ subsec : i+f ] ) in order to minimize the @xmath210 value .
+ the skb source locations ( projected , and therefore concentrated toward the circumference ) are marked as black dots on figure [ fig : maskmap ] and shown on top of the smoothed galactic @xmath0 field with smoothing scale of @xmath247 .
the smooth map is shown in celestial coordinates .
figure [ fig : res_mask ] shows the rm residuals as functions of galactic latitude and longitude , and figure [ fig : maskres_map ] shows the contour map of the smoothed residual field . both figures ( [ fig : res_mask ] & [ fig : maskres_map ] )
show hardly any longitude dependence of the residuals ( in spite of the solar system off - center position ) , and bigger residuals in the @xmath248 strip about the galactic plane ( the same conclusion of oren & wolfe 1995 ) . in an all - sky coverage of sources ,
this residual deviation should be of some worry , but as we shall see ( [ subsec : all_sky ] ) , it does nt bias the result for the most likely power spectrum as calculated by the bayesian analysis .
if the analysis is confined to high latitudes , then no systematic error due to the galactic mask is expected ( [ sec : testing ] ) .
for a few square degrees area , we notice there are typically no significant gradients in the galactic rm field beyond @xmath249 .
this allows us to use the smooth value safely when analyzing a small area with extragalactic rm measurements .
this value is also consistent with all the abovementioned values of @xmath212 as obtained by more detailed analyses .
we would like to get an estimate for the internal @xmath0 contribution and for the foreground screen contribution ( if the latter exists ) . we do not attempt to correct each individual source for the @xmath0 contribution due to the internal and foreground screen terms .
we _ do not _ take the measured @xmath0 and subtract an estimate for @xmath250 , we rather attempt to estimate the distribution of @xmath250 , and treat it as another noise term .
the difference between the two separate terms in eq .
( [ eq : lin_sum ] ) ( @xmath251 and @xmath189 ) emerges from the fact that while the internal contribution is obtained from the same medium that emits the polarized light , the foreground screen only serves as a filter for an already existing signal . in previous investigations
, attempts were made to separate the two contributions ( burn 1966 ; laing 1984 ) , later on it became clear that the signature of the two is rather similar ( tribble 1991 ) .
for our purposes we are interested in all previous calculations left overs " , their signal is our noise and vice versa . for the noise estimate we claim it is legitimate to take the internal contribution and the foreground screen contribution together .
we argue that the two can be dealt with simultaneously . to this end
we shall use the other piece of information provided by the observations the polarization degree .
tribble ( 1991 ) shows the connection between the observed depolarization ( as a function of wavelength ) and the observed rm for an extended source . using this connection , and provided the polarization measurements are available ,
a direct estimate for the internal rm can be carried out .
recall we do not expect any depolarization due to the cosmological rm since the cosmological coherence we are after is orders of magnitude bigger ( @xmath252 a few @xmath253 ) than the source size .
it is interesting to note , that even if we had perfect information about the depolarization due to the source structure and foreground screen contribution and an exact relation between the depolarization degree and the @xmath0 , we could still not subtract this derived @xmath0 from the observed one .
this is due to the fact that the polarization degree does nt specify the @xmath0 direction .
tribble s statistical connection between the observed depolarization and the observed rm is valid only if the correlation scale of the rm structure function for the source or the foreground screen is much shorter than the telescope beam size .
this condition is probably not valid for individual damped lyman@xmath190 systems , and galaxies that serve as foreground screens , and for which long range correlation across the telescope beam may exist .
full justification for neglecting damped lyman@xmath190 systems and galaxies exists only if we can avoid lines - of - sight that cross such systems .
otherwise the order of magnitude of an intervening galaxy contribution is obtained as follows : for a galaxy seen face - on , where there is a danger of a well ordered magnetic field , the disc thickness , to which the magnetic field is presumably connected is of the order of kpc ( simard - normandin & kronberg 1980 ) .
observations show magnetic field magnitude of @xmath254 g , mainly in the disc plane . in order to equate this to a cosmological magnetic field of
the order of ng , and coherent scale of @xmath255 , the average baryon number density across the galaxy should be at least fifty times higher than the cosmological @xmath3 .
however , we do not expect any correlation between the magnetic field orientation of galaxies along the line - of - sight , or galaxies adjacent ( in angle ) to each other .
the worst contribution to the rm , could only come from a random walk of @xmath256 steps where @xmath256 is the number of foreground screens ( intervening systems ) along the line - of - sight . since one can model the probability for an intersecting system along the line - of - sight as a function of the source redshift ( welter , perry , & kronberg 1984 ; lanzetta , wolfe , & turnshek 1995 ) , another noise term can be added .
we did not attempt to model this noise term here , because we believe that the best strategy would be to avoid highly contributing ( damped lyman-@xmath190 and lyman limit ) intervening systems .
low column density lyman-@xmath190 system contribute very little to the rm signal ( see eq .
( 1.2 ) in welter , perry , & kronberg 1984 ) .
an exception may exist , if two sources ( or two separated parts of the same source ) cross the same cluster . in that case
the noise of the two sources ( @xmath250 ) may be correlated , but then we can go back to the technique that uses the depolarization , in an attempt to remove the noise correlation . the best strategy , as stated earlier , would be to avoid intervening systems altogether .
there are two methods for doing this .
the first method is simply to avoid intervening systems by looking for sources that exhibit no absorption lines in their spectra ( we assume spectrography is carried out anyway , to allow redshift determination ) .
the existence of a large fraction of quasar lines - of - sight which do not intersect a dense intervenor is corroborated by mller & jakobsen ( 1990 ) and lanzetta , wolfe , & turnshek ( 1995 ) analyses .
the other method is to put a depolarization limit on the observations ( the one imposed by tabara & inoue 1979 ) .
such a limit naturally reduces the noise term of @xmath257 and does not affect the correlation signal . as a matter of fact , a sequence of such depolarization limits , may be helpful in constituting the estimate for @xmath257 .
there are yet more methods for minimizing the number of sources that manifest internal rm as we shall point out when we consider real observations ( [ sec : discuss ] ) . in the current stage of research
we have very little knowledge about the power spectrum of the cosmological magnetic field .
we can hardly even put limits on its integral value , the rms value of the field .
there are , however predictions regarding the power spectrum shape and amplitude . in the lack of any observational preference towards any of these predictions ( apart from the limits on its rms value )
, we assume that all models are equally probable .
conventional estimates of the @xmath0 correlation , estimates of the sort applied to galaxies are not very useful .
it is difficult to get correct error estimate without simulations , that in turn must assume some magnetic field power spectrum , moreover in the traditional correlation calculation procedure , the errors of different bins in @xmath258 space are correlated . on top of that ,
the quantity we are after is the magnetic field power spectrum , and the inversion from the integrated @xmath0 back to the magnetic field is a non - trivial one .
all of the above lead us to the use of bayesian statistics as a tool for finding the best parameters , and their probability , given a model . in the bayesian formalism ,
the a - posteriori probability density of a model , @xmath259 , given the data , @xmath260 , is @xmath261 as stated earlier , @xmath262 , the model probability density is unknown and therefore assumed uniform for all models .
the data probability , in the denominator is the same for all models , and therefore serves as a normalization factor .
we are thus left with equivalence between the operation of maximizing the probability of the model , given the data [ @xmath263 and the operation of maximizing the probability of the data , given the model [ @xmath264 . this equivalence allows us to write down the likelihood function for @xmath173 sources with measured @xmath0 .
we begin by constructing the @xmath265 symmetric matrix @xmath266 of the expectation values for the @xmath0 correlation ( eq .
[ eq : ups_2 ] ) between these sources ( as determined by their position ) .
we further assume that the errors in each @xmath0 value are uncorrelated and gaussianly distributed .
the variance of the overall error is the sum of the quadratures of the various error sources @xmath267 and the likelihood function becomes @xmath268^{-1/2 } \exp\left ( -{1\over 2}\sum_{i , j}^{n_{src } } { \rm_i \tilde \upsilon_{ij}^{-1 } \rm_j}\right)\,,\ ] ] where @xmath269 .
+ the @xmath270 statistic is defined as @xmath271 .
the @xmath270 statistics as defined , is a @xmath270 distribution ( of @xmath173 degrees of freedom ) with respect to the _ data points _ , not the model parameters .
this @xmath270 is interpreted as a rough estimate for the confidence levels in the _
parameter _ space . before proceeding to the elaborated description of the various tests we performed with artificial data
, we would like to get a rough estimate for the expected rm correlation and the necessary number of sources for a high enough signal - to - noise ratio at various scales .
we temporarily abandon the bayesian approach , and focus on @xmath272 . recall that the noise in the cross correlation is not correlated , but may still dominate the signal if not appropriately averaged over many pairs . the signal - to - noise ratio for an individual measurement of a pair at identical redshift , a separation angle @xmath44 , and in the context of a specific model , is given by @xmath273 where in the last equality we assume no dependence between the various noise terms and the model ( cosmology + power spectrum ) .
the cosmological parameters we choose to use are as follows : we restrict ourselves to flat cosmology with @xmath274 , we choose @xmath275 ( but the scaling with @xmath4 is straightforward ) .
we take the ionization factor to be @xmath276 which according to the gunn - peterson effect ( gunn & peterson 1965 ) is a reasonable choice , and @xmath277 ( tytler , fan , & burles 1996 ) , @xmath278 @xmath279 . for @xmath280 pairs where each source is a member of one pair only , the signal - to - noise can further be increased by @xmath281 . in figure
[ fig : ston ] we plot the necessary number of pairs in order to achieve a signal to noise ratio of @xmath282 .
all sources are assumed to be at the same redshift ( @xmath283 , @xmath243 , or @xmath284 ) , we take the noise values @xmath285 ( this is the value for @xmath247 ) . for the noise term due to the internal rm and the ( possible ) foreground screen we use the tabara & inoue ( 1979 ) catalog as a guideline .
tabara & inoue listed @xmath286 radio sources in their catalog with @xmath0 @xmath287 .
that means a @xmath288 cover , and gives @xmath289 .
the most conservative assumption one can make is to attribute all of this @xmath0 to the internal and foreground screen contribution . + the procedure for deriving the internal and foreground screen contribution from the depolarization ( [ subsec : i+f ] ) bears noise itself and is correct only statistically .
we model its accuracy as a gaussian with one quarter the width of the observed rm distribution . doing so we end up with @xmath290 .
this is a conservative choice , and the true value for @xmath257 is probably much smaller .
note for instance the @xmath291 value as obtained by oren & wolfe ( 1995 ) .
the measurement errors are neglected ( [ subsec : raw ] ) .
two power spectra are considered and they are both normalized to give @xmath292 ng ( the @xmath293 case gives results between the @xmath294 and the @xmath295 power spectra ) . a typical number of pairs for achieving a signal to noise ratio of @xmath195 in the range @xmath296 is between @xmath297 and @xmath298 pairs , depending on the power spectrum .
0.5truecm = 2.7 in 2truecm = 2.7 in this estimate of @xmath9 is not accurate . to begin with , for @xmath173 sources
we expect less than @xmath299 statistically independent pairs .
we do not expect all sources to reside at the same redshift , neither do we expect the errors to be identical for all sources .
furthermore , the separation between the sources is not fixed , but rather is spread from the arc - minute scale to @xmath300 . due to the need for a realistic error estimate , and a demonstration of an actual implementation of the technique we propose , led us to apply it to mock catalogs , for which the underlying power spectrum is known .
as a reliable probe for our method we create artificial catalogs of radio sources for which rm values are measured . we first generate a gaussian random field of the magnetic vector potential , @xmath301 with the power spectrum @xmath302 .
the realization takes place on a @xmath303 grid points equally spaced in comoving coordinates .
the @xmath127 power spectrum is gaussianly smoothed on the one grid cell scale .
the amplitude of the field is arbitrary .
we then derive the magnetic field itself by the real space relation @xmath304 translated to @xmath160 space @xmath305 .
we use the @xmath160 space symmetries due to the real nature of the @xmath301 field ( no imaginary part ) , and go back to real space to obtain a divergence - free magnetic field with the desired correlation function , and periodic boundary conditions .
we check the resultant field by calculating the field divergence about each and every grid point in real space and get @xmath306 . at this point
we select the cosmology into which we embed the simulation .
our standard choice is identical to the one described in the signal - to - noise estimate ( [ subsec : ston ] ) . due to the limited dynamical range , we choose the grid scale to represent @xmath307 , and we are therefore confined to a largest simulation wavelength of @xmath308 .
we then select the source mask , the source distribution across the sky , and their redshifts .
the two - dimensional distribution is taken to be a poisson - like distribution , this choice allows us to have close pairs ( unlike grid selection ) , but to be as conservative as possible in terms of the angular correlation function of the sources . introducing any correlation
as observed for radio sources ( kooiman , burne & klypin 1995 ; sicotte 1995 ; cress 1995 ; loan , wall , & lahav 1997 ) can only increase the number of close pairs on small angles , and thus improve the @xmath0 correlation resultant signal ( see fig . [
fig : ston ] ) .
we do not take explicitly into account very small angle ( sub arc - minute ) separations that may exist in real data sets for resolved extended sources , whose different parts can be used as more than one background source . we employ two distribution schemes , an all - sky coverage , and a cover of @xmath309 deg.@xmath219 area centered about the north galactic pole .
the source redshift distribution is selected according to the @xmath310 of radio sources with flux @xmath311 mjy ( at 4.85 ghz ) as calculated by loan , wall , & lahav ( 1997 ) who used the mean of the theoretical luminosity function models of dunlop & peacock ( 1990 ) and kindly provided their fit .
loan ( 1997 ) @xmath310 is peaked around @xmath312 , and has a shape resembling a low redshift truncated gaussian of width ( @xmath313 ) @xmath314 . in principle , setting flux limit and sky coverage , sets the source number . in practice only a sub - set of all radio sources emit polarized light .
a random sample taken from the nvss survey ( condon 1994 ) shows registered polarization angle for @xmath315 of the sources . refraining from sources with intervening systems and large internal rm
may further decrease the number .
since the source number density in the radio catalogs of this flux limit is a few per square degree , when we prepare an all - sky " coverage catalog ( see below ) , we are always way below the available number of sources . for smaller angular coverage with higher source density
, one has to further decrease the flux limit .
for example in the @xmath316 mjy limit we expect @xmath317 sources per square degree ( nvss , condon 1994 ) , and in the @xmath243 mjy limit we expect source density of about @xmath5 per square degree ( the first " survey ; becker , white , & helfand 1995 ) .
the @xmath310 function of loan ( 1997 ) is therefore interpreted as a normalized selection function , with a cutoff at the highest redshift of the catalog .
we expect this selection function to provide an underestimate for the number of sources at high redshift , if a lower flux limit is set .
since sources at higher redshift typically bear higher signal ( without changing the noise ) , such an imposed selection is a conservative choice in terms of the expected signal - to - noise ratio .
we choose , which is peaked in a relatively low redshift , may be useful in mimicking another effect . as the source redshift increases , the probability for intervening systems along the line - of - sight increases as well .
if in a sample compilation we try to avoid intervening system , this attempt translates to a steeper fall - off of the selection function . ]
+ for each selected source , we integrate along the line - of - sight over the dot product between the magnetic field and the line - of - sight direction using the periodic boundary conditions .
the integration scheme assigns weights to each integration segment , @xmath160 , according to @xmath318 in order to account for the cosmological evolution .
we then add the following noise terms to the resultant @xmath0 1 .
internal variation + foreground screens .
+ we draw this noise term from a gaussian of @xmath319 variance ( [ subsec : ston ] ) .
this is the added rm to the cosmological one .
however , we also try to imitate the procedure of @xmath257 evaluation by the depolarization . to this end , for a specific value of @xmath320 ( drawn earlier ) , we mimic the recovery of the @xmath257 from the depolarization degree , by listing an @xmath257 value scattered about @xmath320 .
the scatter has a gaussian distribution of @xmath321 variance .
this is the error in the noise estimate .
observer " takes into account the scattered value of the @xmath257 and not the value that was actually added to the integrated cosmological rm .
2 . galactic mask .
+ we use the model - estimate of the galactic mask , smoothed to @xmath322 scale ( see [ subsec : gal_mask ] ) .
since the smoothed galactic rm differs from the true @xmath0 to the line - of - sight , we add a @xmath0 drawn from a gaussian distribution with the standard deviation of @xmath323 ( eq . [ eq : gal_res ] ) . at the end of this process
we are left with a mock @xmath0 catalog that consists of source coordinates and redshift , measured @xmath0 value , and the total error in the @xmath0 value ( @xmath257 ) .
we proceed by feeding the mock catalog into the maximum likelihood procedure . in the current application
, we consider only the right cosmology , the one in which the simulation was embedded . in principle
the sensitivity to the cosmology choice can be checked as well .
the two variables for the model are the amplitude and the power index .
it is really the combination of the two that the likelihood procedure constraints best .
+ the statistical significance of the @xmath270 distribution ( as defined following eq .
( [ eq : like ] ) ) is obtained in the standard way .
the number of degrees of freedom is the source number minus the number of fitting parameters , and the goodness of fit is calculated by taking into account the value of @xmath324 .
if the goodness of fit turns out to be very small ( @xmath325 ) , this is a clear indication for convergence to the wrong minimum , and vice versa .
for the right solution the minimum value for the @xmath270 per degree of freedom ( the source number ) is typically unity within @xmath326 accuracy .
we first explore partial sky coverage mock rm catalogs .
figures [ fig : like_part ] show contour maps of @xmath327 with steps of unity .
the thicker lines indicate the approximate @xmath328 , and @xmath329 levels of the probability function as ascribed by using two parameter @xmath270 statistics ( [ subsec : bayes ] ) .
the power spectrum normalization is our standard @xmath330 ng , depending on the power spectrum .
we use @xmath331 sources for the part sky and @xmath309 square degrees coverage . the different source number and @xmath10 values reflect the fact that we have fixed the angular sky coverage .
that in turn translates differently for different spectra .
a redshift cutoff of @xmath332 has been applied .
+ figure [ fig : like_part_high ] shows a particular example for the case @xmath295 , where we have increased @xmath333 to the @xmath334 ng level in order to verify the result dependence on @xmath10 . -1.5truecm
5truecm = 2.7 in -2truecm 0.5truecm = 2.7 in 2truecm = 2.7 in -0truecm we notice that for all power spectra , the bayesian approach recovers the correct value from the artificial catalog ( marked by a cross sign ) to within the @xmath227 level .
the method is more robust with respect to the amplitude determination .
the power index is not well bound unless it is very different from the true one .
the method ( with the limited number of sources ) tends to recover best the @xmath10 value on a certain scale , which is a combination of the power index and the amplitude .
typical values for the ratio between the true @xmath10 and the minimization result for @xmath10 are in the range @xmath335 .
partial sky coverage is preferable in terms of the number of small separation pairs , and the indifference to the galactic mask variation as function of direction .
a collection of several of these small patches is expected to provide an even better constrain on the magnetic field power spectrum , but the analysis becomes a bit more complicated , as one should also include the large scale correlation between the patches without increasing the noise from the auto correlation to an intolerable level . given the relatively big errors , we desire off - diagonal terms in the correlation matrix to be as big as possible ( small separation ) .
this is meant to avoid heavy weight for diagonal terms which are dominated by noise .
the price for choosing small patches with small sky coverage is limited ability to constrain the power index . under special conditions ,
this ability can be partially recovered by turning to an all sky coverage as we discuss in the next section . -1.5truecm
5truecm = 2.7 in -1.0truecm an all sky coverage , with the small correlation scales that are tested here , is impractical without lowering the error level significantly .
otherwise the bayesian analysis will all be dominated by the diagonal terms with a large noise added to them and one loses the advantage of the correlation scheme for noise reduction .
all sky coverage is however still feasible if we design the sample very carefully . in order to minimize the galactic noise
we should observe sources in proximity to the mask sources direction and avoid the necessity for smoothing the galactic mask .
instead , the rm values as deduced from the mask source population can simply be subtracted from the observed rm values . the small variation in the galactic magnetic field on small scales ( simonetti , cordes , & spangler 1984 ; simonetti & cordes 1986 ; minter & spangler 1996 ) allows this subtraction to take place .
if on top of that one can reduce the noise from internal rm to the level of @xmath336 ( one can always avoid foreground screens ) by multi wavelength observations , then it becomes sensible to exploit an all - sky coverage .
figure [ fig : xlike_all_n=0 ] shows an example of a sample taken at the mask source directions with @xmath337 and @xmath338 ng .
the rest of the parameters are identical to those of the part sky coverage ( @xmath295 ) of the last section .
we notice that the power index is somewhat better constrained in an all sky coverage , especially if the amplitude ( or @xmath10 ) are assumed or known from some different source .
the true values are recovered to within the @xmath339 level .
moreover such coverage allows us to calculate all prevailing magnetic fields on the sample scale , expressed as @xmath340 with @xmath341 the dirac delta function and @xmath159 the sample depth .
notice , however that using the formalism proposed in section [ sec : corr ] , we can not take into account power spectra that are not functions of @xmath342 alone so this test is a bit weaker than the dipole test . -1.5truecm
5truecm = 2.7 in -1truecm
we have presented the formalism for the rm correlation function and demonstrated that by using it we can put limits on the power spectrum of the cosmological magnetic field .
these limits are more stringent than limits obtained by any other method that currently exists
. limits of @xmath343 level for magnetic field of @xmath198 ng on inter - cluster scales ( @xmath344 ) can be devised with only @xmath345 sources .
the increment of the source number reflects linearly in the @xmath10 limits .
we performed the statistical analysis following a bayesian approach that seems to be the most adequate for this problem .
the correlation method we propose is also less susceptible to systematic effects such as observational bias ( or bias due to evolution ) toward smaller internal rm measures for higher redshift sources . both of these biases tend to lower the expected correlation signal as the source redshifts increase , whereas a cosmological magnetic field tends to raise the correlation signal between such sources . in order to estimate the statistical significance of limits as derived by the proposed method , we have simulated a few possible cosmological magnetic fields .
the simulations allow us to have a fair estimate of the errors involved in the power spectrum derivation .
the analysis depends , however on a few assumptions , and most crucially on the ability to eliminate or estimate the foreground screen and internal contribution to the observed rm .
this crucial point can be addressed in the process of the sample compilation .
reduction of the foreground screen and internal rm contribution is achievable by applying two cutoff criteria to the sample .
the first cutoff in minimal polarization degree is meant to avoid sources that exhibit large depolarization degree .
these sources are usually interpreted as having a big non - cosmological contribution to their measured rm .
the second cutoff is for sources that do not follow a @xmath346 dependence of the rm .
following laing s ( 1984 ) recipe , in order to minimize the number of sources with internal rm , the wavelength span should allow detection of deviation from the @xmath346 dependence of the rm for @xmath347 .
should this deviation occur , the source is to be eliminated from the catalog as a suspect for substantial internal rm contribution
. the spatial coverage of the sky should be twofold .
small patches of a few square degrees enable the small separation pair number needed for noise reduction .
small patches , however , can not reveal large correlation scales in an effective way .
it is thus desirable to compile a dilute all sky coverage sample , in addition to the high number density patches off the galactic plane .
the two sampling strategies cover a large range of potential correlation scales , without the necessity for an all - sky dense sample .
the all sky coverage should preferably look for sources in proximity to the mask sources directions in order to minimize the galactic noise term .
accurate internal rm for these sources should be evaluated through depolarization measurements in order for them to be useful for the power spectrum evaluation .
ultimately , the sample can always be mimicked in terms of the exact source locations in the framework of all and every detected model .
exact imitation of the source locations ensures the right correlation terms in the correlation matrix .
the limits we have derived here are a combination of limits on the free electron average density , @xmath348 , and the magnetic field .
if by any other fashion ( like hii absorption ) an estimate for @xmath349 can be achieved , then the limits on @xmath79 will become more robust .
the estimate is also a function of the assumed cosmology , and may be entangled with magnetic field evolution in the post recombination era .
we havent modeled such evolution in this paper .
this method of rm correlation can further be generalized to the _ smoothed _ rm correlation .
the data can be smoothed on a certain angular scale , and then either compared to the expectation value predicted by a model , or inverted numerically to give limits on @xmath79 . the bayesian approach , though ceases to be advantageous in the smoothed case , because even though the noise terms decrease , the smoothing introduces correlation among the errors , and the statistical interpretation becomes more hazardous .
a nice feature of the proposed analysis is that it is bound to yield results
. these can be either limits on the magnetic field magnitude , or actual detection of its value .
either way , these results may help in lifting the curtain over the mystery of the cluster and galactic magnetic fields origin .
i would like to thank arthur kosowsky for extensive discussions and helpful assistance , george blumenthal for valuable insight and critical reading of the manuscript , and james bullock for a very carefull examination of the manuscript . +
this work was supported in part by the us national science foundation ( phy-9507695 ) .
adams , j. , danielsson , u.h . , grasso , d. , & rubinstein , h. 1996 , sissa preprint , astro - ph/9607043 barrow , j.d . , ferreira , p.g . , & silk , j. 1997 , sissa preprint , astro - ph/9701063 baugh , c.m .
1995 , sissa preprint , astro - ph/9512011 baugh , c.m . and efstathiou , g. 1993 , , 265 , 145 becker , r.h . , white , r.l . , & helfand , d.j .
1995 , , 450 , 559 boyle , b.j .
1986 , phd thesis , university of durham brandenburg , a. , enquist , k. , & olesen , p. 1996
, sissa preprint , asrto - ph/9602031 broten , n.w . ,
macleod , j.m . , & valle , j , p .
1988 , , 141 , 303 burn , b.j .
1966 , , 133 , 67 camilo , f. , nice , d.j . , & taylor j.h . 1996 , , 461 , 812 cheng , b.l . ,
olinto , a.v . , schramm , d.n .
, & truran , j.w 1996 , , 54 , 4714 condon , j.j . ,
cotton , w.d . ,
greisen , e.w .
, yin , q.f . , perley , r.a . , & broderick , j.j .
1994 , in _ astronomical data analysis software and systems iii , a.s.p .
conference series _ , vol .
dennis , r. crabtree , r.j .
, barnes h. , & barnes , j. p. 155
cress , c.m .
, helfand , d.j . ,
becker , r.h . ,
gregg , m.d .
, white , r.l .
1996 , , 473 , 7 dolgov , a. & silk , j. 1993 , 47 , 3144 dunlop , j.s . & peacock , j.a .
1990 , 247 , 19 foster , r.s . , cadwell , b.j . , wolszczan , a. , & anderson , s.b .
1995 , , 454 , 826 grasso , d. & rubinstein h.r .
1995 , nucl .
b , 43 , 303 grasso , d. & rubinstein h.r .
1996 , phys .
b , 388 , 253 gunn , j.e . & peterson , b.,a .
1965 , , 142 , 1633 harrison , e.r . 1973 , ,
30 , 18 kato , t. , tabara , h. , inoue , m. , & aizu , k. 1987 , , 329 , 223 kolb , e.w . & turner , m.s .
1990 , _ the early universe _ , ( usa : addison - wesley ) p. 84
kooiman , b.l . ,
burns , j.o . , &
klypin , a.a .
1995 , , 448 , 500 kronberg , p.p .
1976 , _ int .
union symp .
_ 74 , 367 kronberg , p.p .
1994 , rep .
phys . , 57 , 325 ( k94 ) kronberg , p.p . & simard - normandin , m. 1976 , , 263 , 653 laing , r.a .
1984 , in _ physics of energy transport in extragalactic radio sources _ , edited by bridle , a.h . & eilek , j.a .
( green bank w.va .
usa : nrao ) p. 90
lang , k.r .
1978 , _ astrophysical formulae _ ,
( new - york : springer - verlag ) p. 57
lanzetta , k.m . ,
wolfe , a.m. , turnshek , d.a .
1995 , , 440 , 425 lee , s. , olinto , a.v .
, & sigl , g. 1995 , , 455 , l21 loan , a.j . , wall , j.v . , & lahav , o. 1997 , , 286 , 994 loeb , a. & kosowsky , a. 1996 , , 469 , 1 masden , m.s .
1989 , , 237 , 109 minter , a.h . &
spangler , s.r .
1996 , , 458 , 194 mller , p. & jakobsen , p. 1990
, , 299 monin , a.s . &
yaglom , a.m. 1975 , _ statistical fluid mechanics _ ( cambridge ; mit press ) , pp .
2952 nissen , d. & thielheim , k.o .
1975 , ap&ss , 33 , 441 oren , a.l . & wolfe , a.m. 1995 , 445 , 624 osmer , p.s . 1981 , , 247 , 762 ostriker , j.p . &
thompson , c. 1987 , , 323 , l97 parker , e.n .
1979 , _ cosmical magnetic fields _
( oxford ; clarendon press ) p. 616
pudritz , r.e . & silk , j. 1989 , , 342 , 650 quashnock , j.m . , loeb , a. , & spergel , d.n .
1989 , , 342 , 650 ratra , b. 1992a , , 45 , 1913 ratra , b. 1992b , , 391 , l1 rees , m.,j .
1987 , , 28 , 197 sicotte , h. , 1995 , ph.d . thesis , princeton university simard - normandin , m. & kronberg , p.p .
1980 , , 242 , 74 simard - normandin , m. & kronberg , p.p .
, & button , s. 1980 , , 45 , 97 ( skb ) simonetti , j.h . , cordes , j.m . , & spangler , s.r .
1984 , 284 , 126 simonetti , j.h . &
cordes , j.m .
1986 , 310 , 160 tabara , h. & inoue , m. 1979 , supp .
, 39 , 379 tadros , h. & efstathiou , g. 1995 , , 276 , 45 tajima , t. , cable , s. , shibata , k. , & kulsrud , r.m .
1992 , , 390 , 309 taylor , j.h .
& cordes , j.m . 1993 , , 411 , 674 taylor , j.h . ,
manchester , r.n .
, & lyne , a.g .
1993 , , 88 , 529 thompson , c. 1990 , _ int .
union symp .
_ , 140 , 127 tribble , p.c .
1991 , , 250 , 726 turner , m.s . & widrow , l.m .
1988 , , 37 , 10 tytler , d. , fan , x. , & burles , s. 1996 , , 381 , 207 vachaspati , t. 1991 , phys .
lett . , 265 , 258 valle , j.,p .
1990 , , 360 , 1 waxman , e. & miralda - escud , j. 1996 , , 472 , l89 weinberg , s. 1972 , _ gravitation and cosmology _ ( wiley & sons , ny ) p. 413
welter , g.l . ,
perry , j.j . , & kronberg , p.p .
1984 , , 279 , 19 zeldovich , y.b . &
novikov , i.d . , 1975 , _ relativistic astrophysics _ , ( chicago ; university of chicago press ) zweibel , e.g. 1988 , , 329 , l1 | this paper introduces the formalism which connects between rotation measure ( @xmath0 ) measurements for extragalactic sources and the cosmological magnetic field power spectrum .
it is shown that the amplitude and shape of the cosmological magnetic field power spectrum can be constrained by using a few hundred radio sources , for which faraday @xmath0s are available .
this constraint is of the form @xmath1 h $ ] nano - gauss ( ng ) on @xmath2 scales , with @xmath3 the average baryon density and @xmath4 the hubble parameter in @xmath5 km s@xmath6 mpc@xmath6 units .
the constraint is superior to and supersedes any other constraint which come from either cmb fluctuations , baryonic nucleosynthesis , or the first two multipoles of the magnetic field expansion .
the most adequate method for the constraint calculation uses the bayesian approach to the maximum likelihood function .
i demonstrate the ability to detect such magnetic fields by constructing simulations of the field and mimicking observations .
this procedure also provides error estimates for the derived quantities . the two main noise contributions due to the galactic rm and the internal rm are treated in a statistical way following an evaluation of their distribution .
for a range of magnetic field power spectra with power indices @xmath7 in a flat cosmology ( @xmath8=1 ) we estimate the signal - to - noise ratio , @xmath9 , for limits on the magnetic field @xmath10 on @xmath11 scale .
employing one patch of a few square degrees on the sky with source number density @xmath12 , an approximate estimate yields @xmath13 .
an all sky coverage , with much sparser , but carefully tailored sample of @xmath14 sources , yields @xmath15 with the same scaling .
an ideal combination of small densely sampled patches and sparse all - sky coverage yields @xmath16 with better constraints for the power index .
all of these estimates are corroborated by the simulations .
epsf.sty = cmr10 # 1#2.5pt 2 degm^-2 2 radm^-2 |
it is well known that several properties of galaxies depend on environment .
pioneering work by @xcite shows that galaxy morphology depends on local galaxy density : late type galaxies prefer low density environments , while dense environments tend to be populated by early type galaxies .
there is also a strong correlation between galaxy star formation rates ( sfrs ) and environment : galaxies in high density environments present a decrease in sfrs compared to field galaxies @xcite .
the effect of local environment on galaxy colours has been studied by @xcite .
this authors find that at @xmath4 faint galaxies show a clear increase in the fraction of red galaxy as the local density increases . on the other hand , bright galaxies , present a constant red galaxy fraction .
there are several known processes that influence galaxy properties .
the hot intra - cluster gas is the main responsible for stripping the gas of galaxies in the core of clusters .
mechanisms such as ram - pressure stripping @xcite and the strangulation or starvation scenarios @xcite lead to a decrease in the star formation rate of cluster galaxies .
tidal interactions of galaxy pairs and mergers affect the morphology of galaxies and can convert spiral galaxies into elliptical and s0s @xcite .
these processes can also trigger star formation , depending on the gas reservoir of the galaxies @xcite . in the extremely dense environments of compact groups ,
galaxies are separated only by a few galaxy radii from each other and with a low relative velocities , an ideal scenario for interactions and mergers @xcite .
several studies have analysed the properties of galaxies in different systems , such as pairs , compact groups , groups with four or more members , and clusters .
nevertheless , there are few works addressing the properties of galaxies in triple systems .
@xcite performed _
surface photometry of a sample of 54 galaxies selected from the catalogue of isolated triplets of galaxies in the nothern hemisphere @xcite and investigate the properties of 34 galaxies in 13 triplets .
these authors found that these systems are spiral dominated and a fraction of 56% of the triplets present morphological signatures associated with interactions .
they also found a fraction of 35% of bars , that can rise up to 66% in the late - type spirals , and a fraction of 20% of rings , hosted preferentially in late - type components .
based on these results the authors suggest that triple systems are essentially different from hickson compact groups and more representative of the field .
however , the results of this work are based mainly in observations of a low number of systems .
for this reason , these authors also highlight the relevance of building a complete sample of local isolated triplet of galaxies , that allows a significant statistical analysis .
@xcite ( hereafter paper i ) constructed a sample of isolated triplets of galaxies brighter than @xmath5 and complete up to @xmath6 from the data release 7 of sloan digital sky survey @xcite .
the aim of this paper is to analyse the properties of a sample of spectroscopic galaxies derived from this triplet catalogue and compare them with the properties of galaxies in different systems such as pairs , compact groups and clusters .
the proximity and low relative velocities of galaxy members in triplets present an ideal scenario for galaxy interactions and mergers , placing these systems in a state of ongoing collapse .
for these reason in this paper we will also investigate the properties of the system as a whole .
this paper is organized as follows . in section 2
we describe the triplet and galaxy system samples used in this work . in this section
we also build control samples in order to avoid biases in the comparison of properties of galaxies in different systems . in section 3
we analyse the specific star formation rate ( @xmath7 ) , the strength of the @xmath8 break , @xmath2 , which is an spectral indicator of the stellar population mean age , and ( @xmath3 ) colour index of galaxies in the different systems under analysis .
we study all these properties as a function of the stellar mass content in order to avoid biases due to differences in this parameter . in section 4
, we analyse de triplet as a system .
we use bootstrap techniques in order to assess the probability that the configurations of galaxies in triple systems are correlated or , on the contrary , can be explained by a random sampling .
we also investigate how the total stellar mass of the system affect the properties of galaxies in triplets and analyse global properties as total star formation rate and global ( @xmath3 ) colour index of the system as a whole .
we also analyse the compactness of the triplets by defining a parameter s that is a measure of the percentage of the system total area that is filled by the light of member galaxies . in this section
we also compare the properties of triple systems with the global properties of compact groups .
finally in section 6 we discuss our main results . throughout this paper
we adopt a cosmological model characterised by the parameters @xmath9 , @xmath10 and @xmath11 .
in this work we analyse several samples of galaxy systems derived from the data release 7 of sloan digital sky survey ( sdss - dr7 , * ? ? ?
sdss @xcite has mapped more than one - quarter of the entire sky , performing photometry and spectroscopy for galaxies , quasars and stars .
sdss - dr7 is the seventh major data release , corresponding to the completion of the survey sdss - ii .
it comprises @xmath12 sq . deg . of imaging data , with an increment of @xmath13 sq .
, over the previous data release , mostly in regions of low galactic latitude .
sdss - dr7 provides imaging data for 357 million distinct objects in five bands , _ ugriz _ , as well as spectroscopy over @xmath14 steradians in the north galactic cap and @xmath15 square degrees in the south galactic cap .
the average wavelengths corresponding to the five broad bands are @xmath16 , @xmath17 , @xmath18 , @xmath19 , and @xmath20 @xcite .
for details regarding the sdss camera see @xcite ; for astrometric calibrations see @xcite .
the survey has spectroscopy over 9380 sq .
deg . ; the spectroscopy is now complete over a large contiguous area of the northern galactic cap , closing the gap that was present in previous data releases . in paper
i we analysed spectroscopic and photometric data extracted from sdss - dr7 in order to build a catalogue of isolated triplets of galaxies .
the spectroscopic data were derived from the main galaxy sample ( mgs ; @xcite ) .
k - corrections for this sample were calculated bandshifted to @xmath21 , using the software ` k - correct_v4.2 ` of @xcite .
the photometric data were derived from the photometric catalogue constructed by @xcite , which contains photometric redshift and k - correction for the photometric data of the sdss - dr7 . for both data sets ,
k - corrected absolute magnitudes were calculated from petrosian apparent magnitudes converted to the ab system .
the final catalogue comprises 1092 isolated triplets of galaxies with absolute r - band magnitude brighter than @xmath5 in the redshift range @xmath22 .
a full description of the algorithm developed to build this catalogue can be found in paper i. one of the aims of this paper is to analyse different properties of galaxies in triplets by including information extracted from their spectra .
for this reason we have considered only triple systems composed by three spectroscopic galaxies that are close in projected separation @xmath23 and with a radial velocity difference @xmath24 . due
to completeness in the spectroscopy , we have restricted our analysis to the redshift range @xmath25 . in order to select isolated systems , we implement two of the isolation criteria described in paper 1 : 1 .
@xmath26 2 .
@xmath27 here , @xmath28 and @xmath29 are the number of galaxies brighter than @xmath5 within @xmath30 and @xmath31 , respectively , from the triplet centre considering the same restrictions on @xmath32 used to identify triplet members . in the same way that paper i
, we calculated the distance of each triplet to the closest neighbour group / cluster , @xmath33 , using two catalogues : @xcite , updated to the sdss - dr7 , for @xmath34 , and for the redshift range @xmath35 , the gmbcg catalogue of clusters @xcite that includes photometric redshift information . for each triplet system candidate we computed the projected distance to the closest neighbour cluster considering a radial velocity restriction @xmath36 when spectroscopic information is available , and @xmath37 when
photometric redshifts are involved .
we acknowledge that the systems obtained by these conditions are consistent with a minimum distance to a rich cluster of @xmath38 , similar to the restriction @xmath39 given in paper i. under these considerations , the final sample of galaxy triplets contain 71 isolated systems with 213 spectroscopic galaxies brighter than @xmath5 in the redshift range @xmath40 .
the aim of this work is to analyse the spectroscopic properties of galaxies in triplets and compare them with the properties of galaxies belonging to different galaxy systems .
here we consider samples of galaxies in compact groups , clusters and galaxy pairs .
@xcite identified compact groups of galaxies , using the photometric data of the sdss - dr6 , through the implementation of the ` hickson criteria ' : 1 .
@xmath41 ; 2 .
@xmath42 ; 3 .
@xmath43 magarcsec@xmath44 , here @xmath45 is the number of galaxies in the magnitude interval @xmath46 $ ] , where @xmath47 is the @xmath48-band magnitude of the brightest galaxy in the group . @xmath49
is the effective surface brightness of the system calculated by distributing the flux of the member galaxies over @xmath50 , where @xmath50 is the angular diameter of the smallest circle encompassing the geometric centres of the galaxies in the group .
finally , @xmath51 is defined as the angular diameter of the largest concentric circle that contains no other ( external ) galaxies within this magnitude range or brighter . @xcite build two catalogues : catalogue a comprising galaxies in compact groups with petrosian @xmath48-band magnitude in the range @xmath52 and catalogue b including galaxies in the broader magnitude range @xmath53 .
due to independent visual inspection by the authors of all galaxy members in catalogue a , the contamination due to gross photometric errors is negligible .
this sample includes compact groups that have @xmath54 where @xmath55 is a measure of the maximum line - of - sight velocity difference between group members for groups with more than two members with spectroscopic redshift information .
catalogue b includes many more groups than catalogue a but has the disadvantage of a larger contamination due to poor photometric classification .
we will use catalogue a as a comparison sample for this work .
nevertheless , this catalogue was constructed from the sixth release of the sdss , therefore we have added to this catalogue spectroscopic redshift information from sdss - dr7 , and in what follows we will use systems in catalogue a that have all their galaxies with spectroscopic measurements .
we have cross - correlated our triplet sample with this compact group catalogue in order to exclude common systems .
we found 5 coincidences of our triplets with the compact groups sample . from these common systems , four consist on three triplet galaxies plus one galaxy with @xmath56 .
the remained common - system is comprised by the three triplet galaxies and one galaxy fainter than @xmath5 .
we have therefore removed these 5 compact groups from the original sample .
this sample was derived from the catalogue of galaxy groups of @xcite updated to the sdss - dr7 .
these authors implement a friends - of - friends algorithm with varying linking lengths @xmath57 and @xmath58 , in the direction perpendicular and parallel to the line - of - sight , respectively , where @xmath59 and @xmath60 .
the spatial scaling @xmath61 takes into account the variation in the space density of galaxies in a flux - limited sample . from this catalogue , we selected a sample of clusters of galaxies considering systems with virial masses @xmath62 and more than 10 galaxy members .
the galaxies belonging to these systems were ranked in @xmath48-band luminosity .
then we compiled a sample selecting the ten first ranked cluster galaxies ( 10frcgs ) of each system , excluding the most luminous galaxy , since these galaxies usually have very distinctive properties .
the sample of galaxy pairs analysed in this work was obtained by @xcite .
these authors construct a galaxy pair catalogue selecting galaxies in the sdss - dr7 , with relative projected separations @xmath63 and relative radial velocities difference @xmath64 .
previous studies of the team found that these limits are adequate to define galaxy pairs with enhanced star formation activity @xcite . in this work
we have excluded from these sample , galaxy pairs that resides in groups , according to @xcite .
these authors analysed in detail the properties of galaxy interactions in high - density environments , identifying pairs that reside in groups by cross - correlating the total galaxy pair catalogue of @xcite with the group catalogue constructed by @xcite , updated to the sdss - dr7 . in order to compare the properties of galaxies in different systems ,
we cross - correlate all the galaxy samples described above , with the derived galaxy properties from the mpa - jhu emission line analysis for the sdss - dr7 . from this catalogue
we extract several galaxy properties : as a spectral indicator of the stellar population mean age , we will use the strength of the @xmath8 break , @xmath2 , defined as the ratio of the average flux densities in the narrow continuum bands 3850 - 3950 and 4000 - 4100 @xcite .
we also use the star formation rate ( sfr ) and specific star formation rates ( @xmath7 ) according to @xcite and total stellar masses ( @xmath65 ) calculated from photometry @xcite . when comparing properties of galaxies that belong to different samples it is important to consider differences in the galaxy redshift distribution .
for this reason , we construct control samples using a monte - carlo algorithm to randomly select galaxies that matched the galaxy triplet redshift distribution , for the different systems samples analysed in this work .
the distribution of @xmath66 of triplet galaxies is truncated by definition at @xmath5 @xcite . in order to avoid biases due to less luminous galaxies , we restrict our analysis to galaxies brighter than @xmath5 .
we performed a kolmogorov - smirnov ( ks ) test between the resdhift distribution of the control samples and the redshift distribution of triplet galaxies . from this test
we obtain a ` p ` value that represents the probability that a value of the ks statistic will be equal or more extreme than the observed value , if the null hypothesis holds . in all cases we obtained
` p`@xmath670.05 for the null hypothesis that the samples were drawn from the same distribution .
table [ t1 ] summarizes the name and number of objects in the triplet galaxy sample and in the control samples analysed in this work .
.name and number of galaxies in the triplet and control samples . [ cols="<,^",options="header " , ] the results obtained from this analysis indicate that both blue triplets and red triplets have such a combination of galaxies that can not be explained by a random reassignment of the galaxies . in particular , we conclude that systems comprising blue , star forming , young stellar population galaxies , have a high probability of being real systems and not a mere configuration of interloping galaxies . in support of this result
, @xcite found that galaxy triplets are spiral rich systems populated mostly by late - type spirals with an excess of @xmath68 mag in the global blue luminosity in comparison to field galaxies .
in the previous section we found that the galaxy configurations of triplets can not be explained by a random sampling of the triplet galaxies .
in particular there are six blue triplets that are comprised by blue , young stellar population and star forming galaxies . in this section
we investigate the dependence of triplet members properties on the total stellar mass of the system considering separately blue and red triplets and by comparison to the results from compact groups .
this joint analysis of our results and those for compact groups is based on their similar stellar mass range , a fact that does not hold for pairs nor clusters .
for the samples of triplets and compact groups described in section 2.1 and 2.2.1 , we calculate the redshift of each system as the mean redshift of their member galaxies .
we restrict our analysis to the redshift range of triplets and perform a ks test to the redshift distributions of triplets and compact groups .
the value obtained for ` p ` is greater than 0.05 so we can not reject the null hypothesis that the redshift of triplets and compact groups are drawn from the same distribution .
the samples of triplets and compact groups used in this section comprises 71 and 163 systems , respectively and we show in figure [ f5 ] the distributions of redshift and log total @xmath69 for these samples .
it can be appreciated that compact groups have a higher stellar mass content than triplets .
this is somewhat expected since triplets consist of three bright galaxies while compact groups are composed by more than four members . in our study , we have considered the total stellar mass range spanned by the triplet sample ( 10.6@xmath70log total @xmath7111.8 ) and analyse the properties of galaxies as a function the total stellar mass content of the system .
the contours shown in figure [ f6 ] represent the properties of compact groups members and the points correspond to galaxies in triplets .
we have considered separately blue triplets ( light - blue filled triangles ) , and red triplets ( red filled squares ) , as well as triplets that do not fulfil these categories ( blue dots ) . the top , middle and bottom panel of this figure correspond to the @xmath7 , the ( @xmath3 ) colour index and the concentration index @xmath72 , where @xmath73 and @xmath73 are the radii containing 90% and 50% of the petrosian galaxy light ] , respectively , as a function of the total stellar mass content .
the @xmath72 parameter is a suitable indicator of galaxy morphological type bimodality : early - type galaxies have @xmath74 while later type galaxies have typically @xmath75 @xcite [ c][c][1.3]log total @xmath65/@xmath76 [ c][c][1.3]c [ c][c][1.3](@xmath3 ) [ b][c][1.3]log sfr/@xmath65 [ 1/yr ] ( 190,610 ) ( 12,400 ) ( 12,200 ) ( 12,0 ) it can be appreciated , as expected , that the compact groups tend to have a tail towards larger total stellar masses .
nevertheless , triplets and compact groups members present a similar behaviour . from this figure
, we can also observe that blue triplets correspond to low total stellar mass systems .
galaxies in these triplets also show a higher star formation activity and lower values of the concentration index , @xmath72 , corresponding to disk objects .
on the other hand , galaxies in red triplets prefer higher total stellar mass systems and tend to be ellipticals with a low star formation efficiency .
the dynamic of galaxy systems suggests that interactions are common in compact systems such as triplets and compact groups @xcite .
@xcite performed dissipationless n - body simulations finding that compact configurations are continually replaced by new systems because within a single rich collapsing group , compact groups of galaxies form continuously . from the observational point of view ,
@xcite combined near - infrared imaging and optical spectroscopy with archival hubble space telescope imaging and spitzer imaging and spectroscopy , revealing that the luminous ir galaxy ( lirg ) iras 19115 - 2124 is actually a triple system where the lirg phenomenon is dominated by the smallest of the components .
since galaxies in triplets and compact groups may finally end in a single system , in this section we analyse different global properties of triplets of galaxies by considering the system as a whole .
we will also perform the same analysis on compact groups in order to compare the results obtained for these systems . for this purpose
we used the samples of triplets and compact groups defined in the previous section . with the aim to complement the previous analysis
, we also explore the dependence of the total star formation rate and total colour as a function of the total stellar mass of the systems.to measure the level of compactness of the systems , we have defined a geometric criteria in order to guarantee that the systems under analysis are not limited by the selection criterion commonly used in determining the system compactness .
radii of compact groups are defined by the hickson criteria as the smallest circle that contains the geometric centres of compact group members . in order to define an homogeneous compactness criterion for triplets and compact groups we calculate the triplets minimum enclosing circle using the code of @xcite .
then we define the compactness @xmath77 , as : @xmath78 where @xmath73 is the radius enclosing @xmath79 of the petrosian flux of the galaxy in the r - band , @xmath80 is the minimum enclosing circle that contains the geometric centres of the galaxies in the system and n is the total number of members of the system . by definition , this quantity is a measure of the percentage of the system total area that is filled by the light of member galaxies . left panels in figure [ f7 ] show the properties of the systems as a whole , as a function of the total stellar mass content . here
again , the contours represent the compact groups , and the points correspond to the triplets .
right panels in this figure show the distribution of the total sfr , global @xmath81 colour , and compactness @xmath77 for both samples .
as it was observed for individual galaxies , it can be seen that the blue triplets have a high total star formation activity and tend to be less massive systems .
red triplets are more massive objects and show a lower total star formation rate . from this figure
it can also be appreciated that there is a correlation between the total star formation rate and total stellar mass for the triplets , while compact groups present an approximately constant trend .
this is reflected in the total sfr distribution ( right upper panel ) : triplets are clearly bimodal , while the compact groups distribution is more consistent with a single population . regarding total colour
, there is a positive correlation with total stellar mass in both samples . as expected by its definition ,
blue triplets are located in the blue , low mass tail of the distribution . in the left bottom panel of figure [ f7 ]
, we show the s parameter as a function of the total stellar mass .
as can be appreciated , the compactness s of triplets increases toward high total stellar masses , and blue triplets tend to have less compact configurations .
on the other hand , compact groups show a nearly constant trend .
besides , as seen in the right bottom panel , it can be appreciated that the light in triplet systems is even more concentrated than in compact groups .
[ c][c][1.3]log total @xmath65/@xmath76 [ c][c][1.3]s [ c][c][1.2]log total sfr [ @xmath76/yr ] [ c][c][1.2 ] total ( @xmath3 ) ( 500,610 ) ( 15,390 ) ( 15,195 ) ( 15,0 ) ( 270,405 ) ( 270,205 ) ( 270,10 )
we have studied 71 galaxy triplets in the redshift range @xmath0 with galaxy members brighter than @xmath82 .
our study also includes a comparative analysis of galaxies in compact groups , the ten brightest members of rich clusters , and pair galaxies .
we also analyse the stellar mass content , the star formation rates , the spectral @xmath2 index and the ( @xmath3 ) colour index of member galaxies .
galaxies in triplets show star formation rates , colours and stellar populations comparable to those of galaxies in compact groups and clusters .
this contrasts with pair galaxy members , which have systematically higher star formation activity indicators .
thus , we conclude that triplets can not be considered as an extension of galaxy pairs with a third extra member . with the aim of determining whether the configurations of galaxies in the sample of triple systems are strongly correlated or , on the contrary ,
can be explained by a random sampling , we generate 1000 random triplet catalogues by a random reassignment of the galaxies from the real sample . from this analysis
we conclude that systems comprising three blue , star - forming , young stellar population galaxies ( blue triplets ) are most probably real systems and not a chance configuration of interloping galaxies .
the same holds for triplets composed by three red , non star - forming galaxies showing the correlation of galaxy properties in these systems . since triplets and compact group members may finally merge into a single system we have computed global parameters to provide a fair comparison of these systems as a whole
we have analysed different properties , such as total star formation rates and total global colours , finding that , in general , triplets and compact groups exhibit a similar behaviour .
nevertheless we find that blue triplets , located in the less massive tail of the total stellar mass distribution , show an efficient total star formation activity with respect to compact groups which present low efficiency in forming new star generations , in the same total stellar mass range .
in contrast , both , high stellar mass red triplets and compact groups in the same stellar mass region , show low total star formation activity . in order to provide a compactness parameter of the triplets that can be suitably compared to the other systems ,
we have defined a parameter s as the sum of the areas comprising 90@xmath83 of the light of galaxies divided into the system minimal enclosing circle area .
this quantity is a measure of the percentage of the system total area that is filled by the light of member galaxies .
the distribution of this compactness parameter shows , surprisingly , that light is even more concentrated in triplets than in compact groups .
the results obtained in this work suggest that triplets composed by three luminous galaxies , have member with properties more similar to compact group members than to galaxies in pairs , and that the behaviour of triplets as a whole is similar to compact groups .
based on the results presented in this work , we argue that triplets are a natural extension of compact groups to systems with lower number of galaxies .
we thank the referee , nelson padilla , for providing us with helpful comments that improved this work .
we thank the authors zapata , perez , padilla & tissera for granting us access to their group catalogue .
this work was supported in part by the consejo nacional de investigaciones cientficas y tcnicas de la repblica argentina ( conicet ) , secretara de ciencia y tecnologa de la universidad de crdoba .
laerte sodr jr and ana laura omill were supported by the brazilian agencies fapesp and cnpq .
funding for the sdss and sdss - ii has been provided by the alfred p. sloan foundation , the participating institutions , the national science foundation , the u.s .
department of energy , the national aeronautics and space administration , the japanese monbukagakusho , the max planck society , and the higher education funding council for england .
the sdss web site is http://www.sdss.org/. the sdss is managed by the astrophysical research consortium for the participating institutions .
the participating institutions are the american museum of natural history , astrophysical institute potsdam , university of basel , university of cambridge , case western reserve university , university of chicago , drexel university , fermilab , the institute for advanced study , the japan participation group , johns hopkins university , the joint institute for nuclear astrophysics , the kavli institute for particle astrophysics and cosmology , the korean scientist group , the chinese academy of sciences ( lamost ) , los alamos national laboratory , the max - planck - institute for astronomy ( mpia ) , the max - planck - institute for astrophysics ( mpa ) , new mexico state university , ohio state university , university of pittsburgh , university of portsmouth , princeton university , the united states naval observatory , and the university of washington . | we analyse a sample of 71 triplets of luminous galaxies derived from the work of @xcite .
we compare the properties of triplets and their members with those of control samples of compact groups , the ten brightest members of rich clusters , and galaxies in pairs .
the triplets are restricted to have members with spectroscopic redshifts in the range @xmath0 and absolute r - band luminosities brighter than @xmath1 . for these member galaxies ,
we analyse the stellar mass content , the star formation rates , the @xmath2 parameter and ( @xmath3 ) colour index .
since galaxies in triplets may finally merge in a single system , we analyse different global properties of these systems .
we calculate the probability that the properties of galaxies in triplets are strongly correlated .
we also study total star formation activity and global colours , and define the triplet compactness as a measure of the percentage of the system total area that is filled by the light of member galaxies .
we concentrate in the comparison of our results with those of compact groups to assess how the triplets are a natural extension of these compact systems .
our analysis suggest that triplet galaxy members behave similarly to compact group members and galaxies in rich clusters .
we also find that systems comprising three blue , star - forming , young stellar population galaxies ( blue triplets ) are most probably real systems and not a chance configuration of interloping galaxies .
the same holds for triplets composed by three red , non star - forming galaxies , showing the correlation of galaxy properties in these systems . from the analysis of the triplet as a whole ,
we conclude that , at a given total stellar mass content , triplets show a total star formation activity and global colours similar to compact groups . however , blue triplets show a high total star formation activity with a lower stellar mass content . from an analysis of the compactness parameter of the systems we find that light is even more concentrated in triplets than in compact groups . we propose that triplets composed by three luminous galaxies , should not be considered as an analogous of galaxy pairs with a third extra member , but
rather they are a natural extension of compact groups .
[ firstpage ] galaxies : general - galaxies : interactions |
quantum parameter estimation , the emerging field of quantum technology , aims to use entanglement and other quantum resources to yield higher statistical precision of a parameter @xmath4 than purely classical approaches @xcite .
the precision of estimation of @xmath4 will depend on the available resources used in the measurement .
it has been shown that standard quantum limit ( sql ) or called shot noise limit scaling as @xmath5 with @xmath6 the number of particles can be surpassed by using coherent light with squeezed vacuum @xcite .
it is also commonly considered that using non - gaussian states like noon states @xcite and quantum entanglement allows one to achieve a sub - shot noise accuracy .
heisenberg limit scaling as @xmath7 is the ultimate limit set by quantum mechanics .
recently , some works @xcite have shown that , without prior information , sub - heisenberg estimation strategies are ineffective .
there are also some papers showing that the heisenberg limit can be saturated without the use of any exotic quantum entangled states @xcite .
interferometric strategies with nonlinear phase encoding are investigated in refs .
practical quantum metrology considering the impact of noise has been considered and studied in refs .
the technique of quantum parameter estimation figures in several metrology platforms , including optical interferometry @xcite , atomic systems @xcite , and bose - einstein condensates @xcite .
in addition , it is at the heart of many modern technologies and researches , such as quantum clock synchronization @xcite , quantum imaging @xcite , and gravitational wave observation @xcite .
general parameter estimation procedure can be divided into three distinct sections : probe preparations , interaction between the probe and the system , and the probe readouts @xcite .
these three sections will be repeated many times before the final construction of the estimation of @xmath4 .
most of the quantum parameter estimation strategies require preparation of @xmath0 copies of entangled states ( @xmath0 is large enough ) .
however , these states are extremely difficult to generate and fragile to the impact of decoherence .
therefore , the method of quantum nondemolition ( qnd ) measurements @xcite initially with entanglement - free states may be a suitable and practical way to overcome these challenges .
the qnd measurements dating back to as early as the 1920s realize ideal projective measurements that leave the system in an eigenstate of the measured observable @xcite . with these ideal projective measurements performed on an initial coherent state , fock states and `` schrdinger cat '' states
can be prepared and reconstructed @xcite .
moreover , with appropriate feedback loops , it is possible to prepare on demand photon states and subsequently reverses the effects of decoherence @xcite . with these merits , we can foresee the widespread applications of this techniques in quantum information and quantum metrology . in this paper
, we present a practical proposal for realizing quantum parameter estimation in `` photon box '' @xcite via qnd measurements .
we show that , with single - mode fock state of @xmath1 photons in the `` photon box '' , this proposal can estimate the parameter @xmath4 within a scaling of @xmath2 . unlike other quantum metrology strategies , our proposal has this advantage - the state of photons can be used circularly .
thus , our scheme performs better than the strategy with noon state when the total resource is taken into consideration .
we also investigate our qnd metrology scheme with cavity damping .
it is shown that our scheme can sustain the quantum advantage for a longer time than the damping time of fock state and is more robust than the interferometric strategy with exotic entangled states .
an improved cascaded estimation scheme is also proposed by adding a real - time feedback interferometric layout @xcite , with which the common @xmath3 periodic error can be handled .
the experimental feasibility of our proposals can be justified with current laboratory parameters @xcite .
we also discuss the possible applications of our qnd metrology scheme .
in our qnd metrology proposal , the experimental setup is similar to the one discussed in refs .
@xcite and is shown in fig .
the core of this setup is a `` photon box '' , which is an open high @xmath8 cavity @xmath9 made up of two superconducting mirrors facing each other ( fabry - prot configuration ) .
qnd probe atoms , generated form the atomic resource @xmath10 , are prepared in circular rydberg states and travel along the transverse direction of the cavity axis .
the atoms cross the cavity @xmath9 sandwiched between two auxiliary low @xmath8 cavities @xmath11 and @xmath12 before being detected in the detector @xmath13 .
the @xmath11-@xmath9-@xmath12 structure can be regarded as a ramsey interferometry .
the microwave field stored in the cavity @xmath9 is with frequency @xmath14 .
the atomic frequency is @xmath15 and is detuned from the cavity mode by @xmath16 ( @xmath17 ) .
, are generated in the atom source @xmath10 .
the interaction between the rydberg atoms and microwave pulse in auxiliary cavities @xmath11 and @xmath12 perform hadamard gate operation .
the unknown parameter is imprinted by the interaction of the atom and the superconducting cavity @xmath9 . after crossing the @xmath11-@xmath9-@xmath12 interferometric arrangement ,
the states of atoms are detected in the detector @xmath13 .
( b ) diagram for our sequential strategy . @xmath0
probe atoms are used .
the fock state of photons in @xmath9 stays unchanged after each qnd measurement.,title="fig:",scaledwidth=40.0% ] + suppose that the state of photons in the cavity @xmath9 is in a superposition of fock states with different photon numbers @xmath18 .
one rydberg atom is prepared in states @xmath19 ; and afterwards , for simplicity , we replace @xmath20 and @xmath21 with @xmath22 and @xmath23 . both pulses in ramsey cavities @xmath11 and @xmath12 are acting as an hadamard operation on each atom which is written as @xmath24 and transforms @xmath22 and @xmath23 to @xmath25 and @xmath26 , respectively .
the interaction between the probe atom and photons contains an unknown parameter @xmath4 ( see appendix [ a ] for details ) and can be expressed as a unitary operator : @xcite @xmath27 $ ] to the lowest order for small @xmath4 , where @xmath28 is the photon number operator in the cavity @xmath9 and @xmath29 is the pauli operator . the final state of photons and one probe atom after the probe atom passing through the @xmath11-@xmath9-@xmath12 is expressed as @xmath30 measurement on the atom in the detector , and the output is @xmath31 or @xmath32 with probability @xmath33 $ ] .
then , the photon state in @xmath9 is affected by this measure due to different outputs : @xmath34}{\sqrt{p(i|\theta)}}|n\rangle_{s}.\ ] ] it is easy to verify that @xmath35|0\rangle_{p}=0 $ ] , which is the general necessary and sufficient condition @xcite that the qnd probe must satisfy .
next , we present the experimental procedure for estimating a parameter , see fig .
we consider that the state of photons in the cavity is prepared as a fock state of @xmath1 photons , @xmath36 .
although generating a single mode fock state of @xmath1 photons in the cavity @xmath9 is a challenging task , it seems nowadays experimentally available @xcite .
a general theoretic review of this method is given in appendix [ b ] .
it is also possible to prepare and lock the field to on - demand photon number states by the real - time quantum feedback techniques reported in refs .
@xcite . preparing
the single - mode number squeezing state is also helpful and urgent for ultrasensitive two - mode interferometry @xcite .
it is easy to verify that , after the qnd measurements procedure discussed above , a fock state stays unchanged @xcite . because @xmath37 $ ]
, the parameter @xmath4 can be estimated from the readouts of @xmath38 measurements performed on a sequence of probe atoms interacting with the light in @xmath9 . for each probe ,
the probability for the readout @xmath31 or @xmath32 is @xmath39 $ ] . with the assumption that the estimation is asymptotically unbiased
, we can utilize fisher information ( fi ) @xmath40^{2}$ ] and cramr - rao bound @xmath41 @xcite to calculate the statistical precision of the estimation .
fi is calculated as @xmath42 which leads to a lower bound : @xmath43 therefore , by using this qnd metrology technique , our proposal is able to achieve the @xmath2 scaling accuracy of parameter estimate with only one initial @xmath1-photon fock state for @xmath0 trials of measurements .
the advantages of this qnd metrology strategy are : the initial photon state is entanglement - free state which is more robust than the exotic states ( e.g. noon states ) , and the fock state stays unchanged after qnd measurements and can be used repeatedly .
technically , for our scheme , the total resource can be written as @xmath44 for a sufficiently large @xmath0 , and the lower bound is expressed as @xmath45 . when @xmath0 copies of noon states are used to achieve the same accuracy ,
the total resource are @xmath46 and the lower bound is @xmath47 .
although both strategies do not achieve the heisenberg limit @xmath48 , given the same total resources , our scheme gives @xmath49 advantage compared with the strategy using noon states and @xmath2 advantage over sql @xmath50 .
since the bohr - einstein photon box thought experiment , experiments with circular rydberg atoms and fabry - prot have become closest to this goal .
they have also led to fundamental tests of quantum theory and various demonstrations of quantum information procedures @xcite . here ,
based on the developments and advances made in the cavity quantum electrodynamics in the microwave domain , we discuss the feasibility of our qnd metrology scheme using current laboratory parameters .
it is reported in ref .
@xcite an ultrahigh finesse fabry - prot resonator @xmath51 ghz with cavity damping time @xmath52 s and cavity quality factor @xmath53 at 0.8 k ( mean number of blackbody photons @xmath54 ) .
the damping rate is given as @xmath55 . in ref .
@xcite , rydberg atoms are prepared by a pulsed process repeated at @xmath56 @xmath57s time intervals with selected atomic velocity @xmath58 m / s . for @xmath0 trials of measurement , the total time is @xmath59 and the photon - loss intensity in the cavity is written as @xmath60 .
in fact , if we shorten the interval @xmath61 , the number of measurement @xmath0 can be large with a low photon - loss intensity .
this long damping time and qnd detection technique can stabilize the fock state in the cavity @xmath9 and make our metrology scheme practicable . in the next section
, we will discuss the effect of cavity damping in detail .
the technologies of other experimental procedures , such as generation of rydberg atoms and polarizing measurement on the atoms , should be feasible and mature referring to ref .
the quantum metrological bounds in noisy systems have become a focus of attentions because in real experiments there will always be some degree of noise and limitation .
the fock state prepared in the cavity @xmath9 mainly suffers from cavity damping .
given a certain the damping rate @xmath62 , the interaction picture of reduced density operator for the field in the cavity @xmath9 under the born - markov obeys the master equation @xcite : @xmath63 where @xmath64 ( @xmath65 ) is the annihilation ( creation ) operator for field in cavity . considering the radiation field with a reservoir at nearly zero temperature @xmath66 , we approximately express the density operator for the field by the well - known photon loss model : @xmath67 .
the initial probe state is still prepared as @xmath68 .
the density operator form of the final state can be written as @xmath69 and the final reduced state for probe atom is @xmath70 where we set @xmath71 and @xmath72 .
then we perform the @xmath73 measurement on the atom and obtain the results @xmath74 and @xmath75 with probabilities @xmath76/{2}$ ] . given photon number @xmath77 , @xmath78 is shown in fig .
[ f2](a ) for different values of lossy intensities .
we can therefore calculate fi as @xmath79\}^{2}}/[{1-r^{2n}\cos^{2}(n\varphi)}]$ ] where we have used @xmath80 and @xmath81^{2}+\sin^{2}\theta(1-\eta)^{2}}+\frac{1}{2n}$ ] . for @xmath77 ,
fi is @xmath4 dependent for a nonzero @xmath82 , see fig .
[ f2](b ) . is set as @xmath77 and cavity damping time is @xmath83 s. ( a ) probability @xmath78 against parameter @xmath4 for @xmath84 , @xmath85 and @xmath86 .
( b ) fi and qfi against parameter @xmath4 for @xmath84 , @xmath85 and @xmath86 .
( c ) the decay of the optimal qfi @xmath87 compared with the decay of fock state .
( d ) the average qfi @xmath88 compared with the optimal qfi @xmath87 against number of measurements @xmath0 with measurement interval @xmath56 @xmath57s.,title="fig:",scaledwidth=48.0% ] + by choosing the optimal measurement , we can obtain the maximum fi which is also called quantum fisher information ( qfi ) . given the spectral decomposition of final reduced state for probe atom , @xmath89 , qfi can be written with condition @xmath90 as @xmath91\ ] ] where we have used @xmath92 and @xmath93 .
comparing fi and qfi in fig .
[ f2](b ) , we can achieve the optimal fi and qfi as we carefully choose @xmath94 : @xmath95^{2}+\eta(1-\eta)n\end{aligned}\ ] ] where for large @xmath1 , we obtain that @xmath96^{2}$ ] .
therefore , with the two - step adaptive method based on bayesian estimation @xcite , the optimal qfi can be achieved with the same measurement on the probe atoms used in the noiseless case .
although the lifetime of the fock state @xmath97 is @xmath98 , much less than the cavity damping time @xmath99 , the optimal qfi for this qnd strategy decays much slower as shown in fig .
[ f2](c ) .
it has been recognized that photon losses in interferometers gradually blur the gain yielded by the special quantum states for parameter estimation ; even with the best strategy , asymptotically the improvement with respect to standard light sources is not by a scale change but only by a limited constant factor @xcite . unlike the interferometric strategy with exotic states , e.g. noon states
, the qnd metrology scheme with fock state will sustain the quantum advantage for a longer time than the damping time of fock state .
for instance , given photon number @xmath77 and cavity damping time @xmath83 s , the quantum advantage remains until @xmath100 s for @xmath101 and the number of trials can be @xmath102 for time interval @xmath56 @xmath57s .
we can also define the average qfi for @xmath0 trials of measurement as @xmath103 , with which we can write the lower bound of estimate accuracy for decoherence scenarios as @xmath104 . since we show in fig .
[ f2](d ) that the quantum - enhanced estimation against decoherence does not limit the number of measurements to be too small ( @xmath105 ) , we can conclude that this qnd metrology scheme is expected to be robust against the cavity damping .
let us assume that the phase to be estimated lies in the interval @xmath106 .
one common but intractable problem in the quantum - enhanced metrology is the @xmath3 periodic error in the estimation of the true phase if @xmath107 @xcite . to address this problem
we will next extend the cascaded protocol reported in ref .
@xcite to our qnd method .
it is realizable with the help of the mature technology of state control in the `` photon box '' @xcite .
our cascaded scheme employs @xmath108 successively larger fock states of @xmath109 photons .
we use @xmath0 rydberg atoms as the qnd probe for each fock state .
the total resource used in this cascaded scheme is @xmath110 .
the interaction with the fock state consisting of @xmath111 photons picks up the phase @xmath112 , where @xmath113 .
the real phase to be estimated can be written in an exact binary representation @xmath114 , with digits @xmath115 . by distinguishing whether the phase is shifted by @xmath116 or not , we can determine the value of the bit @xmath117 according to the relation @xmath118/2\pi$ ] .
we should note that the rounding error @xcite that occurs whenever @xmath119 can be neglected given a large number of trials @xmath0 .
the last group @xmath120 then yields a heisenberg type limited estimate of the parameter with accuracy @xmath121 which is merely less sensitive by a constant @xmath122 compared with eq .
( 3 ) .
axis , we need another real - time feedback interferometric setup to prepare the target fock state in @xmath123 axis : @xmath124-@xmath9-@xmath125 interferometric arrangement and detector @xmath126 . the detection results from detector @xmath127 are sent to the computer based controller @xmath128 .
the controller @xmath128 analyzes each detection result and determines the the real translation amplitude @xmath129 applied by actuator @xmath130.,title="fig:",scaledwidth=38.0% ] + to realize this cascaded qnd metrology scheme , it is necessary to prepare and lock the field to different photon number states during each qnd measurement .
this requirement can be fulfilled by the real - time quantum feedback techniques reported in refs .
@xcite . the experimental layout of our proposal is shown in fig . [ f3 ] .
the layout is supposed to work in two modes : phase estimation mode and target state preparation mode . since only one mode works at the same time , we need two interferometric setups .
in addition to the setup used for the phase estimation mode , another interferometric setup is performed to prepare and stabilize the successively larger target fock states .
the computer based controller @xmath128 controls the conversion between those two modes ( more details are shown in appendix [ b ] ) .
in our scheme , the unknown parameter expressed as @xmath131 is determined by the atom velocity @xmath132 and the position @xmath133 in the cavity @xmath9 , see appendix [ a ] for more details . here
, we use the cylindrical coordinates @xmath134 , @xmath135 is the speed of light in vacuum , @xmath136 is the waist at center @xmath137 and @xmath138 is the vacuum rabi frequency at center .
therefore , this high precision quantum - enhanced measurement can be used to detect and measure the mini - displacement of the cavity @xmath9 along @xmath133 axis . on the practical perspective
, a high sensitivity in @xmath4 leads to the high sensitivity in the displacement @xmath133 when we measure the small displacement around the maximum slope point @xmath139 .
the accuracy can be obtained by straightforward error propagation , @xmath140 where @xmath141 .
using the current laboratory parameters shown in sec .
iii and @xmath142 khz , @xmath143 mm and @xmath144 khz in ref .
@xcite , one obtains that @xmath145 mm . with @xmath146 and @xmath77 ,
the sensitivity is @xmath147 @xmath57 m and it can be improved by reducing the atom - cavity detuning and velocity of atom or increasing @xmath1 and @xmath0 .
however , these methods for improvement may seem challenging in the real experiment , for instance , reducing the wavelength of the light in the cavity will then make it difficult to place the atoms at the maximum slope point .
although this sensitivity by now seems several orders of magnitude worse than the sensitivity needed for gravitational wave observation , our scheme will inspire future experiments demonstrating quantum - enhanced metrology and may be helpful to prospective applications in other experimental platforms .
in this letter , we have presented an experimental proposal for estimating an unknown parameter in `` photon box '' by using the method of qnd measurements .
we have shown that initially with fock state of @xmath1 photons , the @xmath2 scaling accuracy of the estimation can be achieved . moreover
, we do not need to prepare @xmath0 copies of initial state as other metrology schemes , which will give a @xmath49 advantage compared with the scheme using noon states when the total resource is taken into consideration .
we also show that this sub - shot - noise estimation scheme is robust against cavity damping .
the feasibility of our scheme can be met by the current laboratory achievements and it can be improved via a cascaded scheme to overcome the @xmath3 periodic error . furthermore , this proposal with the help of qnd measurements will also be an inspiration to other experimental platforms @xcite for quantum metrology and quantum information techniques .
in addition , our results should be of broad interest as many applications , such as clock synchronization and phase imaging .
since generating and using the noon states with more than @xmath148 photons @xcite for quantum metrology is still an arduous task , researchers have been able to generate fock state with @xmath149 or even more photons in the cavity .
that is to say if our scheme can be realized in the experiments , it will be a great advance in the research area of quantum metrology and quantum physics .
we would like to thank augusto smerzi and mehdi ahmadi for useful discussions .
this work was supported by the `` 973 '' program ( 2010cb922904 ) , nsfc ( 11175248 ) grants from the chinese academy of sciences .
we describe in this section the interaction between the rydberg atom and photons in the cavity in a concise form .
this simple case will provide us the phase shift linearly given by per photon .
the interaction can be described via the hamiltonian of jaynes - cummings model @xcite @xmath150 where @xmath64 ( @xmath65 ) is the photon annihilation ( creation ) operator in the cavity @xmath9 .
let the cavity @xmath9 contain @xmath151 photons . for large detuning frequency ,
the atom - field states @xmath152 and @xmath153 evolve into dressed states and are shifted in angular frequency units by @xcite @xmath154 where @xmath155 is the vacuum rabi frequency following the gaussian distribution at cavity center @xmath156 ; here we use cylindrical coordinates @xmath134 where @xmath133 is the position of the atom along the beam axis .
@xmath135 is the speed of light in vacuum , @xmath136 is the waist at center @xmath137 and @xmath138 is the vacuum rabi frequency at center .
the difference of the phase imprinted on atomic states should be expressed as @xmath157 where @xmath158 with @xmath136 the waist at center and @xmath132 the atom velocity .
therefore , the interaction between the probe atom and photons can be expressed with the parameter @xmath159 as a unitary operator : @xmath27 $ ] to the lowest order for large detuning frequency .
the value of the parameter @xmath159 is determined by the atom velocity and the position in the cavity @xmath9 , which will be of great value in scientific and engineering applications .
in this section , we review two different methods of preparation of fock state in the `` photon box '' .
the stochastic method is expected to require less equipments and be suitable for a demonstrative experiment of quantum - enhanced metrology .
the deterministic method needs additional experimental devices and is able to generate a fock state with on - demand photon number , which is helpful to the cascaded strategy .
+ we start with @xmath18 and set the interaction parameter @xmath160 at an appropriate and definite value such that @xmath161 for all possible photon number @xmath162 , where @xmath163 $ ] .
suppose that @xmath164 atoms cross the @xmath11-@xmath9-@xmath12 interferometric layout and are measured with operator @xmath165 .
the sequence of measurement results for @xmath164 probe atoms , called a event , is expressed as @xmath166 , where @xmath167 and @xmath168 .
the photon number distribution of the final state in @xmath9 can be calculated as @xmath169\sin^{\xi}[(n+\frac{1}{2})\theta_{s}/2]|^{2}}{z(\bm{\omega}_{m})},\ ] ] where @xmath170\sin^{\xi}[\frac{(n+{1}/{2})\theta_{s}}{2}]|^{2}$ ] ; @xmath82 and @xmath171 are the number of @xmath75 and @xmath32 in the event @xmath172 , respectively .
it has been proved in ref .
@xcite that ( @xmath173 ) this photon number distribution converges as @xmath164 becomes infinity : @xmath174 , ( @xmath175 ) the probability for the state in cavity converges to a fock state @xmath97 is @xmath176 , and ( @xmath177 ) the convergence for @xmath178 is exponentially fast .
therefore , we can obtain a single mode fock state @xmath97 by this approach with probability @xmath176 when @xmath164 is large enough .
the final photon number @xmath1 can be determined via analyzing the spin measurement results of the probe atoms @xmath172 . after generating a fock state of a nonzero and known photon number
, we can perform the parameter estimation without adjusting the experimental apparatus .
most commonly the initial state of the light in cavity @xmath9 is a coherent state @xmath179 .
the qnd measurement procedures are numerically simulated via monte carlo method and plotted in fig .
we observe the converging events of different photon numbers in fig .
[ sf1](a ) and [ sf1](b ) . in fig .
[ sf1](c ) and [ sf1](d ) , we present the situation that the special condition @xmath180 is saturated and the convergent states are superposed fock states .
we also numerically simulate the probability for the coherent state converging to a fock state , in fig .
[ sf2 ] , which conforms with the experimental results in ref . @xcite .
the average fisher information for all possible fock states from the initial coherent state @xmath181 can be written as @xmath182 which leads to the heisenberg - type lower bound : @xmath183 $ ] . instead of the coherent state , an efficient method for improving
this strategy is to use the squeezed state @xmath184 which may be generated by first acting with the squeeze operator @xmath185 on the vacuum followed by the displacement operator @xmath186 @xcite .
we can obtain via squeezed state a higher success rate for generating a useful fock state for sub - shot - noise metrology due to its super - poissonian and narrower photon number statistics , see fig .
[ sf2 ] . and
squeezed state ( ss ) @xmath187 are plotted by dashed blue line and dash - dotted magenta line , respectively .
@xmath188 convergence events are simulated to figure out the photon number distribution probability numerically . , title="fig:",scaledwidth=35.0% ]
+ in order to prepare and lock the field to different photon number states during each qnd measurement , it is necessary to use real - time quantum feedback techniques reported in refs .
@xcite to fulfill this requirement .
the experimental layout of our proposal is shown in fig . 3 in the main text .
in addition to the estimation interferometric setup ( @xmath189 direction ) shown in fig . 3 ,
a quantum feedback setup is put in the @xmath123 direction : another atom source @xmath190 generates test rydberg atoms , and two auxiliary cavities @xmath124 and @xmath125 act as two hadamard gates . the information on the interaction between the test atom and the field is assumed to be known .
the measurement results of the test rydberg atoms obtained by detector @xmath127 are sent to the computer based controller @xmath128 . by analyzing each detection result
, the controller @xmath128 updates the photon distribution @xmath191 and controls actuator @xmath192 to feed cavity @xmath9 by diffraction on the mirror edges @xcite .
the controller @xmath128 analyzes each detection result to determine the real translation amplitude @xmath129 applied by actuator @xmath192 which minimizes the distance @xmath193 @xcite between the target @xmath194 and the field estimation @xmath195 . here
@xmath186 is the displacement operator .
when the controller @xmath128 finds that @xmath196 , it stops atom source @xmath190 and the target fock state has been prepared .
it is reported that fock states with photon numbers @xmath197 up to 7 can be prepared with number photon number distribution peaked @xmath198 @xcite .
then , controller @xmath128 activates another atom source @xmath10 and @xmath0 probe states are sent and detected to estimate the unknown phase .
we will next show that this technique based on real - time quantum feedback is also necessary to realize the cascaded scheme @xcite . and
@xmath199 against time for photon number @xmath77 .
solid lines are for @xmath200 and dashed lines are for @xmath199 .
inset : fidelity of @xmath200 and @xmath199 against time.,title="fig:",scaledwidth=35.0% ] + one common but intractable problem in the quantum - enhanced metrology is the @xmath3 periodic error @xcite in the estimation of the true phase if @xmath107 .
this problem can not be settled by simply adding an ancillary phase so that the cascaded protocol is proposed to solve it . to realize the cascaded qnd metrology scheme , it is necessary to prepare and lock the field to different photon number states during each qnd measurement .
the first target state starts with @xmath201 for @xmath202 .
when the target fock state @xmath203 for @xmath113 has been prepared , controller @xmath128 activates another atom source @xmath10 and @xmath0 probe states are sent and detected to estimate the unknown phase with value @xmath204 .
after the estimation using this fock state @xmath205 is finished , controller @xmath128 activates the feedback setup and the target state is changed to the next fock state @xmath206 .
the last target fock state that we consider has @xmath207 photons . when all @xmath204 for @xmath113 have been obtained , we can calculate digits @xmath117 for @xmath208 and retrieve the true value of phase @xmath4 within a accuracy presented in eq .
( 8) . as a brief summary
, the layout is supposed to work in two modes : target state preparation mode and phase estimation mode . since only one mode works at the same time ,
we need two interferometric setups . in addition to the setup used for the phase estimation mode
, another interferometric setup is performed to prepare and stabilize the successively larger target fock states .
the computer based controller @xmath128 controls the conversion between those two modes .
the exact solution of the master equation ( 4 ) in the main text can be expressed by the density operator : @xmath209 with probabilities @xcite @xmath210^{n+n+1}}.\ ] ] in recent experiments @xcite , the mean number of blackbody photons @xmath211 at 0.8 k. thus , we simply choose @xmath212 with @xmath213 and use an approximate solution @xmath67 to calculate the noisy case with cavity damping . in fig .
[ bb ] , we compare the probabilities of these two states against time for photon number @xmath77 ; solid lines are for @xmath200 and dashed lines are for @xmath199 .
we also show in the inset figure of fig .
[ bb ] that the fidelity of these two states @xmath200 and @xmath199 against time is always larger than 0.99 .
therefore , this approximation with which we can obtain a analytical result of fisher information and quantum fisher information is proper .
s. haroche and j. m. raimond , _ radiative progress of rydberg states in resonant cavities _
, in _ advances in atomic and molecular physics _ , edited by d. r. bates and l. esterman ( academic , new york , 1985 ) , vol .
20 , page 347 - 411 . | in quantum metrology schemes , one generally needs to prepare @xmath0 copies of @xmath1 entangled particles , such as entangled photon states , and then they are detected in a destructive process to estimate an unknown parameter . here ,
we present a novel experimental scheme for estimating this parameter by using repeated indirect quantum nondemolition measurements in the setup called `` photon box '' .
this interaction - based scheme is able to achieve the phase sensitivity scaling as @xmath2 with a fock state of @xmath1 photons .
moreover , we only need to prepare one initial @xmath1-photon state and it can be used repetitively for @xmath0 trials of measurements . this new scheme is shown to sustain the quantum advantage for a much longer time than the damping time of fock state and be more robust than the common strategy with exotic entangled states . to overcome the @xmath3 periodic error in the estimation of the true parameter
, we can employ a cascaded strategy by adding a real - time feedback interferometric layout . |
dark energy is widely described as the biggest problem in cosmology today ; one which may demand new physics and a theoretical paradigm shift@xcite . in this paper
i suggest that the solution to the problem of dark energy is intimately related to the correct understanding of observational anomalies in particular , the observed abundance and emptiness of voids , which has also elicited separate calls for a paradigm shift@xcite .
i propose that the paradigm shift that is required to understand both these issues actually entails no `` new '' physics , but a revisiting of fundamental issues relating to the subtlety of the definition of gravitational energy within general relativity , and its relation to the careful modelling of the distribution of matter that we actually observe .
the punchline is that cosmic acceleration can be understood as an apparent effect , and dark energy as a misidentification of those aspects of cosmological gravitational energy which by virtue of the strong equivalence principle can not be localized@xcite , namely gradients in the quasilocal gravitational energy associated with spatial curvature gradients , and the kinetic energy of expansion , between bound systems and the volume average position in freely expanding space . with this interpretation ,
a two scale model can be constructed@xcite , and a simple exact solution@xcite of the relevant equations of cosmic evolution , the buchert equations@xcite , can be found .
this solution provides a fit to type ia supernovae ( sneia ) luminosity distances which is statistically indistinguishable from that of the standard lambda cold dark matter ( ) model , while simultaneously satisfying key independent cosmological tests and offering the potential to resolve significant observational anomalies@xcite . over the past decade
we have come to think of `` dark energy '' as a homogeneous isotropic fluid like quantity , with a local pressure related to its energy density by @xmath0 , where @xmath1 , so that the strong energy condition is violated .
in fact , observations appear to be most consistent with a pure vacuum energy , @xmath2 , and determination of the value of the parameter , @xmath3 , as a function of redshift , @xmath4 , is seen as the goal of `` precision cosmology '' .
what is proposed here , if correct , will turn this situation on its head .
we can look forward to an era of precision cosmology , but one in which the focus will be on the complex hierarchical structure of the universe , rather than any one simple fluid equation of state . nonetheless , while the proposed solution is intimately connected to the growth of inhomogeneities , and their backreaction on the geometry of the universe @xcite , at its heart it addresses the question of the normalization of gravitational energy relative to observers within the observed structure .
thus i claim that the solution to the central foundational question does concern energy , and since `` nothing '' is `` dark '' the terminology `` dark energy '' is actually quite apt for the new proposal , if the community will allow the liberty of a change to the assumed definition of those words . in this paper
i will give an overview of the proposed solution to the problem of dark energy , the extent of its present quantitative successes , and more importantly the directions for future work .
i will present fewer technical details than may be found in papers already published@xcite or in preparation@xcite .
i aim to give a general overview to researchers in astrophysics , particle physics or general relativity , who have no specific prior experience with the averaging of inhomogeneous cosmological models .
our most widely tested `` concordance model '' of the universe relies on the assumption of an isotropic homogeneous geometry at all epochs of cosmic evolution . by the evidence of the cosmic microwave background ( cmb ) radiation ,
the universe was very smooth at the time of last scattering , and these assumptions were completely well justified then .
the departure from homogeneity was of order @xmath5 in photons and the baryons that couple to them , and perhaps of order @xmath6 in non baryonic dark matter , which gives rise to the potential wells responsible for the dominant sachs
wolfe effect . by the copernican principle ,
the assumption of global isotropy and homogeneity is completely justified at the epoch of last scattering , and it is safe to assume that the evolution of the universe was therefore extremely closely modelled by the friedmann lematre robertson
walker ( flrw ) solutions at that epoch .
furthermore , if we consider the spectrum of cmb anisotropies @xcite then overall it appears that globally the universe is very close to spatially flat . its initial evolution at the time of last scattering would therefore have been very close to that of an einstein de sitter model at that epoch , even in the case of the paradigm since `` dark energy '' only becomes appreciable at late epochs . at the present epoch , however , the distribution of matter is far from homogeneous on scales less than 150300 mpc .
the actual universe has a sponge like structure , dominated by huge voids .
these voids are surrounded by bubble walls , and threaded by filaments , within which clusters of galaxies are located .
recent surveys suggest@xcite that some 4050% of the present volume of the universe is in voids of a characteristic scale 30@xmath7 mpc , where @xmath8 is the dimensionless hubble parameter , @xmath9 .
if smaller minivoids @xcite and larger supervoids @xcite are included , then our observed universe is presently void dominated by volume . quite apart from the fact that this observed structure appears emptier than the vistas that newtonian
@xmath10body simulations typically produce , the mere fact that the universe is presently inhomogeneous means that the assumptions implicit in the flrw approximation can no longer be justified at the present epoch in the almost exact sense that they were justified at the epoch of last scattering .
homogeneity only applies at the present day in an _
average sense_. the manner in which we take the average , and the operational issues associated with this are not trivial , since the problem of fitting the local geometry of bound systems to the full dynamical geometry of the evolving universe @xcite is a complicated one .
since a broadly isotropic hubble flow is observed , an almost flrw geometry would appear to be a good approximation at some level of averaging , if our position is a typical one .
i will demonstrate , however , that the assumption that the observed universe evolves exactly as a smooth featureless dust fluid means that we factor out central physically important questions which need to be understood to correctly relate our own observations to the average geometry .
there is a central foundational paradox concerning the expansion of the universe , which others have called the `` hubble de vaucouleurs paradox '' @xcite or the `` hubble sandage paradox '' @xcite , but which i will call the _ sandage de vaucouleurs paradox _ since it was originally raised by sandage and collaborators @xcite in objection to de vaucouleurs hierarchical cosmology @xcite .
the problem is that in the conventional way of thinking about cosmological averages , below the scale of homogeneity we should expect large statistical scatter in the peculiar velocities of galaxies .
in fact , if we were to average on scales of order 20mpc , which is about 10% of the scale of homogeneity , then the statistical scatter should be so large that no linear hubble law can be extracted .
yet 20mpc is the scale on which hubble originally extracted his famous linear law . by conventional understanding
the statistical quietness of the local hubble flow does not make sense .
one might attempt to explain the sandage de vaucouleurs paradox as a consequence of dark energy , since it is well known that in any flrw model which expands forever with or without dark energy peculiar velocities eventually decay .
however , if one tries to numerically model 12.5 gyr of motion of very local galaxies , then it turns out that the initial conditions of each galaxy appear to have the most bearing on the problem@xcite .
in particular , one can fit a realistic motion with a cosmological constant , @xmath11 , or alternatively in an open universe without @xmath11 .
similar conclusions are also reached in constrained dark matter simulations of the local group @xcite .
furthermore , the parameters required for the velocity dispersion predicted by structure formation to match the observed velocity dispersion , do not coincide with the concordance parameters@xcite .
evidently a cosmological constant alone can not explain the quietness and linearity of the local hubble flow .
a related , but not exactly equivalent , issue is that using a conventional kinematic approach , the peculiar velocities of local galaxies appear to be at least a few factors too small to have arisen from a smooth distribution of matter at the time of recombination@xcite .
the present distribution of matter is clearly very complex , and since we can not solve the einstein equations for this distribution of matter analytically , there is an important question as to how we operationally match the average geometry of this distribution to the simple flrw models that we know how to solve . given that the nonlinear growth of structure appears to be roughly correlated to the epoch when cosmic acceleration is inferred to begin , a number of cosmologists have questioned whether the introduction of a smooth dark energy is a mistaken interpretation of the observations @xcite .
attention has focused on the possibility that effects attributed to cosmic acceleration may actually be due to the backreaction from the growth of inhomogeneities in determining the geometry of the observed universe , without exotic dark energy .
different interpretations of a host of complex technical issues have led to a vigorous debate .
there are two large streams of research , which i will not focus on , for physical reasons .
the first is backreaction in perturbation theory about an flrw background , which became the focus of debate in 2005 following public attention generated by papers of kolb and collaborators@xcite . within perturbation theory
one may demonstrate that there is a potentially significant effect from backreaction@xcite
. however , this argument can not be conclusive if the second order terms in the perturbation expansion exhibit an effect which might be interpreted as cosmic acceleration , such an effect may go away when the third order terms are considered , and so on .
perturbation theory is very relevant near the epoch of last scattering , when the assumption of homogeneity was extremely good ; but by the present epoch the nonlinear structures are so numerous and complex that we are beyond the regime of its applicability .
thus perturbation theory can not provide a complete solution , and will not be discussed here . the second stream of research i will not consider are those that involve exact inhomogeneous solutions of the einstein equations , at the expense of introducing matter distributions which are unlikely .
the spherically symmetric lematre tolman
bondi ( ltb ) solutions are perhaps the most well studied class of such models . while they may be very realistic descriptions of single voids , to apply them to the universe as a whole violates the copernican principle , which i shall retain .
one may obtain ltb solutions within which one can obtain reasonable fits to supernovae luminosity distances@xcite .
however , in my view , given their high degree of symmetry , these are at best toy models , which one can not hope to reproduce in structure formation scenarios based on our knowledge of the power spectrum of density perturbations at the time of last scattering . to confront the actual inhomogeneous universe , which has no particular spatial symmetries below the scale of homogeneity
, we must deal with schemes that average the full non linear einstein equations .
there are many schemes for constructing averages , including those of zalaletdinov@xcite and buchert@xcite .
there are various grounds for favouring the approach of zalaletdinov@xcite , which is fully covariant and averages all of the einstein equations .
however , zalaletdinov s scheme is a general one , and for the cosmological problem additional assumptions are required . in buchert
s scheme one just average scalar quantities , and an additional integrability condition is required for the equations to close . however , with reasonable cosmological assumptions , the correlation tensor in zalaletdinov s scheme takes the form of a spatial curvature @xcite , and buchert s scheme can be realized as a consistent limit @xcite . furthermore buchert s scheme yields equations which are close to the friedmann equations .
given that the friedmann equations have worked so well to date , buchert s average would appear to give a natural framework within which corrections to the friedmann evolution can be quantitatively examined for the universe we observe .
i shall adopt buchert s scheme , with caveats to be discussed .
buchert s scheme strictly deals with irrotational dust cosmologies , characterized by an energy density , @xmath12 , expansion , @xmath13 , and shear , @xmath14 , on a compact domain , @xmath15 , of a suitably defined spatial hypersurface of constant average time , @xmath16 , and spatial 3metric , @xmath17 .
angle brackets are taken to denote the spatial volume averages , e.g. , for the scalar curvature @xmath18 with @xmath19 .
the important lesson of buchert averaging is that time evolution and averaging to do not commute@xcite .
generally for any scalar @xmath20 , t-=-[noncom]the fact that the r.h.s . of ( [ noncom ] ) does not vanish , as is the case for the flrw cosmologies , is a manifestation of _
backreaction_. applied to the equations of cosmic evolution one obtains the exact _ buchert equations _ 3 ^ 2 ^ 2&=&8g--,[buche1 ] + 3&=&-4g+,[buche2 ] + _ t&+&3=0 , [ buche3]where @xmath21^{1/3}$ ] , 23(^2-^2)- 2 ^ 2 [ backr]and following integrability condition follows from ( [ buche1])([backr ] ) : _ t(^6)+^4_t(^2)=0.it is observed that the kinematic backreaction term , @xmath22 , enters ( [ buche2 ] ) with the same sign as a cosmological constant in the equivalent raychaudhuri equation in the flrw paradigm , but enters ( [ buche1 ] ) with the _ opposite _ sign to a cosmological constant in the friedmann equation .
given this fact , together with the domain - dependence of the average , the question of whether buchert s scheme can generate sufficient backreaction to give apparent acceleration for realistic initial conditions has been the subject of much debate . on reasonable grounds
one might conclude@xcite that while the effects are real , they are too small in magnitude to give departures sufficiently large from the flrw expectation to register as cosmic acceleration .
it is at this point in the argument caution must be exercised . as every student of general relativity knows , one can not simply write down a time parameter and assume that it is the parameter one measures on one s clock , without specifying how it is related to local invariants .
we actually measure luminosity _ distances _ , and the deduction of cosmic acceleration involves two time derivatives .
we therefore have to be extremely careful to operationally specify how @xmath16 is to be related to our own clocks .
it must be observed that the scale factor @xmath23 is not related to an exact local metric , that has been substituted in einstein s equations .
rather we must solve the buchert equations , and determine how the best
fit almost flrw geometry that is obtained relates operationally to our own measurements .
in a completely arbitrary inhomogeneous universe the calibration of rods and clocks at one point relative to another can vary arbitrarily .
however , our clocks and rods and those of stars in distant galaxies we observe , appear to a very good approximation to be determined by geodesics of ideal solutions the kerr and schwarzschild geometries in which space is asymptotically flat , with an exact time symmetry at spatial infinity .
this time symmetry , mathematically described by a timelike killing vector , must be an idealization which breaks down at some level , since the universe is expanding . in his pioneering work on the fitting problem ,
ellis@xcite suggested its solution should involve the notion of _ finite infinity _ , is appropriate - in latex : @xmath24 .
] , `` @xmath25 '' , namely a timelike surface within which the dynamics of an isolated system such as the solar system can be treated without reference to the rest of the universe . after all , the matter in our galaxy and other typical galaxies broke away from the hubble flow to be come a bound system well over 10 billion years ago . with sufficient computing resources
, we can integrate the motions of stars within our galaxy for billions of years without considering the dynamics of the universe outside the galaxy .
thus within finite infinity spatial geometry might be considered to be effectively asymptotically flat , and governed by `` almost '' killing vectors .
the concept of finite infinity does not appear to have been further developed in the intervening two decades . however , given that the normalization of our clocks in the idealized schwarzschild and kerr geometries is related to the timelike killing vector at spatial infinity , finite infinity would appear to be the appropriate reference surface for the definition of gravitational energy , which in stationary spacetimes is tied to the asymptotically flat region@xcite .
in newtonian terms , it is the scale at which we set the zero of the gravitational potential .
the definition of gravitational energy in general is an extremely difficult problem , on account of the fact that space itself is dynamical and can carry energy and momentum . by the strong equivalence principle , since the laws of physics must coincide with those of special relativity at a point , it is only internal energy that can be localized in an energy
momentum tensor on the r.h.s . of the einstein equations .
any uniquely relativistic aspects of gravitational energy associated with spatial curvature and geometrodynamics can not be included in the energy momentum tensor , but are at best described by a quasilocal formulation@xcite .
einstein himself worried about the problem of quasilocal gravitational energy , in terms of energy
momentum complexes , and many mathematical relativists since einstein have also considered the problem .
it is quite possible that the general problem of a definition of quasilocal energy does not have a solution , since it depends on the split of space and time . here
i will not be interested in the general problem for an arbitrary solution of the einstein equations , but in the specific problem for a universe which was effectively homogeneous and isotropic at last scattering , with nearly scale invariant perturbations .
the fact that quasilocal energy is not part and parcel of theoretical framework of the flrw cosmology , is easily illustrated by the newtonian perspective @xcite .
the l.h.s . of the standard friedmann equation with @xmath26 can be regarded as the difference of a kinetic energy density per unit rest mass , @xmath27 and a total energy density per unit rest mass @xmath28 of the opposite sign to the gaussian curvature , @xmath29 , while the energy momentum tensor on the r.h.s.is the newtonian potential energy per unit rest mass .
the terms in the einstein tensor represent forms of gravitational energy , but since they are identical for all observers in an isotropic homogeneous geometry , one can always synchronize clocks and calibrate rods of ideal isotropic observers unambiguously . as soon as inhomogeneity enters the game , however , one will have gradients in both the kinetic energy of expansion , and in spatial curvature .
given the present value of the hubble constant , it is likely that the spatial curvature gradient is the most significant effect .
quasilocal energy gradients between a bound system and finite infinity could conceivably be relevant for asymptotic galactic dynamics , or even the solution of the pioneer anomaly .
this speculation is left for future work .
the proposal of refs.is concerned principally with dynamically evolving spatial curvature gradients . in this context
, one should observe that the idea that negative spatial curvature is associated with positive gravitational energy , evident already in the newtonian framework , remains true in the ltb models , where the gaussian curvature is replaced by an inhomogeneous energy function @xmath30 .
cosmological gravitational energy is largely uncharted territory in the more rigorous quasilocal framework @xcite within which the problem should ultimately be framed . in any quasilocal formulation results
depend crucially on the reference spacetime and surface of integration .
recently chen , liu and nester@xcite have obtained a result over which they expressed surprise , but which is consistent with the present proposal .
they find that for an isotropic observer in synchronous gauge in a @xmath31 friedmann universe the quasilocal energy in their particular hamiltonian formalism is negative .
a similar result is obtained using a different approach by garecki@xcite .
these results are expected in the current approach , since one is effectively subtracting a fiducial flat spacetime in each case , and the _ relative _ sign of energy depends on the observer .
an isotropic @xmath32 friedmann observer has zero quasilocal energy in the approach of chen , liu and nester ; thus relative to the @xmath31 geometry the @xmath32 geometry has negative quasilocal energy , but conversely relative to the @xmath32 geometry the @xmath31 has positive quasilocal energy .
our viewpoint here will be that the fiducial reference point is the @xmath32 geometry of the finite infinity region .
this agrees with the newtonian version of energy in the friedmann equation , the ltb energy function , and with the idea that binding energy is negative.@xcite
in the standard flrw cosmology the critical density is proportional to the square of the hubble parameter .
however , in an inhomogeneous cosmology this is no longer true , [ feq]as the average hubble parameter , @xmath33 , is related to the spatial scale of averaging and the clock chosen . in the spirit of the spherical collapse model , we are familiar with the idea that a perturbation near the time of last scattering can evolve as an flrw model of different density , until it turns around and collapses , by which point other regions closer to the true mean density are within the past light cone . in a universe which grew from a nearly scale invariant spectrum of perturbations , there will always be spatial correlations of some density perturbation which is relatively far from the mean , once one samples on scales close to the spatial scale of the particle horizon .
this is a simple consequence of cosmic variance , and is illustrated by the following analogy .
take a bucket filled with grains of sand of identical diameter , whose density is gaussian distributed about a mean .
provided the container is much larger than the individual grains then the mean density of the bucket can be expected to be extremely close to the mean density of all the sand from the beach on which it was collected .
if one repeats the exercise for stones of larger and larger diameter , the mean density of the bucket will on average differ more and more from the mean density of the beach as the diameter of the stones becomes comparable to the size of the bucket , by a @xmath34 statistic .
cosmic variance in the actual universe is complicated by three things .
firstly , since general relativity is causal , the bucket in the analogy is the past light cone which grows with time .
secondly , the perturbations are not uniform density but contain other perturbations embedded within them , like russian dolls .
thirdly , below a scale of apparent homogeneity the initial perturbations can no longer be thought of as perturbations .
they have undergone a non - linear transition to form the structure that we see : stars , globular clusters , and clusters of galaxies strung in filaments and bubbles around voids .
every inflationary cosmologist who has thought about the foundations of the subject realises that the present density of the observable universe need not be the same as the whole universe , observable and unobservable , and in general these should differ on account of cosmic variance in a scale free density perturbation spectrum .
nonetheless , in the flrw paradigm we unwittingly end up violating this .
as long as we assume that the average density evolves by the friedmann equation , we implicitly assume that the mean density in our past light cone today is the same as the whole ensemble of observable universes .
as long as we evolve the average geometry purely by the friedmann equation , with equality in ( [ feq ] ) , we will miss an important aspect of actual cosmic evolution .
furthermore , as long we study inhomogeneity in the context of a swiss cheese model @xcite by cutting and pasting holes into a background which still evolves by the friedmann equation alone , we will still fail to uncover essential features of cosmic evolution .
there is perhaps no obvious way to define finite infinity in an arbitrary inhomogeneous background . to proceed i make the crucial observation that since our universe was effectively homogeneous and isotropic at last scattering , a notion of a universal critical density scale did exist then .
it was the density required for gravity to overcome the initial uniform expansion velocity of the dust fluid .
i will assume , as is consistent with primordial inflation , that the present horizon volume of the universe was very close to the critical density at last scattering , with scale invariant perturbations .
in view of the existence of backreaction , what is required is a notion of true critical density without assuming evolution via the friedmann equation .
this should be a dynamically evolving demarcation scale between systems which will become bound and those that are unbound , given available data within the past light cone at any epoch . as outlined in ref .
[ ] , and depicted in fig .
[ fifig ] , finite infinity is defined in terms of a scale over which the average expansion , including virialized , collapsing and marginally expanding regions is zero , @xmath35 , while being positive outside . then @xmath36 .
finite infinity is a non static boundary analogous to the spheres cut out in the swiss cheese model@xcite , but it involves average geometry rather than matching exact solutions , and no assumptions about homogeneity are made outside finite infinity .
finite infinity represents a physical scale expected to lie outside virialized galaxy clusters , but within the filamentary walls surrounding voids .
since it is a scale related to the true critical density , space at finite infinity boundaries can be described by the spatially flat metric ^2=-^2+^2 ( ) , [ figeom]where @xmath37 .
while i use the term `` bubble wall '' it should be recognized that an ideal bubble wall would consist of finite infinity regions touching exactly at their boundaries . due to the existence of minivoids
one should take care in identifying the scale of local walls empirically .
an important issue that has to be faced is the split of space and time .
density is not a covariant quantity , but will depend on the foliation of hypersurfaces chosen to average on , and the specification of the observers within the hypersurfaces .
the hypersurfaces chosen by buchert are the standard ones based on surfaces of constant , @xmath16 , where this is the affine parameter on geodesics of ideal comoving observers .
this choice , which is analogous to the choice of synchronous gauge in flrw models , is always possible for an energy
momentum tensor with dust in the absence of vorticity .
while it is entirely reasonable that buchert s choice of gauge is consistent at the volume average position , provided one averages over suitably large spatial volumes , there is a problem of physical interpretation since observers in bound systems are located in places where the physical assumptions in buchert s gauge choice no longer apply .
as soon as regions start to collapse , geodesics will cross , vorticity comes into play and comoving coordinates can not be chosen globally .
actual observers are in locations where the geometry is better modelled by the vacuum schwarzschild and kerr solutions , rather than by a rotationless dust fluid .
since a nearly isotropic hubble flow is observed , it is clear that at some level a uniform hubble flow should be obtained , despite the large scale inhomogeneity .
i will make a choice of surfaces of homogeneity which _ implicitly solves the sandage de vaucouleurs paradox _ by the assumption that on suitably small scales of averaging _ the bare hubble flow is uniform below the scale of apparent homogeneity _ ,
t_r(t ) = 13=13==(t ) , [ homo]@xmath16 being the local proper time in each averaging region @xmath38 , and @xmath39 ^{1/3}$ ] a relevant proper distance . here
no @xmath38 can be contained within a finite infinity region .
the important point is that although it will appear to any one observer that voids expand faster than bubble walls , the clocks of observers within voids will also tick faster than those of observers at finite infinity on account of the relative gravitational energy gradient .
thus , provided the gravitational energy and spatial curvature gradients are correlated , apparent variance in the hubble flow for any one observer is nonetheless perfectly consistent with uniformity of the bare , or `` quasilocally measurable '' hubble flow . the uniform expansion gauge , which is equivalent to one of the standard gauges in cosmological perturbation theory@xcite , is given as an ansatz for defining the surfaces of homogeneity .
however , it should be noted that this ansatz is _ less restrictive _ than the standard ansatz implicit in the flrw models , where expansion is uniform , and all ideal isotropic observers also have synchronous clocks and measure the same spatial curvature on the relevant surfaces of homogeneity . here
uniformity in the bare hubble expansion is retained , but spatial curvature and gravitational energy of isotropic observers vary in a correlated fashion . it should be emphasized that the copernican principle is retained here .
if we compare ideal isotropic observers , some in galaxies and some in voids , they will also measure an isotropic cmb .
however , they will potentially measure different mean cmb temperatures , and different angular scales in the cmb anisotropies .
these two observations respectively relate to relative gravitational energy and relative spatial curvature .
while the clocks of all ideal observers are synchronized initially , when the flrw approximation was a good one , ultimately they will diverge ; by 38% on average at the present epoch . since we exchange photons with other bound systems , which are within finite infinity regions , such a large clock
rate variance is not directly detectable in experiments conceived to date .
the uniform bare hubble constant gauge is more than simply an ansatz which implicitly solves the sandage de vaucoulers paradox while dealing with the question of cosmological gravitational energy . as is argued in a separate paper@xcite , it may be realized as a consequence of the strong equivalence principle applied to expanding space . among the class of all possible motions of a timelike geodesic congruence
there is a class of conformal motions which _ do not isolate any direction in space as preferred _ , in contrast to individual boosts .
the statement then is that for such motions we can not distinguish the circumstance in which the particles are _ moving _ from a common origin in a static minkowski space , or alternatively are _ at rest _ in the underlying expanding universe .
the two situations are equivalent ; giving a _
cosmological principle of equivalence_@xcite . in terms of the conceptual basis of relativistic cosmology ,
this amounts to a refinement of the principle of inertia .
not only can we not distinguish which one of galileo s ships , or of einstein s trains is the one that is moving ; for conformal motions we can also not distinguish whether it is the collection of ships or trains that are moving , or the ` sea ' or ` railyard ' expanding in between .
in ref . [ ]
i have written down a buchert average of the einstein equations based on two scales , finite infinity regions within which the geometry is given by ( [ figeom ] ) , and voids within which the geometry is negatively curved with local scale factor , @xmath40 . the geometry near the centres of voids is given by ^2=-^2+^2 ( ) .
[ vogeom]the buchert equations are not written in terms of the local geometry ( [ figeom ] ) or ( [ vogeom ] ) at either finite infinity or the void centres , but in terms of an intermediate volume
average location with a volume average time parameter , @xmath16 , and an average scale factor , @xmath41 , defined by ^3=^3+^3,@xmath42 and @xmath43 being the respective initial void and wall volume fractions at last scattering .
initially @xmath44 , and @xmath45 , since the universe is homogeneous and isotropic at last scattering , evolving like an einstein de sitter one .
the buchert average is constructed over the entire present particle horizon volume .
although the actual surfaces of homogeneity will not coincide with buchert s ones on small scales , we can still make use of buchert s scheme , provided that we take care in defining the relationship between the volume average quantities of buchert s scheme , and those in a finite infinity region .
ultimately , in order to deal with actual gradients in spatial curvature between finite infinity and void centres , it may be necessary to use a scheme such as that of zalaletdinov @xcite , or to look at the problem in terms of ricci flow @xcite .
this may be necessary , for example , in determining the most appropriate gauge for the background to structure formation simulations .
the approach adopted in refs.is that the buchert equations are solved in volume average time , but the uniform expansion ansatz is used to calibrate observable quantities relative to observers at finite infinity .
this involves a dressing of cosmological parameters over and above the volume factors considered by buchert and carfora@xcite .
the independent buchert equations , including the integrability condition that ensures their closure , may be written & & + + = 1,[beq1 ] + & & ^-6_t(^2 ^ 6)+^-2_t(^2 ^ 2 ) = 0,[beq2 ] where the volume average matter , curvature and kinematic backreaction parameters are respectively = 8g0 ^ 33 ^ 2 ^ 3;=-^2/3 ^ 1/3 ^ 2 ^ 2;=-^29(1-)^2.[eomq]the average curvature is due to the voids only , which are assumed to have @xmath46 , an overdot denotes a derivative w.r.t .
volume average time , @xmath16 , and @xmath47 is the volume average or bare hubble parameter . although the quantity , @xmath48 , has no particular significance for a general buchert average , for our particular assumptions it represents the underlying uniform quasilocally measured hubble flow .
the quantities @xmath49 , @xmath50 and @xmath51 are then interpreted as the bare cosmological parameters , relevant to a comoving isotropic observer at an average position in freely expanding space , which will be within a void , but not at its centre .
these quantities are the closest analogues of the standard flrw density parameters .
it must be recalled that the scale factor @xmath41 does not correspond to an exact geometry substituted into the einstein equations , and then solved .
rather we integrate the buchert equations and then reconstruct the average spherically symmetric geometry ^2=-^2+^2()^2+a ( , ) , [ avgeom]the area function @xmath52 being defined by a horizon - volume average @xcite . the fact that ( [ avgeom ] ) is spherically symmetric reflects the fact that the geometry is reconstructed by taking an average on our radial null geodesics .
it is therefore _ not _ an ltb model , since exact spherical symmetry has not been assumed in the einstein equations .
since ( [ avgeom ] ) does not correspond to the local geometry of a finite infinity observer , if we assume that our rods and clocks differ little in calibration from those at finite infinity , we still have to be careful in relating our local geometry ( [ figeom ] ) to ( [ avgeom ] ) .
we do this by conformal matching of the radial null geodesics of ( [ figeom ] ) and ( [ avgeom ] ) , which requires that @xmath53 $ ] .
it then turns out that the finite infinity geometry may be rewritten ^2=
-^2+^2 ^ 2[wgeom]where @xmath54 .
the geometry ( [ wgeom ] ) , which now again has a more general spherically symmetric form , is effectively the closest match to the flrw geometry we usually attempt to fit to the whole universe with the assumption that spatial curvature everywhere matches our own , and clocks everywhere are synchronized to our own wall time , @xmath55 . in place of the bare cosmological parameters ( [ eomq ] )
, one can define conventional dressed parameters with respect to the geometry ( [ wgeom ] ) , relevant to `` wall observers '' such as ourselves .
in particular , the dressed density parameter , @xmath56 , defined according to @xmath57 is effectively the conventional parameter , whose numerical value will be similar to that which we infer in flrw models . since ( [ wgeom ] ) is not an flrw metric it does not make particular sense to supplement @xmath56 by additional dressed parameters , as they will not sum to unity . what is more significant cosmologically is that the dressed geometry ( [ wgeom ] ) will yield a dressed hubble parameter , = -=-^-1 .
[ 42 ] where @xmath58 , which differs from the bare hubble parameter , @xmath48 .
similarly , a wall observer will determine a dressed deceleration parameter @xmath59 , which differs from the bare deceleration parameter @xmath60 .
the general solution of the two scale buchert equations was found in ref .
the general solution is specified by four independent parameters , @xmath61 , @xmath62 , @xmath63 and @xmath64 , which are respectively the bare hubble constant , the bare matter density , the present epoch void volume fraction , and a small parameter @xmath65 related to the initial void fraction .
however , two of the four parameters are greatly restricted by taking priors @xcite at the surface of last scattering consistent with the cmb .
furthermore , there is a tracker solution to which all solutions with these priors approach to within 1% by a redshift @xmath66 .
this effectively reduces the number of free parameters to two ; we take these to be @xmath61 and @xmath63 .
the tracker solution is given by & & = 0(3t)^2/32+^1/3 + & & = 3t3t+(1-)(2 + ) , in terms of the volume average cosmic time , @xmath16 .
since the lapse function and bare hubble parameter are related to the void fraction by @xmath67 , the relation @xmath68 may be readily integrated to obtain the wall time , @xmath55 , relevant to observers in galaxies , = 23 + 2(1-)(2+)270(1 + 90 t 2(1-)(2 + ) ) . as is observed in ref .
[ ] , expressions for many relevant observable quantities , including effective dressed luminosity and angular diameter distances , @xmath69 and @xmath70 , are readily obtained .
it is particularly interesting to compare the bare volume
average tracker solution deceleration parameter , = 2 ^ 2(2+)^2,with the corresponding dressed deceleration parameter q=-(8 ^ 3 + 39
^ 2 - 12 - 8)(4++4 ^ 2)^2 , [ qtrack]at early times , when @xmath71 , both deceleration parameters begin at the einstein
de sitter value , @xmath72 .
the bare deceleration parameter always remains positive , meaning that a volume average observer in freely expanding space detects no cosmic acceleration , in accord with the intuition of the critics@xcite .
however , the dressed deceleration parameter changes sign at epoch when @xmath73 , at a zero of the cubic in ( [ qtrack ] ) , so that a wall observer _ does _ detect apparent acceleration . the fractal bubble ( fb ) model provides a solution to the observational coincidence that the onset of cosmic acceleration appears to roughly coincide with the growth of large structures .
cosmic acceleration is an apparent effect that arises from gravitational energy gradients related to spatial curvature gradients .
it begins when the void fraction reaches 59% at a redshift @xmath74 .
this statement itself should be able to be tested in future .
traditionally observational cosmologists define voids in terms of a density contrast threshold , when quoting void volume fractions@xcite . for example
, one might argue as to whether one should take @xmath75 or some other bound . in the present model , the void volume fraction is related technically to those regions within the past light cone which are not within finite infinity . with some further refinement , it should be possible to relate this to an observational definition of void fractions .
then as surveys improve , the claims about the properties of @xmath76 can be compared with observation .
this illustrates the power of the fb model as compared to the spatially flat model .
both models depend on the hubble parameter and one other free parameter .
however , @xmath77 is not directly observable , whereas the void volume fraction , @xmath78 , is empirically observable in principle .
in ref . [ ]
we have examined the fit of the fb model to the riess07 gold data set@xcite of sneia .
we find that with 182 data points and two degrees of freedom the best
fit @xmath79 , i.e. , a @xmath80 of approximately @xmath81 per degree of freedom , which is a good fit . while a marginally lower @xmath82 may be obtained for the best
fit flat model ( with @xmath83 , @xmath84 , @xmath85 ) , the difference is not significant . in fact , a bayesian model comparison of the fb model against a flat model with priors @xmath86 , @xmath87 . gives a bayes factor of @xmath88 in favour of the fb model .
statistically , such a small margin is `` not worth more than a bare mention''@xcite or `` inconclusive''@xcite ; that is to say , the two models are statistically indistinguishable . the residual difference @xmath89 , in the standard distance modulus , @xmath90 , of the best
fit fb model from that of a coasting milne universe of the same hubble constant , @xmath91 , is plotted in fig .
[ dmu ] and compared with binned data from the riess07 gold data set .
apparent acceleration occurs for positive residuals in the range , @xmath92 .
( note , however , that the exact range of redshifts corresponding to apparent acceleration also depends on the value of the hubble constant of the milne universe used to compute the residual . ) the equivalent theoretical residual for the spatially flat model with the same values of @xmath93 and @xmath94 is also shown .
the difference in the gradients of the two curves should lead to observable differences if observing programs which are trying to measure the `` equation of state , @xmath95 , @xmath3 , of dark energy '' can achieve sufficient accuracy . in fig .
[ tests ] we address three major independent cosmological tests .
in addition to statistical confidence limits in the @xmath96 parameter space we also consider the angular scale of the sound horizon in the cmb anisotropy spectrum@xcite , and the comoving scale of the baryon acoustic oscillation ( bao ) seen in galaxy clustering@xcite using the relevant dressed geometry ( [ wgeom ] ) .
ideally we should recompute the spectrum of doppler peaks for the fb model .
however , this is a mammoth task , as the standard numerical codes have been written solely for flrw models , and every step has to be carefully reconsidered .
for this reason we first ask whether parameters exist for which the effective angular diameter scale of the sound horizon matches the angular scale of the sound horizon , @xmath97 rad , of the model , as determined by wmap@xcite .
since there is no change to the physics of recombination , but just an overall change to the _ calibration _ of cosmological parameters , this is entirely reasonable . in fig .
[ tests ] we plot parameter ranges which match the @xmath97 rad sound horizon scale to within 2% , 4% and 6% , using the calculation of the sound horizon given in ref .
[ ] , sec . 7.2 .
the 2% contour would roughly correspond to the 2@xmath98 limit if the wmap uncertainties for the model are maintained . as this can only be confirmed by detailed computation of the doppler peaks , the additional levels have been chosen cautiously . in the case of the bao scale , as we do not yet have the resources to analyse the galaxy clustering data directly , we also begin with a simple but effective check .
since the dressed geometry ( [ wgeom ] ) does provide an effective almost flrw metric adapted to our clocks and rods in spatially flat regions , the effective comoving scale in this geometry should match the corresponding observed bao scale of @xmath99mpc . in fig .
[ tests ] we therefore also plot parameter values which match this scale to within 2% , 4% or 6% .
it should be noted that even if sneia are disregarded , the parameters which fit the two independent tests relating to the sound horizon and the bao scale agree with each other , to the accuracy shown , for values of the hubble constant which include the value of sandage @xcite ; but _ not _ for the values of @xmath93 greater than @xmath100 which best fit the wmap data@xcite with the flrw model .
the value @xmath101 determined by the hubble key team of sandage @xcite has been controversial , given the 14% difference from values which best fit the wmap data with the model@xcite .
however , the wmap analysis only constitutes a direct measurement of cmb temperature anisotropies ; the determination of cosmological parameters involves model assumptions . in the fb model ,
the model assumptions are different . in particular , on account of the differences in gravitational energy and local spatial curvature between observers in bound systems , and those at the volume average in freely expanding space , the calibration of quantities related to the cmb has to be carefully reconsidered . observers at the volume average detect a cooler mean cmb temperature@xcite @xmath102 k than the @xmath103 k we measure . accounting for these differences leads to concordance for parameter values which include the hubble constant of sandage @xcite .
@xmath104 1 @xmath105 statistical uncertainties from sneia alone are shown .
+ [ tab1 ] best fit cosmological parameters are shown in table [ tab1 ] . as statistical bounds are not available for the sound horizon and bao tests , we show 1@xmath98 uncertainties from sneia only .
however , given that there are many independent estimates of the dressed matter density , @xmath94 , we expect that the uncertainties quoted can be significantly reduced if such priors are imposed . the angular scale of the sound horizon and the bao tests have been applied assuming a _ volume average _ baryon to photon ratio in the range @xmath106@xmath107 adopted by tytler @xcite prior to the release of wmap1 . with this range it is possible to achieve concordance with lithium abundances , while also better fitting helium abundances .
this may resolve the primordial lithium abundance anomaly@xcite .
with the 2003 wmap1 release@xcite , the baryon to photon ratio was increased to the very upper range of values that had previously been considered , due to its effect on the ratio of the heights of the first two doppler peaks .
this ratio of peak heights is sensitive to the mass ratio of baryons to non baryonic dark matter rather than directly to the baryon to photon ratio as it depends physically on baryon drag in the primordial plasma .
the fit to the doppler peaks required more baryons than the range of tytler @xcite admitted , when calibrated with the flrw model . in the fb calibration , on account of the difference between the bare and dressed density parameters , a bare value of @xmath108 nonetheless corresponds to a conventional dressed value @xmath109 , and an overall mass ratio of baryonic matter to non baryonic dark matter typically of about 1:3 , which is larger than for .
this would certainly indicate sufficient baryon drag to accommodate the ratio of the first two peak heights .
since the release of the boomerang experiment@xcite in 2000 , it has been generally assumed that the angular position of the first doppler peak is a measure of the spatial curvature of the universe .
however , this inference assumes that the spatial curvature is the same everywhere , as is appropriate for the flrw models . in the present model
there are spatial curvature gradients , and we must revisit the calculation from first principles , as outlined in ref .. insofar as the angular scale of the sound horizon is reproduced for the fb calibration , the overall angular scale can not be regarded as a measure of average spatial curvature .
if there is a sizable negative average spatial curvature at the present epoch , then there must be ways of detecting it other than via the measurement of the angular scale of the doppler peaks .
such effects do indeed arise when one considers more subtle measurements associated with the average geodesic deviation of null geodesics .
indeed one prediction is that there should be nontrivial ellipticity in the cmb anisotropies on account of greater geodesic mixing .
this effect is in fact observed@xcite , and is an important anomaly for the standard paradigm , overlooked by the majority of the community to date
. a detailed quantification of the degree of ellipticity expected , and its comparison with the observed signal , will be an important test of the fb model in future .
it has been recently noted that the values of the normalization of the primordial spectrum @xmath110 and matter content @xmath111 implied by wmap3 are barely compatible with the abundances of massive clusters determined from x ray measurements@xcite .
in fact , sneia also best fit the model with higher values of @xmath94 comparable to the the best
fit value for the fb model , namely @xmath112 . in the fb model the best
fit void fraction , @xmath113 , appears to be the quantity that is mimicked by dark energy fraction , @xmath114 , as far as the wmap normalization to the model is concerned .
the flat model constrains @xmath115 . for the fb model
, the corresponding constraint , @xmath116 , gives quite different larger values , consistent with many other observational determinations of the conventional matter density parameter .
structure formation scenarios in standard model have some difficulty in explaining the observed apparent very early formation of galaxies @xcite .
of course , structure formation contains many model
dependent assumptions , so direct measurements of the ages of metal poor stars in old globular clusters are ultimately a better indicator .
the observational bounds are not very tight , because of the large uncertainties involved in particular nuclear processes .
individual ages are generally consistent with the accepted 13.7gyr age of the universe , but are sometimes in tension with it@xcite , given our uncertainties in knowing how early on the first stars could form .
it is interesting to note that the fb model not only adds a billion years to the age of the universe as measured in a galaxy , but that the percentage difference is larger at earlier times .
for concordance @xmath117 gyr at @xmath118 when distant quasars are seen , and @xmath119 gyr at reionization at @xmath120 .
by contrast for the best
fit fb model @xmath121 gyr at @xmath118 and @xmath122 gyr at @xmath120 , making the universe 34% and 54% older than concordance at the respective epochs .
the fact that the expansion age of the universe determined in a galaxy is greater than that of concordance may alleviate but not completely solve various aspects of the age problem .
however , the fact that the age of the universe is 18.6gyr at the volume average is a basic indicator that the whole issue of structure formation , including calibrations , needs to be re - examined from first principles .
recent analysis of sneia data by jha , riess and kirshner @xcite confirms an effect which has been known about for some time , and has been interpreted as our living in a local void @xcite , the `` hubble bubble '' . if one excludes sneia within the hubble bubble at redshifts @xmath123 , then the value of the hubble constant obtained is lower by @xmath124% @xcite .
it is for this reason that sneia at @xmath125 have been excluded from the riess07 gold data set@xcite .
the hubble bubble is very problematic for the standard dark energy paradigm , as it can contribute a systematic error to the estimate of the equation of state parameter , @xmath126 , which is much larger than the statistical uncertainties that precision cosmologists hope to attain@xcite . in the fb model
the hubble bubble is expected as a feature that results from the implicit resolution of the sandage de vaucouleurs paradox .
since the bare hubble parameter characterizes the uniform `` quasilocally measured '' hubble flow , eq .
( [ 42 ] ) also quantifies the apparent variance in the hubble flow below the scale of homogeneity .
the present bare hubble constant , @xmath61 , is lower than the global average , @xmath93 .
it represents the value we as observers in galaxies would obtain for measurements averaged solely within the plane of an ideal local bubble wall , on scales dominated by finite infinity regions . for us in particular
, measurements of the hubble constant towards the virgo cluster would represent a scale over which we might hope to detect the bare hubble constant , @xmath127 , as this appears to be the direction that most closely represents a local bubble wall .
ultralocal minivoids@xcite complicate any empirical measurement . `
local ' measurements across single voids of the dominant @xmath128 diameter@xcite , probe the scale over which photons on null geodesics encounter the fewest finite infinity domains .
such measurements should give a hubble ` constant ' which exceeds the global average @xmath93 by an amount commensurate to @xmath129 .
as voids are dominant by volume , an isotropic average will generally produce a hubble ` constant ' _ greater than _
@xmath93 for local averages until we sample on large enough scales that the volume average of walls and voids is the same as the global one .
thus the average will steadily decrease from its maximum at @xmath130 mpc until the scale of homogeneity ( @xmath131 ) is reached : the hubble bubble feature . this expectation
is dramatically confirmed by the data points in fig .
[ bubble ] , reproduced from a recent paper of li and schwarz@xcite . with @xmath132 , the maximum hubble ` constant ' , up to about 20% higher than average , should be attained at a scale of about 48mpc , thereafter steadily decreasing until leveling out at the scale of homogeneity , which must be reached before the bao scale of about 168 mpc .
[ bubble ] also illustrates how the resolution of the sandage - de vaucouleurs paradox gives unique predictions .
the two thick lines in fig.[bubble ] show li and schwarz s estimates of the bounds between which the data should lie using buchert averaging , but not accounting for the clock effect .
given that the data indicates a consistently higher hubble constant below the scale of homogeneity , the only explanation for the lack of great statistical scatter between the two lines
if all clocks are effectively synchronous is that by a statistical fluke we happen to be in a large local void which is expanding faster into the surrounding medium@xcite .
however , if the conventional flrw assumptions apply beyond this hubble bubble , then by the standard structure formation scenarios such a circumstance seems impossible . in the fb model
the hubble bubble is not a statistical fluke , but like cosmic acceleration is an apparent effect that arises from clock rate variance .
distant observers in galaxies beyond our own hubble bubble will also detect a hubble bubble centred on their location .
the theoretical derivation of the curve that should pass though the data points in fig .
[ bubble ] will provide an important test of the fb model . to determine this curve , and its variance , requires some knowledge of void statistics , and probably some monte
carlo simulations of the manner in which the scale of homogeneity is filled on average .
particular anisotropies which look like `` large '' voids@xcite might be expected from the variance in alignments of dominant @xmath133 diameter voids below the scale of homogeneity . a systematic error of 12% in the cmb dipole subtraction@xcite
may well be the consequence of a rees sciama dipole resulting from such foreground inhomogeneities@xcite .
while the decades long debate among astronomers about the value of the hubble constant has been mainly dominated by arguments about systematics@xcite , the question of the scale of averaging has no doubt also played a role .
lower values of @xmath93 will be expected if one specifically selects directions and scales which approximate our own bubble wall@xcite .
however , distance determinations on scales @xmath134 mpc will generally give higher values of @xmath93 . since many of the first steps on the cosmic distance ladder are calibrated on such scales , many related issues require careful reconsideration .
the fb model provides a strong new competitor to the standard cosmology , as table [ concord ] shows .
indeed , since it steps out of a paradigm in which everything rests on a single equation of state , it will provide new and unique predictions .
current observational programs of course focus on the measurement of @xmath3 .
the difference in expectations of the fb model for those programs will be highlighted in a forthcoming paper@xcite .
@xmath104 square brackets indicate cases where the present indication is suggestive , but much detailed calculation remains to be done .
+ [ concord ] the quantities that remain to be calculated in table [ concord ] relate largely to the cmb .
detailed construction of the doppler peaks , to enable comparison to wmap , and determination of parameters such as @xmath135 , is a matter of urgency .
it is a rather non
trivial exercise in recalibration of standard quantities , in which all steps need to be carefully reconsidered .
detailed quantitative calculations of cmb ellipticity and the integrated sachs
wolfe effect can only be performed in conjunction with such an analysis . while the calculation of the integrated sachs wolfe ( isw ) effect will differ in the fb model
, one must be careful not to base intuition on that of perturbations on flrw backgrounds .
in particular , the observed isw signal is believed to confirm dark energy since its consequence is a suppression of the gravitational collapse of matter at relatively recent times@xcite . if one replaces the words `` dark energy '' by `` voids '' in the standard qualitative explanation , then a probable description of the isw effect in the fb model emerges .
the same correlation of the isw signal with clumped structures@xcite is expected ; what is important is the magnitude of the effect .
mattsson@xcite has recently proposed an extension of the dyer roeder formalism , which may have some relevance for such calculations . in offering a new paradigm for cosmology ,
every observational test naturally has to be revisited from first principles .
thus there are potential tests which relate to both strong and weak gravitational lensing , for example , which have not been listed in table [ concord ] .
qualitatively one expects voids to be emptier in the fb model than in lcdm structure formation simulations .
however , the specification of the background average surfaces of homogeneity , in terms of a uniform bare hubble expansion gauge , and a correct post newtonian approximation on such a background , need to be examined before structure formation simulations are attempted . ultimately , apparent variance in the hubble flow below the scale of homogeneity will give predictions which can not be reproduced in the flrw scenario .
not only should we determine the average curve that passes through the data points in fig .
[ bubble ] , we should ultimately collect thousands of sneia measurements or other distance measurements on scales up to 200mpc , and test the correlation of the apparent hubble flow with actual structure : a sort of hubble flow tomography of the nearby universe .
a true `` concordance cosmology '' should agree with all reliable observations , and not just a carefully selected subset .
a glance at table [ concord ] reveals that there are many anomalies in the standard model .
much attention has been focused on the cmb anisotropies , on account of the spectacular success of the wmap mission@xcite .
however , it must be recalled that many ingredients go into the analysis of the cmb temperature fluctuations . in looking at these ingredients
we must ask _ `` what is the weakest link '' _ ? in my view , the weakest link is not primordial nucleosynthesis@xcite , which is based on nuclear physics that we understand very well , but the cosmological model .
the weakest link comes from abandoning the theoretical principles that we are careful to apply in other circumstances .
in particular , in general relativity one should model the universe with the matter distribution one observes , rather than trying to impose onto the universe a simple mathematical solution based on some simplifying symmetry .
it is unfortunate that general relativists have been obsessed by exact solutions of einstein s equations , whether they involve likely or unlikely approximations for the matter distribution .
we should face up to the fact that the solution for the actual matter distribution is analytically intractable , and therefore the question of cosmological averaging@xcite is paramount .
furthermore , once we do take an average we must address the fundamental problem that the relationship of rods and clocks at one point to those at a distant point , a conceptual centrepiece of general relativity , is highly non trivial once gradients in spatial curvature and gravitational energy are considered . in an expanding universe
these involve subtle dynamical aspects of general relativity , which can not be localized at a point on account of the equivalence principle .
some colleagues when presented with table [ concord ] , and the fact that the right hand column is based entirely on general relativity with no new or exotic physics ( beyond a need for non baryonic dark matter ) , suggest i should invoke ockham s razor at this point and declare the cosmological constant dead .
other colleagues , who sometimes confess to newtonian intuition , are of the view that a clock rate difference of 38% accumulated between bound systems and the volume average over the lifetime of the universe is nonetheless so great that i must be mistaken somewhere , in spite of table [ concord ]
. in my view caution should always be exercised , but this includes caution with the conceptual basis of our theory and the operational interpretation of measurements . to those who are uncomfortable with my proposal about cosmological quasilocal gravitational energy
let me ask the following : without reference to an asymptotically flat static reference scale , which does not exist given the universe is expanding , and without reference to a background which evolves by the friedmann equation at some level@xcite , an assumption which is manifestly violated by the observed inhomogeneities , _ what keeps clocks synchronized in cosmic evolution _ ?
please explain . in the fb proposal
the fact that our flrw approximation has served so well is understood as a consequence of the fact that the bare hubble flow is uniform , despite large scale inhomogeneities .
uniformity of the quasilocally measured hubble flow does not imply uniformity of spatial curvature , nor of gravitational energy , however .
if we examine the hubble flow over scales on which the average gradients are not large , we will not see large statistical scatter in the hubble flow .
but averaged over larger scales , below the scale of homogeneity , we will see a variance in the apparent hubble flow , that agrees with observations of the statistical properties of voids@xcite , and which seems to accord with fig . [ bubble ] .
the fact that there is a statistically dominant void scale@xcite may well be associated with the evolution of the scale corresponding to the second ( rarefaction ) doppler peak in the cmb anisotropy spectrum , just as evolution of the first ( compression ) peak appears to statistically define the cutoff for the scale of homogeneity . any proposal which seeks to genuinely shift the paradigm against accepted intuition , even a conservative proposal based on a principled approach@xcite to our best theory of gravitation , naturally faces much initial scepticism .
nonetheless , i suspect that in future cosmologists will wonder how we ever could have expected that two decades of precision measurements would simply confirm the nave cosmological models constructed in the 1920s , long before we knew what the universe actually looked like .
i suspect that the coming decades will see us explore and finally understand the arcane territory of general relativity relating to quasilocal gravitational energy@xcite , informed directly by new cosmological observations .
* acknowledgements * i am grateful to many people for discussions and correspondence including , in particular , thomas buchert .
i also thank anthony fairall and dominik schwarz for granting me permission to reproduce their figures [ fig1 ] and [ bubble ] respectively .
66 r. durrer and r. maartens , arxiv:0711.0077 .
peebles , , 495 ( 2001 ) . d.l .
wiltshire , new j. phys .
* 9 * , 377 ( 2007 ) .
wiltshire , , 251101 ( 2007 ) .
t. buchert , , 105 ( 2000 ) ; , 1381 ( 2001 ) .
leith , s.c.c .
ng and d.l .
wiltshire , , l91 ( 2008 ) .
s. rsnen , jcap * 11 * , 003 ( 2006 ) ; m.n .
clrier , new adv .
* 1 * ( 2007 ) 29 [ astro - ph/0702416 ] ; n. li and d.j .
schwarz , , 083011 ( 2007 ) ; t. buchert , arxiv:0707.2153 ; and references therein .
bennett _ et al .
_ , astrophys .
j. suppl .
* 148 * , 1 ( 2003 ) .
spergel _ et al .
_ , astrophys .
j. suppl .
* 170 * , 377 ( 2007 ) .
f. hoyle and m.s .
vogeley , , 641 ( 2002 ) ; , 751 ( 2004 ) .
a.v . tikhonov and i.d .
karachentsev , , 969 ( 2006 ) .
l. rudnick , s. brown and l.r .
williams , arxiv:0704.0908 .
ellis , in b. bertotti , f. de felice and a. pascolini ( eds ) , _ general relativity and gravitation _ , ( reidel , dordrecht , 1984 ) pp .
215288 .
ellis and w. stoeger , , 1697 ( 1987 ) .
f. sylos labini , m. montuori and l. pietronero , phys . rep .
* 293 * , 61 ( 1998 ) .
y. baryshev , aip conf .
proc . * 822 * , 23 ( 2006 ) .
a.d . chernin , i.d .
karachentsev , m.j .
valtonen , v.p .
dolgachev , l.m .
domozhilova and d.i .
makarov , , 19 ( 2004 ) .
a. sandage , g.a . tammann and e.
hardy , , 253 ( 1972 ) .
g. de vaucouleurs , science * 167 * , 1203 ( 1970 ) .
y. hoffman , l.a .
martinez - vaquero , g. yepes and s. gottlber , arxiv:0711.4989 .
m. axenides and l. perivolaropoulos , , 127301 ( 2002 ) .
whiting , , 1996 ( 2006 ) .
kolb , s. matarrese , a. notari and a. riotto , hep - th/0503117 .
kolb , s. matarrese and a. riotto , new j. phys .
* 8 * , 322 ( 2006 ) .
clrier , , 63 ( 2000 ) ; j.w .
moffat , jcap * 05 * , 001 ( 2006 ) ; h. alnes , m. amarzguioui and o. gron , , 083519 ( 2006 ) ; t. biswas , r. mansouri and a. notari , astro - ph/0606703 .
zalaletdinov , , 1015 ( 1992 ) ; * 25 * , 673 ( 1993 ) .
coley , n. pelavas and r.m .
zalaletdinov , phys .
* 95 * , 151102 ( 2005 ) ; a.a .
coley and n. pelavas , , 087301 ( 2006 ) ; * d 75 * , 043506 ( 2007 ) .
a. paranjape and t.p .
singh , , 044006 ( 2007 ) .
a. ishibashi and r.m .
wald , , 235 ( 2006 ) .
j. katz , , 5169 ( 2005 ) ; j. katz , d. lynden - bell and j. bik , , 7111 ( 2006 ) ; d. lynden - bell , j. katz and j. bik , , 024040 ( 2007 ) ; err .
* d 75 * , 044901 ( 2007 ) .
szabados , living rev .
* 7 * , 4 ( 2004 ) . h. bondi ,
_ `` cosmology '' _ , ( cambridge univ .
press , 1960 ) c.m .
chen , j.l .
liu and j.m .
nester , mod . , 2039 ( 2007 ) .
j. garecki , arxiv:0708.2783 .
a. einstein and e.g. straus , rev .
* 17 * , 120 ( 1945 ) ; err .
ibid . * 18 * , 148 ( 1946 ) .
bardeen , , 1882 ( 1980 ) .
t. buchert and m. carfora , , 6109 ( 2002 ) .
t. buchert and m. carfora , , 031101 ( 2003 ) .
riess , , 98 .
kass and a.e .
raftery , j. am .
statist .
* 90 * , 773 ( 1995 ) .
r. trotta , , 72 ( 2007 ) .
eisenstein , ( 2005 ) 560 ; s. cole , ( 2005 ) 505 .
d. tytler , j.m .
omeara , n. suzuki , and d. lubin , phys .
scripta * t85 * , 12 ( 2000 ) .
de bernardis [ boomerang collaboration ] , nature * 404 * ( 2000 ) 955 .
gurzadyan , c.l .
bianco , a.l .
kashin , h. kuloghlian and g. yegorian , ( 2007 ) 121 ; v.g .
gurzadyan , p. de bernardis , g. de troia , c.l .
bianco , a.l .
kashin , h. kuloghlian , s. masi , f. piacentini , g. polenta and g. yegorian , mod .
( 2005 ) 813 .
a. sandage , g.a .
tammann , a. saha , b. reindl , f.d .
macchetto and n. panagia , ( 2006 ) 843 .
g. steigman , int . j. mod
. phys .
* e 15 * ( 2006 ) 1 .
m. asplund , d.l .
lambert , p.e .
nissen , f. primas and v.v .
smith , ( 2006 ) 229 .
g. yepes , r. sevilla , s. gottloeber , and j. silk , , l61 ( 2007 ) .
k. glazebrook , nature * 430 * , 181 ( 2004 ) ; a. cimatti , nature * 430 * , 184 ( 2004 ) .
l.m . krauss and b. chaboyer ,
science * 299 * , 65 ( 2003 ) .
s. jha , a.g . riess and r.p .
kirshner , , 122 ( 2007 ) .
k. tomita , , 287 ( 2001 ) ; prog .
* 106 * , 929 ( 2001 ) .
wood - vasey _ et al .
_ , , 694 ( 2007 ) .
n. li and d.j .
schwarz , arxiv:0710.5073 . w.l .
freedman , , 47 ( 2001 ) .
f. van leeuwen , m. w. feast , p. a. whitelock and c. d. laney , , 723 ( 2007 ) .
m. federspiel , g.a .
tammann and a. sandage , , 115 ( 1998 ) .
freeman , c.r .
genovese , c.j .
miller , r.c .
nichol and l. wasserman , ( 2006 ) 1 .
a. rakic , s. rsnen and d.j .
schwarz , ( 2006 ) l27 .
r. lieu , j.p.d .
mittaz and s .- n .
zhang , , 176 ( 2006 ) .
bielby and t. shanks , , 1196 ( 2007 ) . s. boughn and r. crittenden , nature * 427 * , 45 ( 2004 ) .
t. mattsson , arxiv:0711.4264 . | an overview is presented of a recently proposed `` radically conservative '' solution to the problem of dark energy in cosmology .
the proposal yields a model universe which appears to be quantitatively viable , in terms of its fit to supernovae luminosity distances , the angular scale of the sound horizon in the cosmic microwave background ( cmb ) anisotropy spectrum , and the baryon acoustic oscillation scale .
it may simultaneously resolve key anomalies relating to primordial lithium abundances , cmb ellipticity , the expansion age of the universe and the hubble bubble feature .
the model uses only general relativity , and matter obeying the strong energy condition , but revisits operational issues in interpreting average measurements in our presently inhomogeneous universe , from first principles .
the present overview examines both the foundational issues concerning the definition of gravitational energy in a dynamically expanding space , the quantitative predictions of the new model and its best fit cosmological parameters , and the prospects for an era of new observational tests in cosmology .
= cmr5=cmr7=cmmi7 # 1 _ # 1|#1 # 1 # 1#2#3#1#3#1#2 # 1#2#1#1#2 # 1 _ # 1 _ # 1 # 1#2 # 1 # 1astrophys .
j. * # 1 * # 1phys .
lett . *
# 1 * # 1astron . j. * # 1 * # 1class .
quantum grav .
* # 1 * # 1gen . relativ .
grav . * # 1 * # 1astron .
astrophys . * # 1 * # 1phys .
rev .
lett . *
# 1 * # 1phys .
rev .
* # 1 * # 1mon . not .
r. astr .
soc . * # 1 * |
in this paper we consider a time - periodic ode of the form @xmath0 , where @xmath1 for all @xmath2 with a fixed period @xmath3 , and study the basin of attraction of a periodic solution .
the basin of attraction can be computed using a variety of methods : * invariant manifolds * form the boundaries of basins of attraction , and their computation can thus be used to find a basin of attraction @xcite .
however , this method needs additional arguments to ensure that a certain region is the basin of attraction of an attractor , and that , for example , there are no other attractors in that region . other approaches to compute the basin of attraction are for example the * cell mapping approach * @xcite or * set oriented methods * @xcite which divide the phase space into cells and compute the dynamics between these cells . * lyapunov functions * @xcite are a natural way of analysing the basin of attraction , since they start from the attractive solution , not from the boundary .
moreover , through their level sets , they give additional information about the basin of attraction than just the boundary . *
converse theorems * which guarantee the existence of a lyapunov function under certain conditions have been given by many authors , for an overview see @xcite .
however , all converse theorems offer no general method to analytically construct lyapunov functions . recently
, several methods to construct lyapunov functions have been proposed : hafstein @xcite constructed a piecewise affine lyapunov function using * linear programming*. parrilo @xcite and papachristodoulou and prajna in @xcite consider the numerical construction of lyapunov functions that are presentable as sum of squares of polynomials ( * sos * ) for autonomous polynomial systems .
these ideas have been taken further by recent publications of peet @xcite and peet and papachristodoulou @xcite , where the existence of a polynomial lyapunov function on bounded regions for exponentially stable systems in proven . a different method deals with * zubov s equation * and computes a solution of this partial differential equation ( pde ) . in camilli
@xcite , zubov s method was extended to control problems .
giesl considered a particular lyapunov function satisfying a linear pde which was solved using * meshless collocation * , in particular radial basis functions @xcite .
this method has been extended to time - periodic odes @xcite .
lyapunov functions attain their minimum on the attractor and have a negative orbital derivative for all points in the basin of attraction apart from the attractor .
hence , it is necessary to have * exact information * about the attractor in the phase space before one can compute a lyapunov function . whereas this information might be easy to obtain in special examples , in general this information is not available . *
local contraction property
borg s criterion * another method to characterise the basin of attraction , introduced by borg @xcite , uses a local contraction property and does * not * require information about the periodic orbit .
let @xmath4 be a riemannian metric , i.e. @xmath5 such that @xmath4 is a positive definite , symmetric @xmath6 matrix for all @xmath7 .
then @xmath8 defines a point - dependent scalar product , where @xmath9 .
the sign of the real - valued function @xmath10 , cf .
( [ ldefm1 ] ) , then describes whether the solution through @xmath7 and adjacent solutions approach each other with respect to the riemannian metric @xmath11 .
we define @xmath12w,\label{ldefm1}\end{aligned}\ ] ] where @xmath13 denotes the orbital derivative of @xmath4 , which is the derivative along solutions of @xmath0 .
if @xmath10 is negative for all @xmath14 where @xmath15 is a positively invariant , connected set , then @xmath15 is a subset of the basin of attraction of a unique periodic orbit in @xmath15 . the maximum in ( [ ldefm1 ] ) is taken over all @xmath16 with a norm condition , and @xmath17 is equivalent to @xmath18 given by ( [ cont ] ) , where @xmath19 denotes the maximal eigenvalue of a symmetric matrix . here
, we use that @xmath20 .
@xmath21w\nonumber\\ & = & \lambda_{max}\left(m(t , x)d_xf(t , x ) + d_xf(t , x)^tm(t , x ) + m'(t , x)\right).\label{cont}\end{aligned}\ ] ] we seek to find a matrix - valued function @xmath11 satisfying @xmath18 .
this is equivalent to the condition that the symmetric matrix @xmath22 is negative definite .
as this is a linear matrix inequality , it can be formulated as a constraint of a semidefinite optimization problem . while the sufficiency of this local contraction criterion in the autonomous case goes back to @xcite ,
its necessity was shown in @xcite .
the method was extended to time - periodic systems @xcite .
the advantage of this method over , for example , lyapunov functions , is that it does not require information about the position of the periodic orbit .
moreover , the criterion is robust to small errors .
although the existence of riemannian metrics has been shown @xcite , it remains a difficult problem to construct them for concrete examples .
this is a similar problem to the construction of a ( scalar - valued ) lyapunov function , but borg s criterion requires the construction of a matrix - valued function @xmath4 . in the two - dimensional autonomous case
, however , there exists a special riemannian metric of the form @xmath23 , where @xmath24 is a scalar - valued function @xcite .
this can be used to find an approximation using radial basis functions @xcite . in higher dimensions , however , the existence of such a special riemannian metric is not true in general @xcite . in @xcite , a combination of a riemannian
metric locally near the periodic orbit with a lyapunov function further away was used , and the construction was again achieved by radial basis functions .
this method , however , heavily depends on information about the periodic orbit , which was obtained by a numerical approximation of the periodic orbit and its variational equation . in this paper
, we will develop a new method to construct a riemannian metric to fulfill borg s criterion , which will use semidefinite optimization and does not require any information about the periodic orbit .
* semidefinite optimization * a semidefinite optimization problem for the variables @xmath25 is of the form @xmath26where @xmath27 are symmetric @xmath28 matrices and @xmath29 means that the matrix @xmath30 is positive semidefinite .
the goal of this paper is to formulate the condition of a contraction metric as a semidefinite optimization ( feasibility ) problem . in a subsequent paper
we will discuss the details of how to solve this problem efficiently .
the main idea is to first triangulate the phase space .
the riemannian metric , i.e. the symmetric matrix @xmath4 , will be expressed as a continuous piecewise affine ( cpa ) function , i.e. if @xmath11 is given at the vertices @xmath31 of a simplex , then @xmath32 , where @xmath33 .
the conditions of borg s criterion will become the constraints of a semidefinite optimization problem on the vertices of the triangulation , which will ensure the contraction property for * all * points in the simplices . in @xcite , a contraction metric is also constructed using semidefinite optimization .
there are , however , three main differences to our approach : firstly , adjacent trajectories in * all * directions are contracted , whereas in our case the contraction takes place in the @xmath34-dimensional subspace @xmath35 of @xmath36 , but not in the time - direction . thus , in our case , the attractor is a periodic orbit , whereas in their case , it is an equilibrium point .
secondly , and more importantly , the above paper transforms the construction problem to a linear matrix inequality and solves this using a sum - of - squares approach .
this approach is used to prove global stability , i.e. the basin of attraction is the whole space .
the contraction metric is a polynomial function , and the system considered is assumed to be polynomial , too . in this paper , we study systems which are not necessarily polynomial nor globally stable , and we triangulate the phase space to obtain a large subset of the basin of attraction .
lastly , we are able to prove that the semidefinite optimization problem is feasible if and only if the dynamical system has an exponentially stable periodic orbit , whereas in their paper the equivalence does not hold since the sum - of - squares condition is more restrictive than positive definiteness of matrices .
the paper is organised as follows : in section [ sec0 ] we start with preliminaries , and in section [ sec1 ] we generalise the existing theorem for a smooth riemannian metric @xmath11 to a cpa ( continuous piecewise affine ) riemannian contraction metric .
we show that the existence of such a metric is sufficient to prove existence and uniqueness of a periodic orbit and to determine a subset of its basin of attraction . in section [ sec2 ]
, we describe the triangulation of the phase space into a simplicial complex and state the semidefinite optimization problem .
furthermore , we show that the feasibility of the semidefinite optimization problem provides us with a cpa contraction metric .
we also discuss a possible objective function to obtain a bound on the largest floquet exponent . in section [ sec3 ] , finally
, we show that the semidefinite problem is feasible , if the dynamical system has an exponentially stable periodic orbit .
altogether , this paper shows that the problem of finding a contraction metric is equivalent to the feasibility of a semidefinite optimization problem .
in this paper we consider a time - periodic ode of the form @xmath37 where @xmath1 for all @xmath2 with a fixed period @xmath3 . we denote @xmath38 and @xmath39 .
we study the equation on the cylinder @xmath40 as the phase space , where @xmath41 denotes the circle of circumference @xmath42 .
we assume that @xmath43 .
if even @xmath44 holds , then we obtain improved results , in particular higher order approximations ; this is , however , not necessary to derive the main result .
we denote the ( unique ) solution of the ode with initial value @xmath45 by @xmath46 and denote @xmath47 .
furthermore , we assume that the solution exists for all @xmath48 .
we use the usual notations for the vector and matrix norms , in particular we denote by @xmath49 the maximal entry of a matrix .
@xmath50 denotes the set of all symmetric real - valued @xmath51 matrices . for a symmetric matrix @xmath52
, @xmath53 denotes its maximal eigenvalue , we write @xmath54 if and only if @xmath11 is positive semidefinite and @xmath55 if and only if @xmath11 is negative semidefinite .
the convex hull is defined by @xmath56
it was shown in @xcite that a smooth contraction metric implies the existence and uniqueness of a periodic orbit and gives information about its basin of attraction . in this paper
, we will seek to construct a cpa contraction metric , which is not of the smoothness required in the above paper . in this section
we show that we can relax the conditions on the smoothness of @xmath11 to cover the case of a cpa contraction metric .
we will require that @xmath11 is continuous , lipschitz continuous with respect to @xmath57 , and the forward orbital derivative exists .
we will later relate this to the construction of the cpa contraction metric on a suitable triangulation , and we will prove that such a cpa metric will satisfy all assumption that we make in this section ( lemma [ le4.9 ] ) .
the proof of the main theorem , theorem [ th1 ] , will closely follow @xcite .
first we define a weaker notion of a riemannian metric , which does not assume that @xmath11 is differentiable , but only that the orbital derivative in forward time exists .
[ riem ] @xmath11 is called a riemannian metric for ( [ ode ] ) , if @xmath58 where @xmath4 is positive definite .
moreover , we assume that @xmath11 is locally lipschitz - continuous with respect to @xmath57 , i.e. for all @xmath59 there exists a neighborhood @xmath60 such that @xmath61 holds for all @xmath62 .
furthermore , we assume that the forward orbital derivative @xmath63 is defined for all @xmath64 , where @xmath63 denotes the matrix @xmath65 we have the following statements for functions with a right - hand side derivative .
[ product_rule ] let @xmath66 be rhs differentiable , i.e. @xmath67 exists .
let @xmath68 be differentiable .
then @xmath69 and @xmath70 are rhs differentiable and @xmath71 let @xmath72 , where @xmath73 is open , be rhs differentiable .
let @xmath74\subset i$ ] , and let @xmath75 exist and be finite .
then we have @xmath76 the first two statements follow directly from the usual proofs , replacing the limit with the rhs limit .
the last statement is a result due to lebesgue , formulated originally for dini derivatives .
[ ableitung ] let @xmath77 be a function which is locally lipschitz with respect to @xmath57 and let @xmath78 exist .
then the orbital derivative @xmath79 exists and is equal to @xmath80 recall that @xmath38 and @xmath39 .
we use that @xmath81 note that due to the lipschitz continuity @xmath82 we have for small @xmath83 @xmath84 by the mean value theorem there are @xmath85 $ ] such that @xmath86 hence , @xmath87 since the solution and @xmath88 are both continuous .
altogether , we thus have @xmath89 [ th1 ] consider the equation @xmath0 , where @xmath90 , and assume that @xmath91 and all partial derivatives of @xmath92 order one with respect to @xmath57 are continuous functions of @xmath7 .
let @xmath93 be a connected , compact and positively invariant set .
let @xmath11 be a riemannian metric in the sense of definition [ riem ] .
moreover , assume @xmath94 for all @xmath95 , where @xmath96w\label{lm}\end{aligned}\ ] ] then there exists one and only one periodic orbit @xmath97 which is exponentially asymptotically stable
. moreover , for its basin of attraction @xmath98 holds .
if @xmath99 exists and is finite for all solutions @xmath100 with @xmath101 and all functions @xmath102 , \mathbb r^n)$ ] , then the largest real part @xmath103 of all floquet exponents of @xmath104 satisfies @xmath105 note that in contrast to @xcite , here @xmath106 is not necessarily continuous , since @xmath63 is not continuous in general .
the only parts in the proof that need to be changed slightly from @xcite are the proof of proposition 3.1 and the estimate on the floquet exponents ; the other steps of the proof are exactly the same .
* proposition 3.1 * we replace the temporal derivative in step iii of ( * ? ? ?
* proposition 3.1 ) by the _
forward _ temporal derivative of @xmath107^t m(s_\theta(t_0 , x_0))\nonumber\\ & & \hspace{4.5cm}\cdot \left[s^x_{\theta}(t_0,x_0+\eta)-s^x_\theta ( t_0,x_0)\right ] .
\ ] ] note that the product rule holds for the rhs derivative as usual , cf .
lemma [ product_rule ] .
furthermore , we use the comparison lemma in the version for rhs limits ; a more general version for dini derivatives can be found in ( * ? ? ?
* lemma 3.4 ) .
this shows that the result of ( * ? ? ?
* proposition 3.1 ) remains true for the riemannian metric @xmath11 as in definition [ riem ] .
* floquet exponent * in this part of the proof in @xcite we need to show that @xmath108 where @xmath100 is the periodic orbit and @xmath109 with @xmath110 is a solution of the first variation equation @xmath111 along the periodic orbit .
note that we have already used that the rhs derivative satisfies the product rule , cf .
lemma [ product_rule ] . using the same lemma ,
the composition with the differentiable function @xmath112 is also rhs differentiable . to apply the last statement of this lemma
, we need to show that @xmath113 exists and is finite .
indeed , we have , using the chain rule in lemma [ product_rule ] , that @xmath114 since @xmath115 is positive definite and continuous and @xmath116 , the denominator is bounded away from zero .
we apply the product rule to the nominator and use that @xmath99 exists and is finite by assumption , and that the other terms are smooth .
this shows the theorem .
in this section we will state a semidefinite optimization problem and show that a feasible solution of the semidefinite optimization problem defines a contraction metric , satisfying the assumptions of theorem [ th1 ] .
the solution will be a cpa matrix - valued function @xmath11 , which is defined by the values at the vertices .
we will first define a triangulation of the phase space and then define a riemannian metric by its values at the vertices of this triangulation and affine on each simplex .
this cpa riemannian metric is shown to fulfill all conditions of theorem [ th1 ] , also inside the simplices .
for the algorithm to construct a piecewise affine lyapunov function we need to fix our triangulation .
this triangulation is a subdivision of @xmath117 into @xmath118-simplices , such that the intersection of any two different simplices in the subdivision is either empty or a @xmath119-simplex , @xmath120 , and then its vertices are the common vertices of the two different simplices .
such a structure is often referred to as a simplicial @xmath118-complex .
in contrast to @xcite , we do not need to have a fine triangulation near 0 , but we need to ensure that the triangulation respects the periodicity in @xmath121 of the phase space , i.e. it is a triangulation of the cylinder @xmath40 . for the construction we use the standard orthonormal basis @xmath122 of @xmath123 , where @xmath124 denotes the unit vector in @xmath121-direction .
we also fix a scaling matrix @xmath125 with diagonal entries @xmath126 which fixes the ratio of the fineness of the triangulation with respect to different directions .
we have fixed the @xmath127 in @xmath121-direction to ensure that the simplicial complex is compatible with the periodicity .
further , we use the characteristic functions @xmath128 equal to one if @xmath129 and equal to zero if @xmath130 and the functions @xmath131 , defined for every @xmath132 by @xmath133 where @xmath134 and @xmath135 .
thus @xmath136 puts a minus in front of the coordinate @xmath137 of @xmath138 if @xmath139 .
[ neig ] denote by @xmath140 the set of all subsets @xmath141 that fulfill : 1 .
@xmath142 is compact .
2 . the interior @xmath143 of @xmath142 is connected and open .
3 . @xmath144 .
note that compactness , connectedness etc .
refer to the space @xmath145 . of @xmath146 .
note that the operator @xmath147 puts a minus - sign in @xmath57-directions , but not in the @xmath121-direction , which results in the simplices shown in this figure.,height=143 ] [ triconstr ] let @xmath148 be a given subset of @xmath145 .
we will define a triangulation @xmath149 of a @xmath150 , @xmath151 , that approximates @xmath152 . to construct the triangulation of a set @xmath149
, we first define the triangulations @xmath153 and @xmath154 as intermediate steps . 1 . the triangulation @xmath153 , cf .
figure [ fig1 ] , consists of the simplices @xmath155 for all @xmath156 , all @xmath157 , and all @xmath158 , where @xmath159 denotes the set of all permutations of the numbers @xmath160 .
2 . now scale down the triangulation @xmath153 with the mapping @xmath161 , where @xmath162 and @xmath163 is the fixed diagonal matrix defined above .
we denote by @xmath164 the resulting set of @xmath118-simplices , i.e. @xmath165 note that for each simplex @xmath166 there is a unique @xmath167 such that @xmath168\times \mathbb r^n$ ] .
this follows from the fact that the scaling matrix @xmath163 has @xmath127 as its entry in @xmath121-direction and @xmath169 .
hence , we can and will interpret a simplex @xmath170 as a set in @xmath171 in the following step .
3 . as a final step
define @xmath172 and set @xmath173 note that @xmath174 consists of finitely many simplices due to the fact that @xmath152 is compact and the triangulation respects the periodicity .
consider the sets @xmath152 and @xmath175 from the last definition . then @xmath151 and @xmath150 .
@xmath175 is a closed set containing @xmath176 and thus contains @xmath148 because @xmath177 by property iii ) in definition [ neig ] , so @xmath152 is the smallest closed set containing @xmath176 . that @xmath175 fulfilles properties i ) and iii ) of definition [ neig ] follows directly from the fact that @xmath175 is a finite union of @xmath118-simplices
to see that property ii ) of definiton [ neig ] is also fulfilled , i.e. that @xmath178 is connected , notice the following : the definition of @xmath149 implies that for any @xmath179 we have @xmath180 . because of this , any @xmath181 can be connected to a @xmath182 with a line contained in @xmath183 with a possible exception of the endpoint @xmath138 . because @xmath148 we have that @xmath176 is connected so this implies that @xmath178 is connected too , i.e. @xmath175 also fulfilles property ii ) of definition [ neig ] and therefore @xmath150 .
the triangulation @xmath153 is studied in more detail in sections 4.1 and 4.2 in @xcite .
a sometimes more intuitive description of @xmath184 is the simplex @xmath185 translated by @xmath186 and then a minus - sign is put in front of the @xmath187-th entry of the resulting vectors whenever @xmath188 ; but no change of sign in the @xmath121-coordinate .
@xmath149 is truly a triangulation , i.e. two different simplices in @xmath149 intersect in a common face or not at all , as shown in lemma [ simplemma ] . [ simplemma ] consider the set of simplices @xmath149 from definition [ triconstr ] and let @xmath189 and @xmath190 be two of its simplices .
then @xmath191 where @xmath192 are the vertices that are common to @xmath193 and @xmath194 , i.e. @xmath195 for @xmath196 and @xmath197 , @xmath198 .
the equation ( [ simpeq ] ) follows as in theorem 4.11 in @xcite and the fact that the triangulation respects the periodicity .
one important property of the chosen triangulation is that the simplices are sufficiently regular , i.e. the angles all have a lower bound . to make this precise and to also measure the influence of @xmath15 , we prove the following lemma .
the matrix @xmath199 , as defined in the next lemma , consists of the @xmath200 vectors which span the simplex .
we obtain an estimate on its inverse @xmath201 depending on @xmath15 .
this estimate will later be used to estimate the derivative of an affine function on the simplex .
[ xlemma]using the notation of definition [ triconstr ] , there is a constant @xmath202 , which is independent of @xmath15 and @xmath203 , such that for all simplices @xmath204 we have @xmath205 where @xmath206 , @xmath207 are the vertices of @xmath208 ( in any order ) and @xmath209 . every simplex in @xmath210 is formed from an @xmath118-simplex @xmath211 with corresponding matrix @xmath212 .
the matrices @xmath199 and @xmath212 relate via @xmath213 note that for the matrices @xmath212 , up to translations , there are finitely many different simplices in @xmath153 and also finitely many possibilities of ordering the vertices of any such simplex .
hence , there is only a finite number of possibilities of forming such a matrix @xmath212 .
further , all of them are invertible .
this means that we can define @xmath214 as the minimal eigenvalue of all possible @xmath215 .
note that @xmath216 hence , @xmath217 as @xmath169 .
this shows the lemma with @xmath218 .
note especially that @xmath202 is a constant independent of @xmath15 and @xmath203 . in the next lemma
we show that a cpa function on a simplicial complex as defined above satisfies the assumptions of lemma [ ableitung ] and the technical assumption of theorem [ th1 ] .
let @xmath219 be a simplicial complex in @xmath40 , which is locally finite , i.e. each point has a neighborhood @xmath220 such that @xmath221 only for a finite number of simplices @xmath222 .
denote @xmath223 .
let @xmath224 be a cpa function , which is affine on each simplex of @xmath219 , and let @xmath4 be positive definite for all @xmath7 .
then @xmath11 is lipschitz - continuous on @xmath143 and for each @xmath225 @xmath226 exists .
there also exists ( at least ) one simplex @xmath227 and @xmath228 such that @xmath229 .
\label{t } \end{aligned}\ ] ] @xmath230 restricted to this simplex is an affine function and the expression in ( [ eq * ] ) is equal to * @xmath231 , the smooth orbital derivative of the affine function @xmath230 , * and @xmath232 .
these expressions are the same for all simplices which satisfy ( [ t ] ) . in particular
, @xmath11 is a riemannian metric in @xmath143 in the sense of definition [ riem ] .
moreover , @xmath99 exists and is finite for all solutions @xmath100 with @xmath233 for all @xmath234 $ ] and all functions @xmath102 , \mathbb r^n)$ ]
. let @xmath235 .
since there are only finitely many simplices @xmath236 , which have a non - empty intersection with a neighborhood @xmath220 of @xmath237 , and on each of them @xmath238 is affine and has a finite lipschitz constant , the overall constant can be chosen as the maximum of the finitely many .
now we show that there is a @xmath228 and a simplex @xmath227 such that @xmath239 for all @xmath240 $ ] .
then it is clear by the smooth chain rule that the limit ( [ eq * ] ) exists and is equal to the smooth orbital derivative @xmath241 of the function @xmath11 restricted to the simplex @xmath208 .
furthermore , ( [ eq * ] ) is equal to @xmath242 by lemma [ ableitung ] .
if there are two simplices @xmath193 and @xmath194 with property ( [ t ] ) , then @xmath243 and also @xmath244 by lemma [ ableitung ] since @xmath245 for all @xmath246 with @xmath240 $ ] .
now we show that there exists a simplex with property ( [ t ] ) .
indeed , there is a @xmath247 such that @xmath248 for all @xmath249 where @xmath220 is the neighborhood of @xmath237 from above .
assume that there is no simplex @xmath208 and no @xmath228 such that @xmath239 for all @xmath240 $ ] .
then @xmath250 for all simplices @xmath251 to which @xmath237 belongs , since if one such point was in @xmath252 , then , due to the convexity of @xmath252 , the whole line between that point and @xmath237 would be in @xmath208 .
since there are infinitely many points in @xmath253 which are in @xmath220 , but only finitely many simplices that have nonempty intersection with @xmath220 by assumption , at least one of them , say @xmath208 , must contain infinitely many such points . since the sequence @xmath254 converges to @xmath237 as @xmath255 and @xmath208 is closed , @xmath256 and hence @xmath208 must be one of the @xmath257 defined above , which is a contradiction .
this shows the statement .
for the last statement , we show that the function @xmath258 is rhs continuous
. then the function @xmath259 is rhs continuous and bounded , and thus it is integrable .
fix @xmath237 and a neighborhood @xmath220 such that there are only finitely many simplices with nonempty intersection with @xmath220 .
denote by @xmath260 the subset of those finitely many simplices such that for each @xmath261 there is a @xmath262 with @xmath263 for all @xmath264 $ ] , these are the ones the qualify to be @xmath265 .
now take a sequence @xmath266 and we seek to prove that @xmath267 .
since @xmath268 , if @xmath119 is large enough , there are only finitely many simplices which contain infinitely many elements @xmath269 of the sequence .
we show that these simplices are in fact a subset of @xmath260 as defined above , satisfying property ( [ t ] ) .
if this was not true then there would be a simplex @xmath270 and a sequence ( the subsequence from above ) of points @xmath271 , @xmath272 , but @xmath273 for all @xmath274 .
the simplex @xmath270 is the intersection of finitely many half - spaces .
the point @xmath237 lies on the boundary of ( at least ) one of these half - spaces , since otherwise a neighborhood of @xmath237 would lie in @xmath270 .
hence , there is a half - space of the form @xmath275 with @xmath276 such that @xmath277 .
hence there is an @xmath278 such that @xmath279 for all @xmath280 with @xmath281 .
now we have @xmath282 if @xmath119 is large enough , then @xmath283 fulfills @xmath281 by continuity of @xmath284 and thus @xmath285 , which is a contradiction .
this shows that each simplex which contains infinitely many elements of the sequence fulfills property ( [ t ] ) and thus is a candidate for @xmath265 .
let @xmath270 be one of the simplices which contains infinitely many elements of the sequence and define the subsequence @xmath286 by choosing the next @xmath287 with the property @xmath271 .
we do the same for all ( finitely many ) simplices that contain infinitely many elements . for the convergence of the overall sequence
it is enough to show that every subsequence @xmath269 converges to the same limit , since there are only finitely many . for each subsequence ,
the elements are in one simplex and on this simplex @xmath288 is a continuous function , so it converges .
the limit is the same , as @xmath242 is the same no matter which simplex @xmath265 we choose .
the condition locally finite for the triangulation is indispensable as shown by the following example : let @xmath289 and for every @xmath290 define the triangle @xmath291 .
then clearly there is no @xmath292 with a corresponding @xmath228 such that @xmath293 for all @xmath294 $ ] . for each simplex
@xmath295 we denote @xmath296 note that for our triangulation we have with @xmath297 and @xmath298 the estimate @xmath299 moreover , denote @xmath300 * variables * the variables of the semidefinite optimization problem are 1 .
@xmath301 for all @xmath302 and all vertices @xmath303 of all simplices @xmath304 values of the riemannian metric at vertices 2 .
@xmath305 for all simplices @xmath306 bound on @xmath11 in @xmath208 3 .
@xmath307 for all simplices @xmath306 bound on derivative of @xmath11 in @xmath208 thus we have @xmath308 variables , where @xmath309 denotes the number of simplices and @xmath310 the number of vertices . in this section , there is no objective function , so we are considering a feasibility problem see section [ objective ] for a suitable objective function .
the equality constraint 1 . can be incorporated by choosing the variables in this way and thus is no actual constraint .
note that constraint 3 .
is linear and constraints 2 .
, 4 . and 5 .
are semidefinite .
constraints 2 . and 5 .
need to be satisfied for each simplex and then for each vertex , so vertices common to more simplices need to satisfy several constraints . in constraint 4 .
, however , each vertex only needs to be checked once .
note , however , that we can replace the individual constants @xmath311 for each simplex by their maximum @xmath312 , so that also constraint 2 . only needs to be checked at each vertex .
note that if the triangulation is fine enough and the system has an exponentially stable periodic orbit , this more restrictive form of the constraint can be satisfied , cf .
the proof of theorem [ feas ] . 1 .
* periodicity * + @xmath313 for all @xmath302 and for all vertices at times @xmath314 and @xmath42 , i.e. for all vertices @xmath315 and @xmath316 .
note that by construction of the triangulation @xmath315 is a vertex of a simplex in @xmath174 if and only if @xmath316 is .
2 . * bound on @xmath11 * + @xmath317 for all vertices @xmath303 of all simplices @xmath318 , where the symmetric matrix @xmath319 is defined by setting @xmath320 for all @xmath321 .
* bound on derivative of @xmath11 * + @xmath322 for all @xmath323 , @xmath302 and for all simplices @xmath295 , where @xmath324 for all @xmath325 , which is given by @xmath326 where @xmath327 .
* positive definiteness of @xmath11 * + we fix @xmath328 .
@xmath329 for all vertices @xmath330 of all simplices @xmath295 .
* contraction of the metric * + @xmath331 for all simplices @xmath304 and all of its vertices @xmath332 . here
, @xmath333 and @xmath334 denotes the symmetric @xmath6 matrix with entries @xmath335 , where @xmath336 was defined in ( [ defw ] ) and is the same vector for all vertices in one simplex and @xmath337 & \mbox { if } f\in c^2\\ h_\nu^2n [ \sqrt{n+1}(1 + 4n)b_\nu d_\nu + 2n(n+1)b_{3,\nu}c_\nu]&\mbox { if } f\in c^3 \end{array}\right.\end{aligned}\ ] ] in constraint 3 .
we have claimed that the gradient of the affine function @xmath338 , i.e. @xmath339 , is given by the expression in ( [ defw ] ) . for a proof of this fact and , moreover , that the definition is independent of the choice of the vertex @xmath237 , see ( * ? ? ?
* remark 2.9 ) .
the constraints above are easily transferred into the standard form @xmath340 in the following way : denote by @xmath341 the matrix elements @xmath342 , @xmath302 , by the following @xmath343 elements @xmath344 the matrix elements of @xmath345 etc . and
finish the vector @xmath346 by the @xmath311 and @xmath347 .
this results in @xmath348 as above .
now constraints 2 . and 4 .
are expressed in @xmath349 , @xmath310 blocks of size @xmath34 each , respectively , in the matrices @xmath27 .
constraint 3 . is expressed in @xmath350 conditions ( for each @xmath187 , @xmath351 ) for each of the @xmath200 entries of the vector @xmath352 .
this needs to be considered for each simplex , so that we have @xmath353 blocks of size @xmath127 .
finally , constraint 5 .
is expressed in @xmath349 blocks of size @xmath34 , since for every simplex each of its vertices needs to be considered .
note that @xmath354 depends linearly on @xmath311 and @xmath347 and @xmath352 depends linearly on @xmath355 .
the size of the matrices @xmath27 is thus a block - diagonal structure with @xmath353 blocks of size @xmath127 and @xmath356 blocks of size @xmath34 .
[ con_new ] note that constraint 2 .
implies that @xmath357 since @xmath319 is positive definite .
note also that constraint 3 .
implies that @xmath358 . moreover , constraint 5 .
is equivalent to @xmath359 where @xmath360 denotes the maximal eigenvalue .
a solution of the semidefinite optimization problem returns a matrix @xmath355 at each vertex @xmath303 of the triangulation for @xmath302 .
we define the cpa metric by affine interpolation on each simplex .
[ cpa - def ] fix @xmath152 and a triangulation @xmath174 with @xmath361 .
let @xmath355 be defined by a feasible solution of the semidefinite optimization problem .
let @xmath362 such that @xmath363 with @xmath364 $ ] and @xmath365 . then define @xmath366 [ le1 ] the matrix @xmath367 as in definition [ cpa - def ] is symmetric and positive definite for all @xmath368 .
the function @xmath4 is periodic in @xmath121 with period @xmath42 .
the symmetry follows directly from the symmetry of @xmath319 : @xmath369 the positive definiteness also follows from the positive definiteness of @xmath319 , using that the minimal eigenvalue @xmath370 is a concave function .
indeed , consider @xmath325 .
note that @xmath371 for all @xmath332 due to constraint 4 .
thus , @xmath372 the @xmath42-periodicity follows directly from the definition and the triangulation : @xmath373 we will now relate @xmath374 to @xmath375 , as well as @xmath376 to @xmath377 . for the proof we will need the following auxiliary result , see ( * ? ? ?
* proposition 4.1 and corollary 4.3 ) . [ help ]
let @xmath378 and @xmath379 with @xmath363 , @xmath364 $ ] and @xmath365 .
then for all @xmath380 we have @xmath381 especially @xmath382 where @xmath383 , @xmath384 , @xmath385 and @xmath386 denotes the hessian of @xmath387 . in the following
we restrict ourselves to one simplex @xmath388 .
first we need to define @xmath374 ( the orbital derivative ) for a general point in the simplex @xmath208 .
note that for all points of a simplex the vector @xmath389 is the same , but the contribution of @xmath390 is different .
[ basic]let @xmath367 be as in definition [ cpa - def ] .
fix a point @xmath391 and a simplex @xmath304 such that @xmath392 for all @xmath240 $ ] , cf .
lemma [ le4.9 ] .
then @xmath363 with @xmath364 $ ] , @xmath365 and @xmath393 note that @xmath338 is an affine function and its gradient @xmath336 was defined in constraint 3 . and
is the same vector for all points @xmath394 .
note that for vertices @xmath395 this term appears in constraint 5 .
[ 4.12 ] let @xmath396 and let @xmath367 be as in definition [ cpa - def ] . fix a point @xmath391 and a corresponding simplex @xmath397 as in definition [ basic ] .
then we have the following estimates for all @xmath398 @xmath399 where @xmath384 , @xmath385 and @xmath383 .
altogether we have @xmath400 @xmath401 if @xmath402 we obtain in addition the estimates @xmath403 @xmath404 @xmath405 @xmath404 where @xmath406 , @xmath385 .
altogether we have @xmath400 @xmath407=\frac{e_\nu}{n}.\end{aligned}\ ] ] * step 1 : @xmath408 * fix one simplex @xmath306 of the triangulation and let @xmath394 . by definition of @xmath409 as a cpa function ,
interpolating @xmath355 at the vertices , we have @xmath410 for all @xmath394 , where @xmath11 is restricted to the simplex , cf .
definition [ basic ] .
thus , letting @xmath363 with @xmath364 $ ] and @xmath365 @xmath411 hence @xmath412 since @xmath413 as the @xmath121-component is @xmath314 .
now we use lemma [ help ] for the @xmath414 function @xmath92 , establishing that @xmath415 .
thus,@xmath416 using that @xmath417 holds from remark [ con_new ] , we obtain @xmath418 * step 2 : @xmath419 * we consider @xmath420 .
we first consider two scalar - valued functions @xmath421 and @xmath422 , where @xmath423 and @xmath422 is @xmath424 in @xmath425 .
we have , using taylor expansion for @xmath422 at @xmath303 , i.e. @xmath426 , where @xmath427 lies on the straight line between @xmath303 and @xmath425 , @xmath428\end{aligned}\ ] ] @xmath429 where @xmath430 . applying this to @xmath431 and @xmath432 we obtain with @xmath433 from remark [ con_new ] and @xmath434 for all @xmath435 , using that @xmath436 and @xmath208 is a convex set .
thus , @xmath437 hence , @xmath438 a similar estimate holds for @xmath439 . *
step 2 : @xmath419 * if @xmath402 , then we can derive an estimate for this term which establishes order @xmath440 .
we consider two scalar - valued functions @xmath421 and @xmath422 , where @xmath441 for @xmath363 and @xmath442 is @xmath414 in @xmath425 .
we apply lemma [ help ] to @xmath443 , yielding @xmath444 where the matrix @xmath445 is defined by @xmath446 .
note that @xmath447 and @xmath448 applying this to @xmath431 , we observe that , since @xmath449 is affine on the simplex , @xmath450 and @xmath451 for all @xmath394 . hence , @xmath452 using @xmath432 , we obtain with @xmath453 and @xmath454 , @xmath455 , defining @xmath456 @xmath457 thus , using @xmath458 , as well as @xmath459 , @xmath460 and @xmath461 we obtain @xmath462 using remark [ con_new ] . hence , ( [ help3 ] ) establishes @xmath463 which proves the lemma .
now we can estimate the value of @xmath464 for all points @xmath425 for the cpa metric @xmath11 , given our constraints on the vertices .
[ le2 ] let all constraints be satisfied
. then the cpa metric @xmath11 defined in definition [ cpa - def ] fulfills:@xmath465 for all @xmath466 .
the maximal eigenvalue is a convex and thus sublinear function , i.e. for @xmath467 symmetric ( but not necessarily positive definite ) , we have @xmath468 we show that @xmath469 holds for a symmetric ( but not necessarily positive definite ) matrix @xmath470
. indeed , denote by @xmath471 the eigenvalues of the symmetric matrix @xmath163 with corresponding eigenvectors @xmath472 , forming a basis of @xmath35 .
then @xmath473 thus , @xmath474 is an eigenvector of @xmath475 with eigenvalue @xmath476 .
hence , @xmath477 fix @xmath391 and @xmath295 such that @xmath478 for all @xmath294 $ ] as in definition [ basic ] .
then @xmath479 , @xmath365 and @xmath364 $ ] .
now , using ( [ * ] ) and @xmath480 , where @xmath481 for a matrix @xmath482 , we have with remark [ con_new ] @xmath483 by lemma [ 4.12 ] and the definition of @xmath354 .
we summarize the results of this section in the following theorem .
[ summary ] let @xmath174 be a triangulation as in definition [ triconstr ] and @xmath484 .
let all constraints be satisfied .
then the cpa metric defined in definition [ cpa - def ] fulfills : 1 .
@xmath367 is symmetric , positive definite and a @xmath42-periodic function and thus defines a riemannian metric on @xmath178 in the sense of definition [ riem ] .
@xmath485 for all @xmath466 , where @xmath486 , @xmath487 is defined in theorem [ th1 ] and @xmath488 .
if @xmath489 is positively invariant , then @xmath99 exists and is finite for all solutions @xmath100 with @xmath101 and all functions @xmath102 , \mathbb r^n)$ ] . hence , @xmath11 satisfies all assumptions of theorem [ th1 ] .
. follows directly from lemma [ le1 ] and [ le4.9 ] . for part 2 . , note that @xmath490 exists and is positive , since @xmath11 is positive definite for all @xmath425 and depends continuously on @xmath425 . then @xmath491w}{w^t m({\tilde{x}})w}\\ & \le&\frac{1}{2}\frac{\lambda_{max}(m({\tilde{x}})d_xf({\tilde{x}})+d_xf({\tilde{x}})^tm({\tilde{x } } ) + m'_+({\tilde{x}}))}{\lambda_{max}(m({\tilde{x}}))}\\ & \le&-\frac{1}{2\mu_{max}}\end{aligned}\ ] ] using lemma [ le2 ]
. moreover , constraint 2 .
yields that @xmath492 .
. follows from lemma [ le4.9 ] using that @xmath489 is positively invariant .
this shows the theorem .
while we are primarily interested in the calculation of any riemannian contraction metric , i.e. a feasible solution of the semidefinite optimization problem , the optimization problem also allows for an objective function .
theorem [ summary ] suggests a possible objective function , namely simply the maximum over all @xmath311 , i.e. @xmath493 .
we can either implement this by adding @xmath312 as an additional variable with constraints @xmath494 for all @xmath203 .
or we can replace the variables @xmath311 in constraint 2 . by the uniform bound @xmath312 directly , which reduces the number of constraints 2 .
to @xmath310 , the number of vertices . by minimizing the constant @xmath312 , we minimize the bound @xmath495 on the maximal floquet exponent .
note , however , that @xmath312 has a lower bound given by @xmath496 .
this means that by choosing @xmath496 too large , the estimate on the floquet exponent will be very rough .
in the next theorem we assume that there exists an exponentially stable periodic orbit .
then we can show that the semidefinite optimization problem has a feasible solution and we can thus construct a suitable riemannian metric .
we have to assume that the triangulation is fine enough , i.e. in practice we start with a coarse triangulation and refine until we obtain a solution .
the triangulation has to stay suitably regular , i.e. the angles in simplices have lower and upper bounds . for simplicity
, we use the reference simplicial complex and scale it uniformly as described in section [ tria ] , but other refinements are also possible .
[ feas ] let the system @xmath0 , @xmath497 have an exponentially stable periodic orbit @xmath104 with basin of attraction @xmath498 .
let @xmath499 , @xmath500 , @xmath501 be a compact set in the cylinder @xmath171 .
fix @xmath328 .
then there is a @xmath502 such that the semidefinite optimization problem is feasible for all triangulations @xmath174 as described in section [ tria ] with @xmath503 .
note that we can choose the constants @xmath311 and @xmath347 to be the same for each simplex , i.e. @xmath504 and @xmath505 for all @xmath203 . 1
. * smooth riemannian metric * + denote the maximal real part of the floquet exponents of the exponentially stable periodic orbit @xmath104 by @xmath506 and set @xmath507 .
+ since @xmath499 is compact and @xmath498 open , there is a positive euclidean distance between @xmath152 and the boundary of @xmath498 .
let @xmath508 denote this distance if it is finite and otherwise set @xmath509 .
now define @xmath510 to be the set of all @xmath511 that have euclidean distance less than or equal to @xmath512 to @xmath152 .
clearly @xmath513 .
for all large enough @xmath514 the euclidean distance from the boundary of @xmath515 to @xmath152 is bounded above by @xmath516 , see ( [ hnu ] ) .
thus , there is a @xmath517 such that @xmath518 for every @xmath519 . +
now apply the following theorem to the above defined @xmath510 and @xmath278 .
+ [ lyapweight ] assume that @xmath520 is a periodic function in @xmath121 with period @xmath42 , and all partial derivatives of order one with respect to @xmath57 are continuous functions of @xmath7 .
let @xmath521 be an exponentially asymptotically stable periodic orbit , @xmath498 be its basin of attraction , and let the maximal real part of the floquet exponents be @xmath506 .
then for all @xmath522 and all compact sets @xmath510 with @xmath523 there exists a riemannian metric @xmath524 , such that @xmath525 holds for all @xmath526 .
+ the proof of this theorem , cf .
* theorem 4.2 ) shows that we have @xmath527 , if @xmath43 , as is the case by our assumptions .
+ since both @xmath528 and @xmath529 are continuous functions and @xmath530 is positive definite for all @xmath531 , there exists an @xmath532 such that @xmath533 for all @xmath531 , as @xmath510 is compact .
+ moreover , since @xmath534 is a continuous function ( note that @xmath530 is smooth ) satisfying @xmath535 for all @xmath531 , we have @xmath536 for all @xmath531 .
as @xmath537 and @xmath360 are continuous functions , there exists an @xmath538 such that @xmath539 for all @xmath531 , as @xmath510 is compact . +
now define @xmath540 .
then @xmath541 for all @xmath542 .
+ @xmath11 is @xmath414 on the compact set @xmath510 , so that we can define the following constants @xmath543 where @xmath544 denotes the hessian of @xmath409 .
2 . * assigning the variables of the optimization problem * + set @xmath545 + finally , we define , using the constant @xmath202 from lemma [ xlemma ] , @xmath546^{-1}&\mbox { if } f\in c^2\\ \,[2n(\sqrt{n+1}(1 + 4n)bd+2n(n+1)b_3c)]^{-1/2}&\mbox { if } f\in c^3\end{array}\right.\\ h_2^*&:=&\frac{1}{2nfc^*}. \end{aligned}\ ] ] now define @xmath547 let @xmath503 and consider the triangulation @xmath548 .
note that , since @xmath503 , we have by ( [ hnu ] ) @xmath549 interpolate @xmath367 on this triangulation , i.e. assign the variables @xmath355 for all vertices @xmath303 of the triangulation ; moreover , set @xmath504 and @xmath505 for all @xmath203 . by ( [ hnu2 ] ) and the definition of @xmath550 this ensures @xmath551 3 . * auxiliary results * + to check the feasibility in the next step , we need some estimates .
+ consider a simplex @xmath552 .
denote , as in constraint 3 . , ( [ defw ] ) @xmath553 , where @xmath554 and @xmath555 note that @xmath409 is two times continuously differentiable on @xmath556 and for @xmath557 we thus have by taylor s theorem @xmath558 where @xmath559 is the hessian of @xmath238 and @xmath560 for some @xmath5610,1[$ ] . + by rearranging terms and combining , this delivers with ( [ wknu ] ) @xmath562 we have @xmath563 note that @xmath564 since the simplex is convex .
hence , by ( [ wmxnw ] ) , @xmath565 + now we need to obtain an estimate on @xmath566 . for @xmath567 , @xmath568 ,
there is a @xmath569 on the line segment between @xmath570 and @xmath571 , such that @xmath572 where @xmath573 denotes the @xmath574-th component of @xmath575 and @xmath576 is the gradient of this function .
then we have @xmath577 hence , @xmath578 from this , lemma [ xlemma ] and ( [ ain1 ] ) we obtain the estimate @xmath579 using ( [ hnu ] ) and the definition of @xmath580 . + a further useful consequence , which we need later ,
is that @xmath581 4 . *
feasibility * + now we check that the assignment in step 2 . is a feasible point of the semidefinite optimization problem .
we discuss each constraint in the following . 1 .
this follows directly from the periodicity of @xmath11 .
2 . this follows directly from the definition of @xmath312 .
3 . we have for all @xmath582 and any @xmath323 by ( [ cfirst ] ) @xmath583by definition of @xmath584 as @xmath585 .
4 . this follows from ( [ cond4 ] ) .
this is the main step and will be shown below .
+ we fix a simplex @xmath586 and a vertex @xmath303 , @xmath587 . with remark [ con_new ]
we need to show that @xmath588 we have @xmath589 , cf .
( [ enu ] ) . hence , using the sublinearity of @xmath360 , we have @xmath590 where we have used ( [ cond5 ] ) and @xmath591 , cf .
( [ 2-eig ] ) .
+ using @xmath592 , we obtain @xmath593 with ( [ xfirst ] ) , the definition of @xmath594 and using @xmath595 .
e. m. aylwarda , p. a. parrilo & j .-
e. slotine : _ stability and robustness analysis of nonlinear systems via contraction metrics and sos programming_. automatica * 44 * ( 2008 ) 8 , 21632170 .
r. baier , l. grne & s. hafstein : _ linear programming based lyapunov function computation for differential inclusions_. discrete contin . dyn .
b * 17 * ( 2012 ) 1 , 3356 .
m. branick & s. wiggins : _ an adaptive method for computing invariant manifolds in non - autonomous , three - dimensional dynamical systems_. physica d * 238 * ( 2009 ) , 16251657 .
f. camilli , l. grne & f. wirth : a generalization of zubov s method to perturbed systems .
_ siam j. control optim .
_ * 40 * ( 2001 ) 2 , 496515 .
s. hafstein : _ an algorithm for constructing lyapunov functions_. volume 8 of _ electron . j. differential equ .
_ texas state univ . ,
dep . of mathematics ,
san marcos , tx , 2007 .
available electronically at ` http://ejde.math.txstate.edu/ ` .
b. krauskopf , h. osinga , e. j. doedel , m. henderson , j. guckenheimer , a. vladimirsky , m. dellnitz & o. junge : _ a survey of methods for computing ( un)stable manifolds of vector fields_. internat . j. bifur
. chaos appl .
* 15 * ( 2005 ) 3 , 763791 .
a. papachristodoulou & s. prajna : _ the construction of lyapunov functions using the sum of squares decomposition_. in : proceedings of the 41st ieee conference on decision and control ( 2002 ) , 34823487 .
m. peet & a. papachristodoulou : _ a converse sum - of - squares lyapunov result : an existence proof based on the picard iteration_. in : proceedings of the 49th ieee conference on decision and control ( 2010 ) , 59495954 . | a riemannian metric with a local contraction property can be used to prove existence and uniqueness of a periodic orbit and determine a subset of its basin of attraction . while the existence of such a contraction metric is equivalent to the existence of an exponentially stable periodic orbit , the explicit construction of the metric is a difficult problem . in this paper , the construction of such a contraction metric
is achieved by formulating it as an equivalent problem , namely a feasibility problem in semidefinite optimization .
the contraction metric , a matrix - valued function , is constructed as a continuous piecewise affine ( cpa ) function , which is affine on each simplex of a triangulation of the phase space .
the contraction conditions are formulated as conditions on the values at the vertices .
the paper states a semidefinite optimization problem .
we prove on the one hand that a feasible solution of the optimization problem determines a cpa contraction metric and on the other hand that the optimization problem is always feasible if the system has an exponentially stable periodic orbit and the triangulation is fine enough .
an objective function can be used to obtain a bound on the largest floquet exponent of the periodic orbit . |
there is currently much interest in performing ancestral inference from molecular population genetic data . to facilitate this inference ,
there has been an explosion of research in developing computationally efficient methods .
these techniques are designed either to compute the likelihood , for maximum likelihood estimation , of a sample of genes or for deriving the posterior distribution on parameters in coalescent models , which describe the ancestry of the genes .
broadly speaking there are three main approaches to inference in molecular population genetics : ( i ) importance sampling for likelihood evaluation , whose application in population genetics was pioneered by ( griffiths & tavar , 1994a , b , c ) ( ii ) markov chain monte carlo methods ( e.g. kuhner et al .
( 1995 ) , wilson & balding ( 1998 ) ) ( iii ) approximate bayesian computation ( abc ) ( del moral et al .
( 2009 ) , marjoram et al .
( 2003 ) ) .
see stephens ( 2004 ) for a review .
in this paper we concentrate on likelihood - based methods .
molecular data have a sampling distribution which is a mixture over possible ancestries .
the state space of the ancestries is huge and closed - form expressions are available only in the simplest cases .
the objective is to calculate a parameter @xmath0 ( @xmath1 ) such that @xmath2 for some observed genetic data @xmath3 , parameter @xmath4 , probability density @xmath5 on @xmath6 and @xmath7 an integrable function .
note that @xmath8 is typically the genetic types of a random sample of chromosomes .
in addition , @xmath9 denotes the coalescent history , i.e. the set of ancestral configurations at the embedded events in a markov process where coalescence , mutations or other events take place .
@xmath10 denotes the current state , while @xmath11 is the state when a singleton ancestor is reached . statistical inference associated to @xmath12
can be regarded as a missing data problem and could , in principle , be tackled by the em algorithm ( dempster et al . 1977 ) and its monte carlo extensions ( e.g. fort & moulins ( 2003 ) ) .
however , @xmath13 , the stochastic tree , can be computationally expensive to simulate and such techniques are typically avoided .
for example , for the coalescent ( kingman , 1982 ) and ancestral recombination graphs ( e.g. fearnhead & donelly ( 2001 ) ) , the standard approach is to use is ( de iorio & griffiths , 2004a ; griffiths & tavar , 1994a ; stephens & donelly , 2000 ) and smc methods ( chen et al .
2005 ) to approximate .
these approximations are usually computed on a discrete grid @xmath14 and the estimate of @xmath15 corresponds to the largest approximated likelihood on @xmath16 .
see also olsson & rydn ( 2008 ) for an alternative procedure for state - space models .
techniques such as abc and composite likelihood ( wiuf , 2006 ) do not give solutions which are exact w.r.t . the original model whilst , when possible , exact inference is of interest .
this is because , given a reasonable stochastic model , the approach allows investigators to exactly ( up - to a numerical error ) average over the uncertainty in the tree structure when estimating genetic parameters of interest .
one of the main drawbacks of existing exact is / smc schemes is the simulation of the tree backward in time , from observed data , until the tree coalesces . in many scenarios , especially for large data sets , when getting close to the top of the tree , it often takes a long time to coalesce .
this is due to genetic parameters ( e.g. mutation rates ) that can be very large relative to the size of the data .
consequently , it can take a very long time to simulate the tree back to the mrca . as a result
, the variance of the estimate of the likelihood can be higher than is desirable , along with long cpu times .
it should be noted that the calculation of the likelihood at these points , @xmath4 , can be inferentially important .
in addition , it is seldom possible to speed up the simulation via importance sampling as the variance of the weights can become too large .
that is , by adapting the parameter of the proposal to lead to a fast coalescence , the discrepancy between the true process and the proposal leads to a very inefficient algorithm w.r.t . variance .
the approach proposed in this paper is based on is .
stephens & donelly ( 2000 ) proposed a way to use is efficiently to simulate ancestral trees by characterizing an optimal proposal distribution and similar methods have since been developed for a variety of genetic scenarios ( e.g. de iorio & griffiths ( 2004a , b ) ) .
the basic idea is to define an efficient proposal distribution on ancestral histories which allows us to reconstruct markov histories backwards in time from the sample @xmath8 to an mrca .
we introduce a stopping time in the is proposal , backward in time , to stop the simulation before the mrca is reached . then using a simple stopped identity , forward in time we are able to characterize the bias introduced in the evaluation of the likelihood due to stopping the simulation of the stochastic tree .
the bias can be understood by considering two aspects : 1 .
the underlying mixing of the evolutionary process[pt : evpr ] 2 .
the last exit time distributions on the process .
[ pt : lepr ] in the context of ( [ pt : evpr ] ) , the idea is that for many models , close to the top of the tree , the process is able to forget its initial condition . as a result ,
stopping the simulation is reasonable , because the place where it is stopped is forgotten by the process forward in time ; we formalize these ideas later on . in reference to ( [ pt : lepr ] ) , the more information there is on the true marginal distributions of the process , the more it is possible to reduce the bias .
ideas from the theory of population genetics models ( ethier & griffiths 1987 ; ewens , 1972 ) will be used to achieve the latter . in reference to a comment of edwards ( 2000 ) , our method is termed the ` time machine ' .
this is because , estimation is performed saving the simulation time of going all the way back in time to the mrca .
a similar idea , in the context of filtering , can be found in the work of olsson et al .
( 2008 ) and also in option pricing avramidis & lcuyer ( 2006 ) . in our context
, we have a simpler underlying process than in filtering , but the ergodicity conditions considered there do not apply here .
the mixing conditions that they require only apply locally and thus the proofs have to be modified .
recall that approximate tools for inference from stochastic trees ( e.g. del moral et al .
( 2009 ) , meligkotsidou & fearnhead ( 2007 ) , tavar et al .
( 2000 ) ) are available . however , our approach is ` less approximate ' , in that our point - wise estimate of the likelihood is significantly less - biased , but costing more in computational - time .
this paper is structured as follows . in section
[ sec : motex ] we introduce a motivating example , the coalescent model , which will help to illustrate our ideas . in section
[ sec : stopsimos ] our methodology is described ; section [ sec : bias ] features an analysis of the bias of the approach ; section [ sec : simos ] presents a simulation study to demonstrate the performance of our algorithm and we conclude the paper in section [ sec : summary ] .
appendix 1 contains some proofs , appendix 2 details of our numerical implementations .
our ideas are illustrated in the context of the coalescent .
however , the formulation is kept as general as possible , as the framework can be extended to other tree models , such as the infinite sites model . in appendix 3
we show how this can be done .
the coalescent model is used as a motivating example for our work .
some notations are first introduced . in particular
, we consider the case in which the type space @xmath17 for the collection of the @xmath18 genes / chromosomes is finite and the only genetic process of interest is mutation .
denote by @xmath19 a measurable space .
for two @xmath20finite measures @xmath21 and @xmath22 mutual absolute continuity is written @xmath23 and the radon - nikodym derivative as @xmath24 . given a markov kernel @xmath25 $ ] , let @xmath26 , ( the dirac measure ) and write the composition for @xmath27 as @xmath28 , with a corresponding composition of inhomogeneous kernels as @xmath29
. write @xmath30 as the indicator of a set . for @xmath31 @xmath32
denotes the class of stochastic matrices for which there exist a stationary distribution @xmath33 .
the collection of bounded and measurable function are denoted @xmath34 .
the supremum norm is written @xmath35 .
the total variation distance between two probability measures @xmath21 and @xmath22 on @xmath19 is @xmath36 .
given a probability measure @xmath37 , and a @xmath38 , the product measure is written @xmath39 , @xmath40 .
the vector notation @xmath41 is adopted .
in addition , let the @xmath42-dimensional vector @xmath43 where the 1 is in the @xmath44 position .
the @xmath45 norm of a vector is written @xmath46 . for @xmath47 , @xmath48 .
define the tree model on the measurable space @xmath49 , with @xmath50 .
let @xmath51 .
the basic idea is to maximize , w.r.t @xmath4 , the quantity @xmath52 where the observed data is @xmath53 , @xmath54 , normally the identity , for @xmath55 @xmath56 and @xmath57 for some @xmath58 and @xmath59 depending upon the model under study . in all of our examples
, @xmath60 corresponds to the density of a non - decreasing ( in some sense ) markov process in discrete time , stopped at a random time @xmath61 ; that is @xmath62 throughout the article it is assumed that @xmath63 , i.e. that the stopping time is a.s . finite w.r.t @xmath5 .
the stopping time will be determined by the first time that the tree is of ` size ' @xmath64 .
introduce an absolutely continuous distribution @xmath65 on @xmath6 and sample @xmath66 according to @xmath65 , then the is estimator of @xmath12 is @xmath67\ ] ] where @xmath68 and @xmath69 the empirical measure of the simulated samples .
denote the number of genes of type @xmath70 at event @xmath71 of the process as @xmath72 , with @xmath73 .
the objective is to find the genetic parameters @xmath74 where @xmath75 and @xmath76 , @xmath77 .
@xmath78 is the mutation rate per chromosome per generation and mutations along the edges of the tree occur according to a markov chain with transition matrix @xmath79 .
the various components of the identity ( [ eq : likelihood_id ] ) for the coalescent model are defined as : @xmath80 with @xmath81 the identity function , @xmath82 and finally , @xmath83 where @xmath84 @xmath85 and @xmath86 write @xmath87 ( here @xmath88 is counting measure ) .
note that for any fixed @xmath51 , @xmath4 , @xmath89 .
for simplicity of exposition , the results are given with only mutation .
however , they can be easily extended to the case of migration as well ( e.g. de iorio & griffiths ( 2004b ) ) . to compute the likelihood , for a given @xmath4
, importance sampling is adopted .
an importance distribution , @xmath65 , is introduced to simulate the tree backward in time to the mrca ; this ensures that the data is hit . in details ,
let @xmath90 denote the reverse chain backward in time and write @xmath91 instead of @xmath13 ( this convention is used throughout the article , see also figure [ fig : coalgraph ] ) .
let : @xmath92 for some markov transition @xmath93 ; see stephens & donelly ( 2000 ) for the optimal @xmath65 .
then the likelihood is @xmath94 the simulation proceeds by sampling from @xmath95 and computing the weight @xmath96 simulations backward in time are carried out until we reach the mrca , i.e. when there is only one individual in the sample . this procedure is repeated @xmath97 times to provide a monte carlo estimator for the likelihood @xmath98\ ] ] where @xmath99 are the simulated samples , @xmath100 for every @xmath101 and @xmath102 this can be repeated for many @xmath103 using a driving value ( griffiths & tavar , 1994 ) or bridge sampling ideas ( e.g. fearnhead & donelly ( 2001 ) ) .
in addition , to deal with the problem of weight degeneracy ( e.g. doucet et al .
( 2001 ) ) resampling steps can be added .
see , for example , chen et al .
( 2005 ) .
( 1,1 ) ( .5,.75)(0,1)0.25 ( 0.25,.75)(1,0).5 ( 0.75,.75)(0,-1)0.5 ( 0.25,.75)(0,-1)0.1 ( 0.125,.65)(1,0)0.25 ( 0.125,.65)(0,-1)0.65 ( 0.375,.65)(0,-1)0.15 ( 0.3125,.5)(1,0)0.125 ( 0.3125,.5)(0,-1)0.5 ( 0.4375,.5)(0,-1)0.5 ( 0.625,.25)(1,0).25 ( 0.625,.25)(0,-1).25 ( 0.875,.25)(0,-1).25 ( 0.1,0.5)(0.1,0)8(1,0)0.05 ( 0.95,.95)@xmath104 ( 0.01,.95)@xmath105 ( 0.95,.75)@xmath106 ( 0.95,.65)@xmath107 ( 0.95,.5)@xmath108 ( 0.95,.25)@xmath109 ( 0.95,.01)@xmath110 ( 0.01,.5)@xmath111 ( 0.01,.25)@xmath112 ( 0.01,.01)@xmath113 ( 0.25,.45)@xmath114 ( 0.7,.45)@xmath115
it is now detailed how we stop the simulation of the stochastic tree back in time before the mrca is reached . in the next section we provide theoretical results and connections to the theory of smc
are established .
for the purpose of stopping the simulation , introduce two stopping times ( forwards in time ) : the first hitting time of the set @xmath116 @xmath117 and some stopping time @xmath118 associated to the hitting of a set @xmath119 @xmath120 such that @xmath121 where @xmath122 is the @xmath123probability .
for example , in the context of the coalescent , it is suggested to take , for @xmath124 @xmath125 let @xmath126 denote the expectations w.r.t the process @xmath127 .
then the likelihood ( [ eq : likelihood_id ] ) can be written as @xmath128\ ] ] and applying the strong markov property we have @xmath129\;\ ] ] that is , @xmath130 dz_{\alpha } \label{eq : stoplike}\end{aligned}\ ] ] where @xmath131 the equation ( [ eq : stoplike ] ) will be the starting point for constructing our biased estimates of the likelihood function .
consider the coalescent model . specifically ,
define , for @xmath132 the stopping time @xmath118 @xmath133 which is the first time the forward process has @xmath134 individuals .
using equation ( [ eq : stoplike ] ) , we have @xmath135dz_{\alpha } \label{eq : coallikestp}.\ ] ] in words this means that to have @xmath134 chromosomes , we need a minimum of @xmath136 steps in the process and @xmath137 has to be at least @xmath138 steps . in this case ,
write @xmath139 note this is well - defined due to the fact that the size of the population is non - decreasing , and then , for any @xmath140 @xmath141 where @xmath142 that is , given @xmath118 , the distribution of the chromosome counts at the first entrance time of @xmath143 can be written as the composition of : * the distribution of the counts at the last exit time from @xmath144 * and the markov transition .
returning to the likelihood ( [ eq : coallikestp ] ) and making the substitutions , @xmath145 , @xmath146 , it thus follows that @xmath147 here @xmath148 is the time from the last time there are @xmath136 chromosomes to @xmath64 chromosomes .
now set @xmath149 in other words the simulation is stopped the first time there are @xmath136 chromosomes .
our approximation of the likelihood is then @xmath150 on the basis of the above analysis , it is then clear that if @xmath151 then the approximation of the likelihood is exact .
that is , to minimize the bias an approximation of the true distribution of the counts at the last time there are @xmath136 chromosomes should be used .
the ideas and notation are clarified in figure [ fig : coalgraph ] .
in our biased simulation , using the decomposition , the procedure will approximate @xmath152 where our notation is such that : * @xmath153 is the time reversed process * @xmath114 is a first hitting time associated to @xmath153 * @xmath154 an approximation of a marginal probability .
we begin by giving a simple result on the error bounds for smc algorithms .
the result applies to the standard is algorithms , for example in de iorio & griffiths ( 2004b ) , stephens & donelly ( 2000 ) , and for the smc algorithms as in chen et al .
the simulation is to be performed backward in time , as in section [ sec : likelihood_computation ] .
the ideas here are adapted from the theory of del moral ( 2004 ) .
the biased estimates are denoted as @xmath155 , @xmath156 , where @xmath157 depends upon whether is or smc is implemented .
for example , in the is case : @xmath158 where @xmath159 is such that @xmath160 is the set associated to @xmath161 ( @xmath162a.s . ) , @xmath163 and @xmath164 .
below expectations w.r.t the stochastic process that is simulated by the algorithm are written as @xmath165 and it is assumed @xmath166 [ prop : lpbound ] for any @xmath51 , @xmath167 , @xmath4 , @xmath168 , there exists a @xmath169 such that : @xmath170^{1/p } \leq \frac{b_{p , n}(\theta)}{\sqrt{n } } + |l_b(y_{1:n};\theta ) - l(y_{1:n};\theta)|.\ ] ] * remark*. _ the result shows the standard variance - bias type decomposition .
that is , @xmath171 can be thought of as a bound on the variance and @xmath172 is the bias .
our estimate converges to @xmath173 , and it is sought to control the bias term , which , in our case can be approximately written in the form @xmath174(p_{1:k}(f))| \label{eq : biasdecomp}\ ] ] for @xmath175 two probability measures and @xmath176 a sequence of non - homogenous markov kernels ( @xmath103 is suppressed on the r.h.s ) . _ a simple technical result is now given which shows how to control the bias term ( [ eq : biasdecomp ] ) .
some assumptions are now made , that can be satisfied by many stochastic tree models .
introduce a sequence of time inhomogeneous markov kernels @xmath177 , on space @xmath178 and a sequence of sets @xmath179 .
[ hyp : p_nassump ] _ stability of @xmath177 .
_ * \(i ) * initial probability measures*. @xmath175 are concentrated on @xmath180 .
* \(ii ) * absorption of @xmath177*. for every @xmath181 , @xmath182 we have @xmath183 * \(iii ) * local mixing of @xmath177*. for every @xmath181 , there exist @xmath184 , @xmath185 concentrated on @xmath186 , such that for all @xmath182 @xmath187 the assumption ( a[hyp : p_nassump ] ) ( which is comprised of ( i)-(iii ) ) will refer to the fast mixing of the process close to the top of the tree .
the absorption type assumption refers to the birth process associated to coalescent type chains .
[ prop : biascontrol ] assume ( a[hyp : p_nassump ] ) .
then , for any @xmath188 , define : @xmath189 and we have @xmath190p_{1:k}\|_{tv } \leq \vartheta_k\|\lambda_1-\lambda_2\|_{tv}.\ ] ] * remark 1*. _ the result helps to bound the bias as @xmath191(p_{1:k}(f))| \leq \|f\|_{\infty}\|[\lambda_1-\lambda_2]p_{1:k}\|_{tv}.\ ] ] essentially , the fast mixing of @xmath192 within the domain it is constrained to allow the composition of kernels to forget its initial distribution at an exponential rate . in addition , as in olsson et al .
( 2008 ) , assuming @xmath193 is uniform in @xmath194 , the benefits of stopping , in terms of variance / bias trade off can be substantial .
_ * remark 2*. _ one point of interest in the sequel is that , if the mixing condition ( a[hyp : p_nassump ] ) does not hold , it is possible to establish a similar bound when the initial measures @xmath21 and @xmath22 are similar .
that is to say , when @xmath23 and @xmath195 such that @xmath196 this is unsurprising as it implies that if the kernels do not mix , we need to ` match ' @xmath21 and @xmath22 for the bias to be small .
_ ( a[hyp : p_nassump ] ) is now discussed in the context of the coalescent .
note that the results follow , with some extra work , for coalescent processes with migration .
readers interested in how the method may be applied can skip to section [ sec : simos ] , with no loss in continuity .
suppose that the transition matrix @xmath79 satisfies , for any @xmath197 , @xmath198 and probability @xmath199 , @xmath200
@xmath201 this condition implies that @xmath79 mixes extremely quickly .
let @xmath202 ; this corresponds to the space of @xmath203 .
also let @xmath204 .
it is clear that @xmath205 : since we start with at most 3 chromosomes and the most possible after 3 steps is 6 .
now it can be seen that , for any @xmath206 @xmath207 and @xmath208 with @xmath209\bigg(\frac{\mu}{\mu+2}\bigg)^2\bigg]^3.\ ] ] here the minorising probability @xmath210 puts all its probability on having 3 chromosomes .
then it can be subsequently seen that @xmath211 satisfies condition ( [ eq : mixingcond ] ) , with @xmath212 and so fourth . in effect
the condition ( [ eq : mixingcond ] ) holds with @xmath213 ; that is , the closer to the top of the tree we stop , the faster the process will mix forward in time . as a result , to bound the bias we can write it , approximately , in the form , for @xmath214 @xmath215\ ] ] with @xmath21 as in ( [ eq : lastexit ] ) ,
@xmath216 , @xmath217 , @xmath218 is associated to the fact that we need to iterate the kernels to satisfy ( [ eq : mixingcond ] ) and @xmath219 .
@xmath220 is an integer big enough ( say @xmath221 ) where we suspect that the possibility of generating a tree of length @xmath194 and hitting the data is extremely small , so we can neglect the upper term .
thus , approximately , the bound shows that the bias falls geometrically as we stop closer to the top of the tree .
note , however , it can not go to zero unless @xmath21 and @xmath22 are equal . to an extent ,
finding good approximations is more difficult than being able to stop the tree , which is why we focus on this . *
remark 1*. _ the result given here mirrors one proved by donelly & kurtz ( 1999 ) for fleming - viot models . in theorem 9.4 of that paper they show that the particle process is uniformly ergodic , if the mutation process is .
this is very similar to the property established above . _ * remark 2*._the information , in terms of when to stop the simulation , that is contained in the bound on the bias is as follows .
if the mutation process mixes quickly , as above , then the bias falls at a geometric rate : we should stop the simulation when the process starts to mutate many times .
this could be measured in terms of the effective sample size ( e.g. liu ( 2001 ) ) , if trees are simulated in parallel , or alternatively , if @xmath222 , for @xmath223 a large multiple of the current size of the tree .
_ * remark 3*. _ in terms of the expression @xmath224 , one could adopt a parent - independent mutation ( pim ) marginal .
if we have @xmath225 where @xmath226 is the transition for the pim , and the mutation vector is @xmath227 , then ideas from perturbed markov chains ( e.g. mitrophanov ( 2005 ) ) can be adopted to determine a quantitative bound .
we are currently investigating a meaningful bound . _
to illustrate our approach , we consider three simulation scenarios : two pim models and one parent dependent mutation model ( pdm ) .
the two pim models , denoted pim 0.5 - 0.5 and pim 0.1 - 0.9 are based on the following per - locus transition matrices : @xmath228 while the per - locus mutation probability matrix underlying the pdm model is @xmath229 in all three scenarios , the initial population was set to 100 sequences and we considered a single - locus case ( @xmath230 with 2 possible types ) . for the pdm model only , we also considered the case of 10 loci ( @xmath230 @xmath231=1024 different types ) .
irrespective of the number of loci considered , the distribution of the 100 initial sequences among the different types was sampled from a multinomial distribution with a probability vector @xmath79 defined as the invariant point , solution of equation , @xmath232 , where @xmath194 denotes the number of loci considered and @xmath233 is the full mutation probability matrix . the algorithm description is given in appendix 2 .
for the function @xmath234 , we use the distribution of an un - ordered sample from a pim model ( which , even for the pim cases , is not the correct distribution in the bias term ) .
simulations were carried out until there were @xmath235 sequences left in the population .
@xmath236 exactly corresponds the approach in stephens & donelly ( 2000 ) and is subsequently referred to as @xmath237 .
for each simulation , we examined 60 values for @xmath78 ranging from 0.1 to 30.1 .
we report in figure [ graphe_distrib ] the estimated log - likelihood distribution , based on 100,000 samples for all four simulations scenarios ( presented in lines ) and three values of @xmath78 ( presented in columns ) .
estimated distribution of the likelihood for the four simulation scenarios and for three values the mutation rate @xmath238 and @xmath239 . in each model
, results are presented for the six stopping times in the simulation of the genealogical tree .
plots are based on 100,000 samples . ] in figure [ graphe_distrib ] it is clear that as expected , uniformly across the values of @xmath78 , the closer to the mrca the algorithm is stopped , the more accurate the distribution of the likelihood is estimated .
however , up to tm 10% ( and even tm 25% for @xmath240 ) our results suggest that the time machine approximation and correction provides an accurate estimate of the distribution of the likelihood .
conversely , when the algorithm is stopped too early ( tm @xmath241 25% ) the biased estimator underlying the time machine approach leads to very inaccurate estimates of the likelihood . for even more extreme cases ( tm 50% for @xmath242 ) , this results in a highly shifted estimated distribution of the likelihood .
the above observations are also reflected in the mean likelihood ( figure [ meanl ] ) . for every model considered here , the simulations of the time machine up to tm 25% seem to provide estimates of the mean likelihood that are similar to the sd approach , although for larger values of @xmath78 , tm 25% seems to overestimate the mean likelihood .
furthermore , the time machine approach seems to accurately locate the value of @xmath78 maximizing the likelihood for tm@xmath243 10% , and to provide acceptable approximations for for this when tm@xmath244 10% , regardless of the simulation scenario .
estimated likelihood for the four simulation scenarios as a function of the mutation rate @xmath78 .
plots are based on 100,000 samples . ] in figure [ meant ] , the average computation time per iteration is plotted as a function of @xmath78 for the pdm-10 loci simulations .
results for all other models led to the same conclusions and are therefore not shown . from this figure , the computation time appears to be a linearly increasing function of @xmath78 : increasing the mutation rate naturally decreases the probability of simulating a coalescent event and therefore tends to increase the time to reach the mrca ( or any population size ) .
however , it seems that stopping the simulation when there are only more than 5 sequences left in the population drastically reduces the computation time : for tm 5% the simulation run is on average more than twice as fast as the sd simulation , and for tm 25% , the time machine is more than 3 times more time - efficient than the sd algorithm .
it should also be noted that ` large ' values of @xmath78 ( around 10 ) , for which the time savings are most significant , also seem to be inferentially important ( see the fourth panel of figure [ meanl ] ) . in figure [
fig : relative_sd ] the relative standard deviation across our 100 repeats of the algorithm , of the time machine to sd are plotted for all the scenarios considered .
it can be seen , as expected , that there is some variance reduction and , for example for the tm 5% pdm , the variance reduction is of the order 1.5 .
average computation time as a function of the mutation rate @xmath78 .
figures are based on 100,000 samples . ]
relative standard deviation across 100 repeats of the time machine to sd .
figures are based on 100,000 samples . ] on the basis of our experiments , combining both computational efficiency and the numerical accuracy , the use of the time machine with tm 5% is an efficient alternative to the sd algorithm .
the c++ code is available upon request from the third author .
in this paper we have considered a new approach for simulation of stochastic trees and likelihood calculation of sample probabilities in population genetics models .
the approach consists in stopping the backward simulations before the top of the tree is reached .
we have provided theoretical results on the bias introduced in the estimation of the likelihood .
some extensions to our work are described below .
firstly , to extend our analysis to different models .
the paper has been written to facilitate such analysis and we believe it is rather simple to deal with other stochastic tree models . also , some further empirical investigations would help support the simulations and theoretical analyses presented here .
our methodology would be further enhanced with gpu technology ( e.g. lee et al .
( 2010 ) ) , and this is one area that we are currently investigating . secondly , to look at the consistency ( in a likelihood sense ) of our biased monte carlo estimator . as we observed in section
[ sec : simos ] , it appears that the time machine seems to recover the maximum likelihood estimator .
therefore consistency , or potential asymptotic bias is of genuine interest .
there are very few results in the context of consistency , due to the dependency in the data , after integrating out the tree .
that is , it is difficult to apply uniform laws of large numbers to complex dependency structures .
none - the - less , we suggest the work of douc et el .
( 2004 ) , fearnhead ( 2003 ) , olsson et al .
( 2008 ) , olsson & rydn ( 2008 ) as possible starting points for a proof .
thirdly , the time machine can be used in the context of markov chain monte carlo ( mcmc ) .
if one is interested in bayesian parameter inference , then a stopping - time smc algorithm can be used within an mcmc algorithm ( particle mcmc ( andrieu et al .
significant time savings per iteration can be gained by using the time machine ; see jasra & kantas ( 2010 ) for some details .
we thank prof .
arnaud doucet for some valuable conversations related to this work .
we give the proofs of propositions [ prop : lpbound ] and [ prop : biascontrol ] . in the case of
is , the result follows by adding and subtracting @xmath173 applying minkoswki and the marincinkiewicz - zygmund inequality . in the case of the smc algorithm
the proof follows from the fact that the algorithm approximates a multi - level feynman - kac formula ; see chapter 12 , proposition 12.2.3 del moral ( 2004 ) .
note that this point is apparently over - looked in chen et al .
( 2005 ) , and such a result helps to verify the convergence of the algorithm .
in addition , note that the proposition 12.2.3 of del moral ( 2004 ) does not depend on the importance weights being upper - bounded by 1 .
hence , due to the boundedness of the weights , the same proof as for is applies , except the @xmath245 bound for particle approximations of feynman - kac formulae is used instead of the marincinkiewicz - zygmund inequality .
the proof is fairly simple and combines the proof of lemma 3.9 and theorem 4.1 of le gland & oudjane ( 2004 ) ( see also theorem 3.1 of tadi & doucet ( 2005 ) ) .
the idea is to use the contraction property of the total variation distance and hilbert metric , as well as the relation between the two ( see lemma 6.1 of tadi & doucet ( 2005 ) ) .
the only real complication is using the local mixing condition ( [ eq : mixingcond ] ) to derive a bound on the radon - nikodym derivatives @xmath246 consider @xmath247 , clearly @xmath248 in addition @xmath249 since @xmath250 it follows that @xmath251 the proof can then be concluded by following the arguments of le gland & oudjane ( 2004 ) , lemma 3.9 and theorem 4.1 .
let @xmath252 , @xmath253 , the population size within each of the @xmath42 states at time @xmath81 . the algorithm will simulate backward in time genealogical trees for an initial population , @xmath254 the ( @xmath255dimensional ) counts associated to the observed data , until there are @xmath256 sequences left in the population .
the case where @xmath256=@xmath257 corresponds to ordinary coalescent and @xmath258 to the time machine .
most of the notations can be found in sections [ sec : motivation ] and [ sec : id_interest ] . for any generation @xmath81
, there are @xmath259 sequences left in the population , the following steps will be iterated until @xmath259=@xmath256 : 1 . sampling the type of the offspring sequence ( @xmath70 ) with probability @xmath260 2 . getting the type of the ancestor sequence ( @xmath71 ) .
+ a sequence of a given type @xmath70 can have arisen from an ancestor sequence of type @xmath71 through : 1 . a coalescent event with a probability proportional to @xmath261 2 . a @xmath71 to @xmath70 mutation event ( inclusive of self mutations , from type @xmath70 to type @xmath70 ) , with probability proportional to : @xmath262 where @xmath263 3 .
updating the population sizes within each type .
+ @xmath264 4 .
calculate the contribution to the likelihood of the simulated event ( suppressing the subscript @xmath103 ) + @xmath265 where @xmath266 and @xmath267 5 .
updating the log likelihood + @xmath268 6 . assessing the stopping criterion .
+ when the time machine is used ( @xmath230 @xmath258 ) , steps 1 to 5 are repeated until @xmath269 .
otherwise , when the full tree is simulated ( @xmath256=@xmath257 ) , steps 1 to 5 are repeated until there are 2 sequences left in the population .
then , mutations are simulated until both remaining sequences are of the same type , based on the following three steps : 1 .
choose one of the two sequence , of type @xmath70 , with probability 0.5 .
2 . simulate the mutation event from an ancestor of type @xmath71 ( to type @xmath70 ) according to the probability defined in equation ( [ p_mut ] ) , and setting the coalescent probability to 0 .
calculate the corresponding weight for the sampled @xmath71 to @xmath70 simulated transition . at this final stage
there are only two individuals in the population ( @xmath270 ) , hence @xmath271 , and @xmath272 , then : @xmath273 when the final generation is reached @xmath274 , and @xmath275 for any other iterations in that step .
the weight for each generation , derived from is then defined as : @xmath276 4 .
update the likelihood according to equation ( [ update_l ] ) this step is specific to the time machine ( i.e. if @xmath277 ) . recalling that @xmath114 is the generation at which the iterative algorithm was stopped ,
the bias induced by stopping the simulation before reaching the mrca is estimated as : @xmath278 where @xmath279 denotes the gamma function .
the likelihood of the tree is then updated @xmath280 the above algorithm is independently repeated @xmath97 times , the estimate of the log - likelihood is @xmath281 where @xmath282 is the value of the final weight for sample @xmath70 and @xmath283 .
we now consider our results in the context of the infinite sites model .
we concentrate upon likelihoods associated to rooted genealogical trees ; see ethier & griffiths ( 1987 ) or griffiths & tavar ( 1995 ) for more details .
the model is based upon the simulation of distinct dna sequences , and the multiplicity of the sequences . in more details ,
the simulation begins with a single dna sequence @xmath284 , and counts @xmath285 .
the process can then undergo a mutation ( rate @xmath78 ) or a split . if a mutation occurs ( to the first sequence say ) we have the new state @xmath286 , @xmath287 and @xmath288 , otherwise the new state is @xmath284 and @xmath289 .
the key point is that new mutations introduce a new site ( that is a new integer number ( which is larger than all others currently present ) to the start of a selected sequence ) and hence dna sequence , whilst splits only increase the number of an existing sequence .
the state - space consists of the @xmath255distinct sequences ( vectors of potentially different length sequences @xmath290 ) and the respective counts @xmath291 of the sequences that have been simulated .
that is , in the previous notation @xmath292 the simulation stops , as before , when @xmath293 . in general ,
transitions are governed by the following markov kernel .
a mutation ( rate @xmath78 ) , at time @xmath294 , of the @xmath295 sequence occurs with probability @xmath296 and a split of the @xmath295 sequence occurs with probability @xmath297 see ethier & griffiths ( 1987 ) and griffiths ( 1989 ) for details on the transition dynamics . in this scenario ,
the state - space is more complicated .
let @xmath298 here @xmath299 are the lengths of the distinct sequences , and the ordering constraint notes that the discovery of a new site is added to the beginning of the segment vector . in addition , let @xmath300 then @xmath301 there are three trans - dimensional aspects to the state - space ; the time to simulate @xmath64 sequences ; the number of distinct sequences and the respective lengths of the distinct sequences ( which is determined in part by the first two aspects ) . for the infinitely - many - sites model , we will use the idea of the first time the number of segregating sites is @xmath136 ( or mutations here ) to stop the simulations backward in time . in a similar manner to section [ sec : stopcoal ] , it can be established that we want the approximating function @xmath302 to be the marginal of the process at the last time we have @xmath136 segregating sites . in the context of the infinitely - many - sites model , the bias is controlled by our ability to approximate this marginal ( see remark 2 in section [ sec : contrbias ] ) .
this is because the markov transitions can only change the multiplicity of counts , or increase the number of distinct sequences ; we are unable to change the beginning of sequences . as a result
, it is not possible to establish conditions such as ( a[hyp : p_nassump ] ) .
we propose the following approximation of the marginal , based upon the theoretical properties of such models ( ethier & griffiths , 1987;griffiths , 1989 ) and the relation to the infinitely - many - alleles model ( e.g. griffiths ( 1979 ) and the references there - in ) .
let us consider the marginal distribution , call it @xmath303 .
we extend the state - space to include uncertainty on @xmath42 , the number of distinct types , @xmath304 and @xmath305 the number of segregating sites , and adopt the decomposition @xmath306 now , under certain conditions , there are results about the exact densities @xmath307 ( ewens , 1972 ) and @xmath308 ( watterson , 1975 ) . in the case
@xmath309 , as noted by griffiths ( 1979 ) , for large populations ( such that diffusion results can apply ) the infinitely - many - sites and infinitely - many - allele frequencies are not too different .
therefore , we propose to use the probability ( as in ewens ( 1972 ) ) @xmath310 with @xmath311 are stirling numbers of the first kind . for the quantities @xmath312 and @xmath313 , we use approximations . for the former ,
a uniform distribution is adopted @xmath314\bigg)^{-1}\ ] ] where @xmath315 that is , it is a simple task in combinatorics to show that if there are @xmath305 mutations with @xmath316 repetitions of mutations @xmath257 to @xmath305 , ( subject to the constraint that each mutation can only occur at most once in each sequence and that order of allocating a mutation does not matter ) then there are @xmath317 possible sequences ; summing over all the possible multiples yields the desired cardinality of the state - space .
@xmath313 is not known ( except as the marginal of a recursion ( as in ethier & griffiths ( 1987 ) ) ) and is assigned @xmath318 ( poisson ) distribution ( at time @xmath71 ) . in practice
, it may not be possible to evaluate some of these quantities and a further monte carlo simulation / numerical approximation ( for the integral over @xmath305 and the normalizing constant of @xmath319 ) will be required .
that is to say , we set @xmath320 the approximation will be different for every simulated sample .
lee , a. , yau , c. , giles , m. b. , doucet , a. & holmes , c.c .
( 2010 ) . on the utility of graphics cards to perform massively parallel simulation of advanced monte carlo methods , _ j. comp . graph .
_ , ( to appear ) . | in the following paper we consider a simulation technique for stochastic trees .
one of the most important areas in computational genetics is the calculation and subsequent maximization of the likelihood function associated to such models .
this typically consists of using importance sampling ( is ) and sequential monte carlo ( smc ) techniques .
the approach proceeds by simulating the tree , backward in time from observed data , to a most recent common ancestor ( mrca ) .
however , in many cases , the computational time and variance of estimators are often too high to make standard approaches useful . in this paper
we propose to stop the simulation , subsequently yielding biased estimates of the likelihood surface .
the bias is investigated from a theoretical point of view .
results from simulation studies are also given to investigate the balance between loss of accuracy , saving in computing time and variance reduction .
+ * key words * : stochastic trees , sequential monte carlo , coalescent . |
vortices play a fundamental role in a large variety of physical systems , ranging from ordinary fluids over condensed matter to the early universe .
this variety is reflected in the mathematical models used to describe the formation , structure and dynamics of vortices . in fluid dynamics ,
the basic ingredient of the mathematical model is the velocity field of the fluid , and the vortex is a particular configuration of that velocity field . in condensed matter theory , the mathematical models are field theories . the basic field of such models is a complex valued scalar field , and vortices are particular configurations of that field .
one important difference between vortices in ordinary fluids and those in condensed matter systems is that the total vorticity can take arbitrary real values in the former but is quantised in the latter .
nonetheless , it is sometimes possible to establish a precise mathematical connection between field theory models and those used in ordinary fluid dynamics . in this paper
we are concerned with vortices in the ginzburg - landau model . for general background and references
we refer the reader to the book @xcite .
we are interested in the dynamics of such vortices in the plane , so the basic field of the model is the function : ^2 , depending on time @xmath0 and spatial coordinates @xmath1 .
sometimes we also use polar coordinates @xmath2 in the plane and parametrise the field @xmath3 in terms of its argument and modulus via @xmath4 , where @xmath5 . in this paper @xmath3
will be used to denote both time - dependent and static configurations . for static configurations
we often suppress the first argument and simply write @xmath6 . for configurations varying with time
we sometimes write @xmath7 for the partial derivative @xmath8
. often we write @xmath9 for @xmath10 , @xmath11 for @xmath12 and so on .
the ginzburg - landau energy is the following functional of @xmath3 : [ glenergy ] e [ ] = 12 |+ 12 ( ||^2 - 1)^2 d^2x .
we impose the boundary condition [ boundary ] _
r|(r,)|=1 , which implies that for sufficiently large @xmath13 the field _ r()= ( r , ) is a well - defined map from the spatial circle of large radius @xmath13 to the unit circle in @xmath14 .
it therefore has an associated integer winding number or degree , which can be computed via the following line integral over the circle @xmath15 of radius @xmath13 [ degform ] [ ] = 1 2i_s^1_rd .
this integer , denoted @xmath16 in the following , is also called the vortex number .
the variational derivative of the ginzburg - landau functional ( [ glenergy ] ) with respect to @xmath17 is [ functionalder ] = -1 2 ( + ( 1- ||^2 ) ) and , because of the reality of @xmath18 , = .
the ginzburg - landau equation is the equation for critical points of the ginzburg - landau energy , so it reads [ lgequation ] -+ ( ||^2 -1)=0 .
a ginzburg - landau vortex is a solution of this equation with non - vanishing vortex number @xmath16 .
it is shown in @xcite that for any configuration of non - vanishing vortex number which attains the limit ( [ boundary ] ) uniformly in @xmath19 the ginzburg - landau energy is necessarily infinite .
the origin of this divergence is not difficult to understand and also explained in @xcite .
the essential point is that the gradient terms in the energy density contain the term @xmath20 which , for @xmath21 , leads to a logarithmic divergence .
since the divergence only depends on @xmath16 and not on any other details of the configuration , it can be removed by introducing a smooth cut - off function _ r(x)=\ { cl 0 & |x|r + 1 & |x|r+r^-1 . and defining the renormalised energy functional e_[]= 12 |- _
r + 12 ( point of view of physics and in numerical investigations .
the point is that the total vortex number @xmath16 is conserved during the time evolution .
when studying the interacting dynamics of @xmath16 vortices during a finite time interval we can choose @xmath13 so large that all the vortices remain well inside the disc of radius @xmath13 during that time interval .
the renormalisation procedure removes a divergence which only depends on the conserved quantity @xmath16 but not on any other details of the dynamics . in studying the dynamics of vortices
one has a choice of several different ways of extending the static ginzburg - landau equation to a time - dependent evolution equation .
we are interested in the following first - order equation of schrdinger type , often called the gross - pitaevski equation : [ gpequation ] i = -+ ( ||^2 -1 ) .
the vortex dynamics dictated by this equation has been studied in a large number of publications . in two seminal papers @xcite neu showed that in a scaling limit where the vortex size shrinks to zero , the time evolution according to the _ partial _ differential equation ( [ gpequation ] ) reduces to a set of coupled _ ordinary _ differential equation for the centres of vorticity .
moreover , he showed that set of ordinary equations to be the kirchhoff - onsager law for the motion of fluid vortices in incompressible , nonviscous two - dimensional flows .
neu s work has in turn inspired a number of authors .
some , mathematically motivated , have investigated the scaling limit further and have put his work on a more rigorous mathematical footing , see @xcite and @xcite .
others , starting with the interpretation of neu s work as a finite - dimensional approximation to the infinite - dimensional dynamical system governed by the gross - pitaevski equation , have tried to go beyond this approximation .
physically , one may think of neu s approximation as a point - particle approximation to vortex dynamics in a field theory .
this approximation is expected to be reasonable when the vortices are well - separated and moving slowly . however , when the vortices overlap , the point - particle approximation is poor .
similarly , when the vortices move rapidly we expect them to excite radiation in the field theory which is not captured by the point - particle approximation .
more recently , ovchinnikov and sigal have derived neu s approximation from a different point of view and have computed leading radiative corrections to it in a series of papers @xcite .
they work with the lagrangian from which the gross - pitaevski equation can be derived and explicitly study the truncation of the infinite dimensional gross - pitaevski dynamical system to a finite - dimensional family of multi - vortex configurations with pinned centres of vorticity .
the induced lagrangian on that finite dimensional family reproduces the kirchoff - onsager law when the vortex centres are well - separated .
moreover , the approach makes it possible to study the coupling between the vortex motion and radiative modes and to compute radiative corrections .
however , ovchinnikov and sigal s family of pinned multi - vortex configurations is less useful when trying to understand the dynamics of overlapping vortices .
when vortices get close together pinning the vortex centres is mathematically awkward and physically unnatural . in this paper
we study the truncation of the gross - pitaevski dynamical system to a finite dimensional dynamical system using a different set of multi - vortex configurations .
our configurations are similar to those of ovchinnikov and sigal when the vortices are well - separated , but we believe they provide a more accurate description of vortex dynamics when the vortices overlap .
our approach is inspired by an approximation scheme proposed by manton @xcite in the context of lagrangian soliton dynamics .
manton considered the problem of defining a smooth finite - dimensional family of multi - soliton configurations which could be used as the configuration space for a finite - dimensional low - energy approximation to multi - soliton dynamics .
his proposal is to consider an auxiliary evolution equation , namely the gradient flow in the potential energy functional , and to use the unstable manifold of a suitable saddle point as the truncated configuration space .
such an unstable manifold is the union of paths of steepest descent from the saddle point . in practice
this scheme is not easy to implement , and so far it has only been used to study soliton dynamics in one spatial dimension @xcite . in this paper
we show that it is very well suited to studying the dynamics of overlapping ginzburg - landau vortices .
the basic reason is that in the ginzburg - landau model there are well - known rotationally symmetric saddle point solutions for all vortex numbers @xmath22 which can be used for the construction of the unstable manifold .
we focus on the case @xmath23 and find that the relevant unstable manifold is two - dimensional .
we describe its geometry , compute the induced lagrangian and use it to study the relative motion of two vortices at arbitrary separation .
for our study of the gross - pitaevski equation it is essential that we can derive it from a lagrangian .
the required lagrangian is the following functional of time - dependent fields @xmath3 [ gplag ] l[]=t[,]-e [ ] , where the kinetic energy functional @xmath24 is given by t(,)=-12 ( | ) d^2x and the potential energy functional @xmath18 is the ginzburg - landau functional ( [ glenergy ] ) .
one checks that , essentially because of the linearity of the kinetic energy in @xmath25 , the total energy or hamiltonian is equal to the potential energy @xmath18 . using the formula ( [ functionalder ] ) , the euler - lagrange equation of @xmath26 _ t=-2i ,
is readily seen to be the gross - pitaevski equation ( [ gpequation ] ) .
we define the configuration space to be the space of smooth fields @xmath3 satisfying the boundary condition ( [ boundary ] ) , i.e. [ confspace ] c=\{:^2| _ |x| |(x)| = 1}. as mentioned earlier , the boundary condition means that every @xmath27 has an associated integer winding number or degree .
using stokes s theorem and taking the radius @xmath13 to infinity in ( [ degform ] ) one shows that the degree can also be written as the integral [ ] = _ ^2 , where the integrand [ vorticity ] = d|dis called the vorticity .
the integer degree can not change under the continuous time evolution , which can therefore be restricted to one of the topological sectors _
n=\{|[]=n}. for each @xmath16 , the pair @xmath28 is an infinite dimensional lagrangian dynamical system . from the lagrangian viewpoint @xmath29
is the configuration space and a motion is a path @xmath30 in @xmath29 which satisfies the evolution equation ( [ gpequation ] ) .
alternatively , we can adopt the hamiltonian point of view
. then @xmath29 should be thought of as the phase space and the kinetic part of the lagrangian ( [ gplag ] ) defines a symplectic structure on @xmath29 .
this is a two - form @xmath31 on the tangent space of @xmath29 . for a more careful discussion of that tangent space
we refer the reader to @xcite , but for our purpose it is sufficient to think of tangent vectors as complex valued functions which tend to zero at infinity . if @xmath32 is such a tangent vector we use the notation introduced in @xcite for elements of the complexified tangent space : [ vectnot ] = . if @xmath32 and @xmath33 are two elements of the same tangent space the symplectic form is defined via [ symform ] ( , ) = ( |-| ) d^2x .
it follows that the poisson bracket of @xmath3 and @xmath34 is [ poisson ] \
{ , |}=2i .
one checks that the gross - pitaevski equation can then be written in the canonical form = \{e[],}. as an aside we point out that the total vorticity or degree is related to the pull - back of the symplectic form @xmath31 evaluated on the vector fields @xmath9 and @xmath35 on @xmath36 : ^*(_1,_2 ) = -_^2 .
thus if @xmath37 then @xmath38 .
the lagrangian @xmath26 has a large invariance group .
writing @xmath39 for spatial rotations and @xmath40 for translations in the plane , elements @xmath41 of the euclidean group e_2= so(2)^2 act on fields via pull - back ( r , d)^-1 , i.e. @xmath42 .
both the lagrangian and the degree are invariant under this action .
similarly , unit complex numbers @xmath43 act on fields @xmath3 as phase rotations e^iand leave both the lagrangian and the degree invariant . spatial reflections about any axis leave the lagrangian invariant but change the sign of the degree .
the same is true for the combination of complex conjugation with time reversal @xmath44 .
however , with @xmath45 , the combination [ refsym ] c:|(st ) leaves the degree and the lagrangian invariant . to sum up , for each @xmath46 the group [ symgroup ] c(e_2u(1 ) )
acts on @xmath29 preserving the lagrangian @xmath26 and hence the equations of motion .
the two - dimensional galilean group provides an additional more subtle symmetry . if @xmath47 solves the gross - pitaevski equation , then so does the configuration [ galileo ] _
v(t , x)=e^i(1 2 xv -1 4 v^2t)(t , x - vt ) , where the parameter @xmath48 is physically interpreted as the boost velocity . to end this section we note the conservation laws which follow from the symmetry group ( [ symgroup ] ) .
the noether charge which is conserved due to the invariance under phase rotations is [ qcharge ] q[]=_^2||^2 d^2x .
invariance under spatial rotations leads to the conserved charge [ jcharge ] j[]=_^2 ( | _ ) d^2x and invariance under translations leads to the conservation of the vector @xmath49 with components p_i[]=_^2 ( |_i ) d^2x i=1,2 .
all of the above charges have to be handled with care because the integrals defining them do not generally converge for configurations with non - vanishing degree . in @xcite
a related field theory was studied and it was pointed out that the noether charges can be related to moments of the vorticity . in our cases the moments [ jmoment ]
j [ ] = -_^2 ( x_1 ^ 2+x_2 ^ 2 ) and [ pmoment ] p_1 [ ] = 2_^2 x_2 p_2 [ ] = -2_^2 x_1 are also conserved during time evolution according to ( [ gpequation ] ) and can be obtained from the noether charges @xmath50 , @xmath51 and @xmath52 by integration by parts .
for some configurations the integrals defining @xmath53 and @xmath54 are convergent even when those defining @xmath50 and @xmath55 are not .
the galilean symmetry also implies a conservation law , but it will not be required in this paper .
imposing extra symmetry is the easiest way of finding static solutions of the ginzburg - landau equations .
the largest invariance group one can impose on configurations of degree @xmath16 is a group containing reflections and combinations of spatial rotations @xmath56 by an angle @xmath57 with suitable phase rotations .
more precisely we define the subgroup [ sadsym ] r_n = c \ { ( r(),e^in ) so(2)u(1)|[0,2 ) } of the symmetry group ( [ symgroup ] ) .
configurations invariant under this group are of the form _ n(r,)=f_n(r)e^in , where @xmath58 is real and satisfies the boundary condition [ unique ] f_n(0)=0 n0 _ rf_n(r)=1 .
the ginzburg - landau equation implies the following ordinary differential equation for @xmath58 ( r ) -f_n + ( 1-f_n^2)f_n=0 . as proved for example in @xcite , this equation has a unique solution satisfying the boundary conditions ( [ unique ] ) for each @xmath16 , and in the following we use @xmath58 to denote that solution . near @xmath59 and for large @xmath60 the equation
can be solved approximately f_n(r)&~ & a_n r^n r , + f_n(r)&~ & 1- r , where @xmath61 are real constants .
however , no exact expression for @xmath58 in terms of standard functions is known .
for @xmath62 the resulting vortex configuration is a stable solution of the ginzburg - landau equation . for @xmath63
the solution is an unstable saddle point .
bounds on the number of unstable modes were derived by ovchinnikov and sigal in @xcite .
it follows from their analysis that for @xmath23 there is precisely one unstable mode .
since this mode plays an important role in our analysis , we describe and compute it explicitly . for further general background
we refer the reader to @xcite .
the required unstable mode is an eigenvector of the hessian of the ginzburg - landau energy e[]=. for the saddle point @xmath64 the hessian takes the form e[_2]= 1 2 , where @xmath65 . in terms of the notation ( [ vectnot ] ) the eigenvalue equation [ hesseigen ] 2e[_2]= is thus equivalent to [ fluctuate ] -+ ( 2_2 - 1)+_2 e^4i |= , which can also be obtained by linearising the ginzburg - landau equation around the saddle point @xmath64 .
if @xmath32 solves ( [ fluctuate ] ) then the leading term in the difference @xmath66 - e[\psi_2]$ ] is the second order term e^(2)()&=&1 2 ( ) ^ e[_2 ] + & = & , , where we used the inner product [ metric ] , = 1 2 ( | + |)d^2x . to study the eigenvalue problem ( [ fluctuate ] ) and to compute the variation in the energy it is best to expand the @xmath19-dependence of @xmath32 into a fourier series .
the term @xmath67 in the equation leads to a coupling between fourier modes @xmath68 with @xmath69 differing by @xmath70 , but all other modes decouple .
as explained in @xcite , the unique eigenfunction with a negative eigenvalue is of the form [ eigenfun ] ( r , ) = u(r ) + v(r)e^4i . inserting this expression into ( [ fluctuate ] ) leads to coupled equations for the radial functions @xmath71 and
@xmath72 [ coupled ] -1 r ( r ) + ( 2_2 -1)u + _ 2 v & = & u + -1 r ( r ) + ( + 2_2 - 1)v + _ 2 u & = & v . in terms of @xmath71 and
@xmath72 the second variation of the ginzburg - landau energy for a fluctuation @xmath32 of the form ( [ eigenfun ] ) about the stationary point @xmath64 is [ energyvar ] e^(2 ) ( ) = _ 0^r dr , where we have assumed regularity of @xmath71 and @xmath72 at the origin and exponential decay at infinity in the integration by parts .
this assumption will be justified below .
note that , in terms of the functions @xmath71 and @xmath72 , , = 2_0^ ( u^2 + v^2)r dr .
the eigenvalue ( [ coupled ] ) was first studied numerically by dziarmaga in @xcite , using a shooting method .
near @xmath59 the leading terms for a regular solution are u(r)~1-r^2 v(r)~a r^4 , where @xmath73 is an undetermined parameter . for large @xmath60 , it is convenient to write the equation in terms of the linear combinations @xmath74 and @xmath75 : [ pmcoupled ] -1 r ( r ) + ( + 3 - 1)u_+ u_-&= & u_+ + -1 r ( r ) + ( + - 1 ) u_- + u_+ & = & u_- .
now the equations decouple for large @xmath60 and one deduces the asymptotic form u_+&~ & a + b + u_-&~ & c + d. solving the eigenvalue problem with a shooting method means determining the two parameters @xmath73 and @xmath76
so that the solutions @xmath77 and @xmath78 both decay exponentially for large @xmath60 i.e. @xmath79 .
such a parameter search in a two - dimensional parameter space is tricky .
we start our search by assuming that @xmath72 is negligible compared to @xmath71 and take the trial function @xmath80 where u_0(r ) =( r),one then finds that e^(2 ) ( ) = -0.3781 2 _ 0,_0 , showing that @xmath81 and hence that the value of @xmath82 for the eigenvalue given in @xcite is incorrect . in our search , starting with @xmath83 , we find the eigenvalue = -0.41869 , with eigenfunctions @xmath71 and @xmath72 shown in figure 1 .
note that @xmath71 is qualitatively very similar to @xmath84 .
however , @xmath72 is non - vanishing and negative .
this is not surprising .
looking at the energy expression ( [ energyvar ] ) we note that it is energetically favourable for the functions @xmath71 and @xmath72 to have opposite signs .
=3 in [ fig1 ]
the unstable manifold of a saddle point is the union of paths of steepest descent from the saddle point . to define paths of steepest descent ,
one requires both a metric and a potential .
suppose @xmath85 are local coordinates on a finite - dimensional manifold with potential energy @xmath86 and the metric is represented by the matrix @xmath87 .
then the paths of steepest descent are solutions to the gradient flow equations g_ij = - , with @xmath88 approaching the relevant saddle point as @xmath89 .
it is important to keep track of the metric even when it is flat .
for example , gradient flow on @xmath36 with the standard flat metric takes the following form in complex coordinates @xmath90 : = -2 . in the field theory we are considering here
the potential energy is the ginzburg - landau potential energy functional ( [ glenergy ] ) and the metric is the flat metric .
since we work with complex - valued fields , tangent vectors are also complex - valued functions and the metric is the inner product of tangent vectors defined in ( [ metric ] ) .
the gradient flow equation is the following non - linear heat equation [ heat ] _ = -2 = + ( 1- we distinguish notationally from the time variable @xmath91 used in the gross - pitaevski equation .
we have solved this equation numerically on a grid of size @xmath92 , corresponding physically to the square @xmath93\times [ -20,20]$ ] in the @xmath94-plane . to generate the gradient flow curve we start with the saddle point configuration @xmath64 and add a small perturbation @xmath95 , where @xmath96 is the negative eigenmode ( [ eigenfun ] ) found in the previous section . at the boundary of the lattice
we impose the dirichlet boundary condition @xmath97 .
the resulting gradient flow curve does not depend significantly on @xmath98 provided it is small enough .
when @xmath99 , the discretisation effects provide sufficient perturbation to make sure that the gradient flow curve moves away from the saddle point @xmath64 , but the initial flow is very slow .
it is therefore numerically convenient to use a non - vanishing value of @xmath98 . in figures 2 and 3
we show snapshots of the field and of the energy density during the early stage of the gradient flow , generated by starting with @xmath100 . during the gradient flow ,
the unstable @xmath23 vortex splits into two @xmath101 vortices which drift apart .
the saddle point solution @xmath64 has an energy density which has a degenerate maximum on a ring .
the splitting process begins with the development of two local maxima , as shown in configuration a in figure 3 .
as the splitting process continues , these local maxima become two distinct vortices .
note , however , that during the early phase of the splitting process , the vortices overlap and deform each other significantly .
this is clearly visible in configuration b in figure 3 . only in the later stages of the splitting process
do the vortices resemble two standard @xmath101 vortex solutions .
= 2.3 in [ fig2 ] = 3.0 in [ fig3 ] in our numerical simulation of the gradient flow we generated 700 configurations . the most convenient way of labelling these configurations is in terms of their ginzburg - landau energy rather than by the configuration number or the gradient flow `` time '' @xmath102 , which have no physical significance in the gross - pitaevski model .
the finite lattice used in the computation provides a natural renormalisation of the energy of the vortex configurations , assigning the value @xmath103 = v_{\mbox{\tiny max}}=31.982 $ ] to the saddle point configuration @xmath64 . the configuration a in figures 2 and 3 has the energy @xmath104 and the configuration b has the energy @xmath105 .
there is another way of parametrising the gradient flow which we need to address , namely the parametrisation in terms of vortex separation .
vortex separation is used as a parameter for pinned vortex configurations in the work of ovchinnikov and sigal . in view of our later comparison with that work
we therefore make a few comments on it here .
the pictures of the vortex configurations and energy densities shown in figures 2 and 3 illustrate that the notion of a separation is not well - defined during the early stages of the gradient flow .
the best one can do is to define a functional of vortex configurations which reproduces the distance between the maxima of the energy density ( or , equivalently , between the zeros of the field @xmath3 ) for well - separated vortices and which is computationally convenient . in the following we use the vorticity ( [ vorticity ] ) instead of the energy density .
the vorticity looks qualitatively very similar to the energy density during all stages of the gradient flow and in particular it is strongly peaked around the zeros of the field @xmath3 for well - separated vortices .
the advantage of using the vorticity rather than the energy density is that the total vorticity is conserved whereas the total energy changes during the gradient flow . to motivate our approach further ,
note that the configurations generated during the gradient flow break the invariance group @xmath106 of the saddle point to the group @xmath107 generated by reflections about the @xmath108 and @xmath109 axes , each combined with the complex conjugation of @xmath3 . as a result the vorticity @xmath110 of all configurations is reflection symmetric about two orthogonal axes . for our gradient flow the reflection axes are the coordinate axes , and the vortices separate along the @xmath109-axis because we have broken the rotational symmetry @xmath106 of the saddle point solution explicitly by using the perturbation @xmath32 ( [ eigenfun ] ) .
a different perturbation would have led to an isomorphic reflection symmetry about a different set of orthogonal axes , and a different separation direction .
we use the reflection axis orthogonal to the separation direction to divide @xmath36 into two half spaces @xmath111 and @xmath112 .
then we define the separation functional as [ distdef ] a[]= _ h_r x_2 d^2x -_h_l x_2 d^2x= 2_h_r x_2 d^2x .
as expected , the separation functional ( [ distdef ] ) increases monotonically during the gradient flow , but one interesting feature is that it assigns a non - vanishing value to the `` coincident '' vortex configuration @xmath64 .
this is a familiar feature of separation parameters used for other solitons , such as monopoles in non - abelian gauge theory @xcite .
the point is that solitons lose their separate identity when they overlap and separation ceases to be a meaningful concept . in particular
there is no reason to insist that a separation parameter vanish for a particular configuration .
numerically , we find the minimal separation to be @xmath113=4.815 $ ] , which corresponds roughly to the diameter of the ring on which the energy density of @xmath64 is maximal . for the configuration a in figures 2 and 3
we find @xmath114 and for configuration b we find @xmath115 .
the lowest energy configuration we generated has the energy @xmath116 and separation @xmath117 . for the construction of the unstable manifold we return to the parametrisation of the gradient flow curve in terms of renormalised energy .
thus we obtain a family of fields @xmath118 , labelled by the value @xmath86 of the renormalised ginzburg - landau energy . according to the general prescription of @xcite
we should now act with the full symmetry group ( [ symgroup ] ) to generate a family of saddle points and a family of gradient flow curves , and use their union as the collective coordinate manifold for the truncated dynamics .
however , we are only interested in the relative motion of the two vortices in the fission process and not in their centre - of - mass motion .
the latter can be found by applying galilean boosts ( [ galileo ] ) to the entire configuration .
furthermore , we exclude collective coordinates which change our boundary condition .
both phase rotations and spatial rotations change the field `` at infinity '' ( or at the boundary of our lattice ) , but the subgroup @xmath106 defined in ( [ sadsym ] ) respects the boundary condition .
it also leaves the saddle point solution @xmath64 invariant but it acts non - trivially on the gradient curve emanating from @xmath64 .
thus we obtain a family of fields [ thefamily ] ( v,;r,)= ( v;r,-)e^2i depending on two collective coordinates @xmath119 .
the former measures the energy and the second labels the spatial orientation of the two - vortex configurations generated during the gradient flow . since the saddle point configuration with energy @xmath120 does not depend on the angle @xmath57 , the collective coordinate manifold is topologically a plane , with @xmath121 as polar coordinates .
we denote this manifold by @xmath122 .
the range of the angle @xmath57 is @xmath123 since the configurations @xmath118 generated during the gradient flow are invariant under a rotation by @xmath124 ( which is the product of the reflections at the coordinate axes discussed above ) .
note that @xmath125 .
it is now a simple matter to compute the restriction of the gross - pitaevski lagrangian ( [ gplag ] ) to @xmath122 .
we allow the collective coordinates @xmath86 and @xmath57 to depend on time and insert the fields @xmath126 into ( [ gplag ] ) .
the result is [ indlag ] l[(v,;)]= - 12 ( 2q(v)-j(v ) ) - 12g(v)v - v , where we have written @xmath127 and @xmath128 for the noether charges ( [ qcharge ] ) and ( [ jcharge ] ) evaluated on @xmath129 , i.e. q(v)=|(v,;x)|^2 d^2x and j(v)= ( @xmath86 are both divergent , but the combination [ kfunctional ] k(v)= j(v)-2q(v ) occuring in the lagrangian is finite . to see this note that for the saddle point configuration @xmath64 one computes @xmath130 , so that @xmath131 . as the configuration evolves away from the saddle point @xmath132 grows but it remains finite for any finite value of @xmath86
this follows from the continuity of the semigroup defined by the evolution equation [ heat ] in a suitable norm , see @xcite and @xcite , and from the continuity of the funtionals @xmath50 and @xmath133 ( [ qcharge ] ) and ( [ jcharge ] ) with respect to that norm .
finally , the function @xmath134 is [ gcharge ] g(v)= ( |(v,;x)(v,;x ) ) d^2x .
neither @xmath132 nor @xmath135 depends on @xmath57 because of the rotationally invariant integration measure and because the integrands are independent of the phase of @xmath136 .
the term @xmath137 is a total time derivative which does not affect the equations of motion .
we therefore omit it in our final expression for the induced lagrangian @xmath138 on @xmath122 : l_2 = 12 k(v ) -v .
the euler - lagrange equations are very simple because our truncated system , like the original gross - pitaevski dynamics , is rotationally invariant and has a conserved energy .
thus @xmath86 remains constant during the time evolution , and the angle @xmath57 changes according to [ chidot ] ( v ) = 2 k(v ) .
it follows from the constancy of @xmath86 that @xmath139 is also constant during the time evolution .
we introduce the abbreviation ( v ) = ( v ) for the rate of change of the angle @xmath57 .
the dependence of this angular velocity on @xmath86 is the main dynamical information we extract from our truncated dynamics .
it is is given by the differential quotient [ angvel ] ( v ) = 2 .
numerically , we compute @xmath140 by evaluating finite difference quotients @xmath141 of successive configurations .
our results are displayed in figure 4 . in the early phase of the gradient flow ( roughly the first 100 configurations )
both @xmath142 and @xmath143 change very slowly and accurate numerical computation of the difference quotient is difficult .
we have omitted these difference quotients from our plot .
since the energy @xmath86 hardly changes during this part of the gradient flow , this omission makes no visible to difference to the graph @xmath144 shown in figure 4 .
the next section is devoted to a detailed discussion and interpretation of our results .
=3 in [ fig4 ]
our two - dimensional lagrangian dynamical system @xmath145 is designed to approximate the slow interactive dynamics of two vortices moving according to the gross - pitaevski equation ( [ gpequation ] ) .
the model predicts that the two vortices will orbit each other , with the angular frequency depending on their total energy as shown in our figure 4 .
while the qualitative behaviour of the two vortex system is not surprising and familiar from vortices in ordinary fluids and also vortices in the gauged ginzburg - landau model @xcite , the precise prediction of the variation of the angular frequency with the energy for both overlapping and well - separated vortices is new . in @xcite dziarmaga
computed the rotation frequency of an overlapping vortex pair by linearising the gross - pitaevski equation around the saddle point solution @xmath3 .
the linearised solution is [ linsol ] ( t , x ) = _ 2(x ) + e^-it ( u(r ) + e^4iv(r ) ) , where @xmath76 and @xmath146 are the eigenvalue and eigenfunctions of the eigenvalue problem ( [ coupled ] ) .
the interpretation of the linearised solution in terms of rotating vortices requires some care . as we saw in section 3 ,
the radial eigenfunction @xmath72 is much smaller than @xmath71 .
thus , as a first approximation , we may think of ( [ linsol ] ) as a field whose magnitude depends on @xmath60 according to @xmath147 and whose direction changes with angular frequency @xmath76 , all superimposed on the static @xmath23 solution .
the effect of such a superposition is a configuration with two zeros ( whose separation depends on @xmath98 ) which rotate around one another with angular velocity @xmath148 .
the factor @xmath149 arises because for @xmath64 a phase rotation by @xmath150 is equivalent to a spatial rotation by @xmath151 . with our value for @xmath76
, we thus find @xmath152 .
in the limit @xmath153 our numerically computed function @xmath144 approaches a value which is close to but not exactly equal to @xmath154 .
the discrepancy arises because we neglected the radial eigenfunction @xmath72 in our interpretation of the linearised solution .
we can establish a precise relation between the eigenvalue @xmath76 and the limiting angular frequency [ maxfre ] _ : = _ vv _ ( v ) as follows . using the definition ( [ kfunctional ] ) and noting that both @xmath86 and @xmath132 are originally defined along the gradient flow curve parametrised in terms of the auxiliary variable @xmath102 ( [ heat ] ) we write ( v ) = 2 = 2 /.
recall that we computed our gradient flow curve by starting at @xmath155 with the configuration @xmath156 , where @xmath32 is the eigenmode ( [ eigenfun ] ) and @xmath98 is small ( in practice @xmath157 ) . at that point
the gradient of the ginzburg - landau energy functional is [ _ 2 + ] & & + | + & = & 1 2 , where we have used that the first functional derivative of @xmath18 vanishes at @xmath64 and that @xmath32 is an eigenfunction of the hessian , i.e. it satisfies ( [ hesseigen ] ) .
hence from ( [ heat ] ) [ useful ] _ _ |_=0- _ |_|_=0 -| and thus .|_=0 & = & [ _ 2 + ] _ _ |_=0 + [ _ 2 + ] _
|_|_=0 d^2x + & & -^2 ^ 2|d^2x + & = & -^2 ^ 2 2 ( u^2 + v^2)rdr , where we used ( [ useful ] ) and inserted the expression ( [ eigenfun ] ) for @xmath32 .
it is easy to check that any @xmath106-invariant configuration ( [ sadsym ] ) and hence in particular @xmath64 is a stationary point of the functional @xmath158 $ ] ( [ kfunctional ] ) .
thus @xmath159 and by a similar calculation to the one given above we have .|_=0 & = & [ _ 2 + ] _ _ |_=0 + [ _ 2 + ]
_ |_|_=0 d^2x + & & -^2| + | + & = & -2 ^ 2(|_-2| ) d^2x + & = & 4 ^22 ( u^2 -v^2)r dr , where we used that @xmath160 .
the approximate equalities become exact in the limit @xmath161 .
thus we find the following exact expression for the angular frequency @xmath162 ( [ maxfre ] ) in terms of the eigenvalue @xmath76 : [ goodformula ] _ = -2 . in view of the numerical difficulties in computing @xmath140 near the saddle point
this formula is very useful in practice . evaluating it numerically we find [ omval ] _
= -2 1.0200 = 0.21354 .
this value provides a check on our computation of @xmath144 near the saddle point and is marked with a `` @xmath163 '' in figures 4 and 5 . for well - separated vortices we can compare our results with those obtained by neu in @xcite via scaling techniques or those obtained by ovchinnikov and sigal in @xcite based on pinned vortex configurations .
both find that well - separated vortices are governed by the kirchhoff - onsager law . according to that law two vortices separated by a distance @xmath73 have an interaction energy given by [ endist ] v(a ) = -2a + c , where @xmath164 is an arbitrary normalisation constant , and orbit each other with angular frequency [ simplelaw ] _
it follows that the rotation frequency is given as a function of the energy via [ kolaw ] _ ( v ) = c e^v , where @xmath165 .
the constant @xmath164 ( and hence @xmath166 ) is determined once the dependence of the energy on the vortex separation is known . to make contact with the kirchhoff - onsager relations we thus have to make use of the separation functional introduced for our vortex configurations in ( [ distdef ] ) .
we pick a vortex configuration consisting of clearly separated vortices and compute its energy to be @xmath167 and its separation to be @xmath168 .
inserting these values into ( [ endist ] ) fixes the value of @xmath164 and hence @xmath166 .
we have plotted the resulting function ( [ kolaw ] ) in figure 4 .
it is important to be clear about the interpretation of the two curves in figure 4 . for well - separated vortices we can meaningfully associate a separation parameter to a vortex configuration .
the kirchhoff - onsager relation gives the angular frequency of point - vortices with that separation .
the fact that for low energy configurations consisting of well - separated vortices the kirchhoff - onsager relation agrees with our results shows that well - separated vortices may be treated as point - vortices , in agreement with the results of ovchinnikov and sigal in @xcite .
the two curves in figure 5 differ for configurations near the saddle point , not so much because the predictions of our model disagree with those of the point vortex model but because the two models are no longer compatible in that regime . from the point of view of the kirchhoff - onsager relation ,
the dependence ( [ simplelaw ] ) of the angular velocity on separation is more fundamental than the dependence on the energy . in figure 5
we plot the angular velocity @xmath140 as a function of the separation parameter ( [ distdef ] ) and compare it with the simple kirchhoff - onsager prediction . again
the two plots agree for large separations , and the comments made above on the comparison near the saddle point apply here , too .
however , it is worth emphasising that the kirchhoff - onsager relation predicts a divergence of the angular velocity when the separation tends to zero .
our results show that for the ginzburg - landau vortices this divergence is removed , essentially by cutting off small separation parameters .
this is yet another example of a soliton model `` regularising '' the singularities of a point - particle model .
=3 in [ fig5 ]
the unstable manifold method used in this paper to approximate the dynamics of interacting vortices in the gross - pitaevski model allows one to analyse the relative motion of two vortices at arbitrary separation .
when the vortices are overlapping our results reproduce those found with linearisation methods and when the vortices are well - separated our results agree with the kirchhoff - onsager law .
our method gives a unified description of both regimes , and successfully interpolates between them .
when vortices evolve according to the gross - pitaevski equation they will in general also emit radiation .
this effect is not captured in our approximation .
however , it is possible to compute radiative corrections to the vortex motion predicted by our approximation using the methods developed by ovchinnikov and sigal in @xcite .
in particular , ovchinnikov and sigal studied the radiative corrections to the orbiting motion of a vortex pair and found that , for large vortex separation @xmath73 , the power emitted in the radiation is proportional to @xmath169 . as a result of this energy loss , the orbiting vortices drift apart slowly as they orbit each other .
however , since the power loss is small , the change of the separation during one period of rotation is also small ( much less than 1 % ) . combining these results with our calculations , we arrive at the following qualitative picture of vortex fission in the gross - pitaevski model .
suppose the @xmath23 saddle point solution is perturbed and allowed to evolve according to the gross - pitaevski equation .
the perturbed vortex configuration rotates and emits radiation .
the radiation carries away energy and the vortex configuration adjusts by finding a configuration of lower energy . the most efficient way of doing so is by following a path of steepest ( energy ) descent .
as we have seen , the @xmath23 vortex breaks up into two @xmath101 vortices along the path of steepest descent .
thus we can use our truncated dynamical system @xmath145 and expect the vortex pair to orbit with the angular frequency depending on the energy according to ( [ chidot ] ) .
the continued rotation results in further emission of radiation and loss of energy , to which the vortex pair adjusts by drifting further apart .
the fission process is thus governed by the combination of two effects : the radiation carries away energy and the vortex configuration adjusts by sliding down the unstable manifold of the @xmath23 saddle point .
it would be interesting to make this picture more precise by studying the radiation effects quantitatively .
this should probably be done in conjunction with a more careful study of the conservation laws in the theory , similar to the investigation for gauged vortices in @xcite .
as mentioned briefly in section 2 , the conservation of angular momentum can also be expressed as the conservation of the moment ( [ jmoment ] ) of the vorticity .
just looking at vortex motion during the fission process it seems that the moment ( [ jmoment ] ) increases as the vortices drift apart
. it would be interesting to understand how the radiation makes up for this change , possibly by carrying negative vorticity off to infinity .
we would like to emphasise that the work described here is the first application of the unstable manifold method to soliton dynamics in more than one spatial dimension .
as such , it contains a number of useful general lessons . at first sight
it may not seem sensible to use one non - linear partial differential equation ( the gradient flow equation ) to approximate another non - linear partial differential equation ( the gross - pitaevski equation ) .
however , generally speaking , gradient flow is easier to understand and to compute numerically than hamiltonian flow like that defined by the energy - conserving gross - pitaevski equation .
one reason for this is that the typical time scales of the two evolution processes are very different .
our discussion of the vortex fission process illustrates this point .
the solution of the gradient flow equation over a relatively short cpu time ( roughly 700 steps ) maps out all the interesting configurations between the @xmath23 saddle point and the well - separated vortices .
the fission process according to the gross - pitaevski equation , by contrast , is expected to take several orders of magnitude longer , making it difficult to maintain numerical accuracy .
the gradient flow equation provides an efficient way of mapping out those configurations which are relevant for slow ( or low - energy ) dynamics . at a more practical level
, we found that the explicit study of the linear negative modes of the saddle point was essential for an efficient computation of the unstable manifold .
we expect that this will become even more important for saddle points with more than one negative mode .
thus one could repeat the calculations performed here for the saddle point solutions @xmath170 with vortex number @xmath171 . as explained in @xcite
, there is now more than one unstable mode . as a first step in constructing the unstable manifold of these saddle points one needs to compute the negative eigenmodes of the linearised ginzburg - landau equation explicitly .
while the explicit construction of the unstable manifolds and the computation of the induced dynamics may be difficult , it would be interesting to understand general geometrical features of the unstable manifolds and use them to derive qualitative features of multi - vortex dynamics in the gross - pitaevski model .
it remains an open problem to justify the unstable manifold approximation used in this paper analytically .
the fact that our results agree with those obtained via more familiar approximations in limiting regimes provides an encouraging check .
however , this agreement also raises a question . near the saddle
point our method agrees with results obtained by linearisation , and the small parameter controlling the validity of the approximation is the amplitude @xmath98 in ( [ linsol ] ) . as explained , this is related to the separation of the two zeros of the field @xmath3 . far away from the saddle point
, our approximation agrees with the point - vortex model implicit in the kirchhoff - onsager law .
the small parameter controlling the validity of the approximation in this regime is the _ inverse _ separation of the zeros of @xmath3 ( i.e. the inverse distance between the vortices ) .
since our method interpolates between the two approximations it is not clear which , if any , small parameter controls the approximation in the intermediate regime .
more generally , proof of the validity of the approximation , possibly along the lines of the proof of the geodesic approximation for gauged vortex dynamics in @xcite , would be very desirable .
a useful starting point for such an investigation may be the following natural relation between the gradient flow equation ( [ heat ] ) and the gross - pitaevski equation ( [ gpequation ] ) .
both define flows in the space @xmath172 ( [ confspace ] ) in terms of the ginzburg - landau energy .
however , the gradient flow uses the inner product ( [ metric ] ) whereas the gross - pitaevski equation is a hamiltonian flow using the symplectic structure ( [ symform ] ) .
the inner product ( [ metric ] ) and the symplectic form ( [ symform ] ) are equal to , respectively , the real and imaginary part of the sesquilinear form g(,)=_^2| d^2x . one interesting manifestation of this relationship is that gradient flow trajectories and solutions of the gross - pitaevski equation are orthogonal with respect to the inner product ( [ metric ] ) whenever they intersect .
most of the research reported here was carried out while ol was an msc student at heriot - watt university .
ol thanks the daad for a scholarship during that time and dugald duncan for advice on the numerical calculations .
bjs acknowledges an epsrc advanced research fellowship .
99 bethuel f , brzis h and hlein f 1994 _ ginzburg - landau vortices _
( basel : birkhuser ) ovchinnikov y n and sigal i m 1997 ginzburg - landau equation i. static vortices , in partial differential equations and their applications _ crm proceedings and lecture notes _ * 12 * 199220 , eds greiner p et al ( providence : american mathematical society ) neu j c 1990 vortices in complex scalar fields , _ physica _ * d 43 * 385406 neu j c 1990 vortex dynamics of the non - linear wave equation _ physica _ * d 43 * 407420 lin f - h and xin j x 1999 on the imcompressible fluid limit and the vortex motion law of the nonlinear schrdinger equation _ commun . math .
phys . _ * 200 * 249274 lin f - h and xin j x 1999 on the dynamical law of the ginzburg - landau vortices on the plane _ comm .
pure appl .
_ * 52 * 11891212 colliander j e and
jerrard r l 1998 vortex dynamics for the ginzburg - landau - schrdinger equation _ internat .
. notices _ * 7 * 333358 ovchinnikov y n and sigal i m 1998 the ginzburg - landau equation iii . vortex dynamics _
nonlinearity _ * 11 * 12771294 ovchinnikov y n and sigal i m 1998 long - time behaviour of ginzburg - landau vortices _ nonlinearity _ * 11 * 12751310 manton n s 1988 unstable manifolds and soliton dynamics _ phys .
lett . _ * 60 * 19161919 manton n s and merabet h 1997 @xmath173 kinks gradient flow and dynamics _ nonlinearity _ * 10 * 318 papanicolaou n and tomaras t n 1993 on the dynamics of vortices in a nonrelativistic ginzburg - landau model _ phys .
lett . _ * a 179 * 33 dziarmaga j 1996 dynamics of overlapping vortices in complex scalar fields _ act .
phys . pol .
_ * b27 * 19431960 ginibre j and velo g 1996 the cauchy problem in local spaces for the complex ginzburg - landau equation i. compactness methods _ physica _ * d 95 * 191228 ginibre j and velo g 1997 the cauchy problem in local spaces for the complex ginzburg - landau equation ii .
contraction methods _ commun . math .
_ * 187 * 4579 manton n s 1997 first order vortex dynamics _
annals phys .
_ * 256 * 114131 manton n s and nasir s m 1999 conservation laws in a first order dynamical system of vortices _ nonlinearity _ * 12 * 851865 atiyah m and hitchin n 1988 _ geometry and dynamics of monopoles _ ( princeton : princeton university press ) stuart d 1994 dynamics of abelian higgs vortices in the near bogomolny regime _ commun .
phys . _ * 159 * 5191 . | the time evolution of several interacting ginzburg - landau vortices according to an equation of schrdinger type is approximated by motion on a finite - dimensional manifold . that manifold is defined as an unstable manifold of an auxiliary dynamical system , namely the gradient flow of the ginzburg - landau energy functional . for two vortices
the relevant unstable manifold is constructed numerically and the induced dynamics is computed .
the resulting model provides a complete picture of the vortex motion for arbitrary vortex separation , including well - separated and nearly coincident vortices .
nlin.ps/0201047 + hwm01 - 44 24 pt unstable manifolds and schrdinger dynamics + of ginzburg - landau vortices o. lange and b. j. schroers + department of mathematics , heriot - watt university + edinburgh eh14 4as , united kingdom january 2002 ams classification scheme numbers : 35q55 , 37k05 , 70k99 |
since the beginning of the 21st century , the spectroscopy of spin noise @xcite has emerged as a competitive tool to study spin dynamics in close to thermal equilibrium conditions .
the rapid development of this technique has started with detailed investigations of the spin fluctuations in atomic gases in 2004 @xcite , and then moved to studies of spin noise in semiconductor systems , such as bulk crystals @xcite or nanostructures @xcite , because this class of materials may be the building blocks of future spin - based devices .
for many applications , self - assembled semiconductor quantum dots ( qds ) have been considered , because qds can be grown epitaxially and implemented into established semiconductor environment @xcite .
the direct band gap in combination with a giant optical dipole moment allows for spin control operations that can be performed at a terahertz rate or even faster , see , e.g. , @xcite . to that end , it is mandatory for such applications to have detailed knowledge on the interactions of the involved carrier spins and the resulting dynamics .
the spin dynamics of the confined spins of single electrons and holes are mostly determined by the interaction with the surrounding nuclei in the dot , which are theoretically treated usually by the central spin model ( csm ) .
first experimental investigations on the spin noise of singly - charged qds were undertaken by crooker _
et al . _
further studies on such qds included , for example , the observation of the hole spin decoupling from the surrounding nuclei on the fluctuation level @xcite .
another achievement was related with the investigation of hole spin noise correlations probed by two beams in a two - colour optical spin noise technique , making it possible to reveal the homogeneous lifetime of the optical transition in the dot , despite the inevitable inhomogeneous broadening of the qd ensemble @xcite . however , until now the objects of investigation of the spin noise in quantum dot ensembles were , to the best of our knowledge , resident hole spins only .
if at all , electrons were additionally excited by a weak non - resonant pump laser , by which the spin fluctuations of the optically excited electron spins became observable @xcite .
in contrast to the lack of comprehensive experimental investigations of the electron spin noise in @xmath0-doped quantum dots , theoretical treatments do already exist .
two complementary approaches , based on a semiclassical treatment utilizing the bloch equations and averaging over the static nuclear spin fluctuations @xcite , and a fully quantum mechanical approach that is based on the chebyshev polynomial technique @xcite , lead to almost identical predictions of spin noise spectra @xcite .
this work aims at a detailed experimental and theoretical investigation of resident electron and hole spin fluctuations in qd ensembles .
we present experimental results on these fluctuations in absence of external fields , where the electron spin precession around overhauser field fluctuations is revealed .
we also demonstrate comparable time scales for long - term spin decoherence in @xmath0- and @xmath1-type qds .
this result , being unexpected within the simple csm , is explained by the inclusion of nuclear quadrupole interactions @xcite .
further , we study the behaviour of the spin noise under application of transverse and longitudinal static fields . we demonstrate that , like in the case of hole spins @xcite , the electron spins also undergo decoupling from the nuclear spins .
this effect takes place at somewhat stronger longitudinal fields , demonstrating a stronger coupling to the surrounding bath of nuclei .
additionally , we are able to trace the redistribution of the spin noise power between the nuclei - induced carrier spin precession and the quasi - static contributions at zero magnetic field . the paper is organized as follows : sec .
[ sec : theory ] introduces the theoretical approaches to calculate the spin noise ; the description of samples and setup is presented in sec .
[ sec : samples ] .
section [ sec : discussion ] presents the experimental results and their discussion in relation to the theoretical predictions .
conclusions are given in sec .
[ sec : conclusion ] .
in optical spin noise spectroscopy , spin fluctuations are monitored as noise in the faraday or kerr rotation of a continuous - wave ( cw ) linearly polarized laser beam .
the fluctuation angles @xmath2 ( faraday effect , transmission geometry ) and @xmath3 ( kerr effect , reflection geometry ) are proportional to the spontaneous fluctuations of the charge carrier spins .
the autocorrelation functions of the spin signals are proportional to the correlation functions of electron and hole spin-@xmath4 components , @xmath5 and @xmath6 , with the @xmath7 $ ] axis being the direction of the probe beam propagation coinciding with the quantum dot growth axis @xcite .
we consider an ensemble of electronically isolated quantum dots ( qds ) in an external static magnetic field @xmath8 .
each qd can be charged either with an electron or a heavy hole . as a result , it is sufficient to calculate the spin fluctuation of a charge carrier within a dot and afterwards average over the quantum dot distribution . since the qds are independent of each other , the cross - correlations between the spins in different qds are negligible .
further , no correlation between electron and hole spins appears , see refs .
@xcite for details .
hence , we are interested in the frequency spectra of the electron and hole spin fluctuations defined as @xcite : @xmath9 here @xmath10 for electrons and @xmath11 for holes , @xmath12 is the quantum - mechanical operator of the cartesian component @xmath13 of the carrier spin , the angular brackets denote the quantum mechanical and statistical averages , and @xmath14 stands for the symmetrized product of the operators . in accordance with the general theory @xcite
the spectrum is an even function of @xmath15 . as a result , for a qd ensemble one obtains the following expression for the faraday rotation fluctuation spectrum : @xmath16 where @xmath17 and @xmath18 are constants which depend on the details of the studied sample , the radiation propagation geometry and the probe frequency @xcite .
furthermore , @xmath19 and @xmath20 are the numbers of negatively and positively charged qds within the probe laser spot , respectively .
the hamiltonian of a single qd is given by @xmath21 here @xmath22 is the frequency of the carrier spin precession about the external magnetic field applied in the @xmath23 plane , @xmath24 , @xmath25 denotes the unit vector along the @xmath26-axis , and @xmath27 and @xmath28 are the longitudinal and transverse components of the @xmath29-factor tensor . for electrons
the @xmath29-factor is nearly isotropic in the qds studied here , @xmath30 , while for heavy holes @xmath31 as a rule @xcite .
we also do not consider any in - plane anisotropies of the carrier @xmath29-factors .
the operator @xmath32 in the hamiltonian describes the hyperfine coupling between the charge carrier and nuclear spins , @xmath33 .
\label{omegan}\ ] ] the summation in eq
. runs through the large number of host lattice nuclei in the qd ( @xmath34 ) , @xmath35 are the spin operators of the @xmath36 nucleus , @xmath37 denotes the hyperfine coupling constants proportional to the probability to find an electron ( hole ) at a given nucleus , and @xmath38 is the anisotropy parameter .
in particular , @xmath39 for electrons and @xmath40 ( typically @xmath41 ) for heavy holes @xcite .
we note that the pronounced anisotropies of both the zeeman effect and the hyperfine interaction for heavy holes results from the @xmath1-type character of its bloch functions as compared with the @xmath42-type character of the electron bloch functions .
the last term in eq . denotes the hamiltonian for the nuclear spin system . due to the smallness of the nuclear gyromagnetic ratios , the effect of the external magnetic field on the nuclei
can be disregarded @xcite . however , in iii - v qds the spins of all nuclear isotopes exceed @xmath43 , which makes the quadrupolar splittings of the nuclear spin states due to strain and electric fields important @xcite .
when neglecting the relatively weak nuclear spin - spin interactions , which can be important for nuclear spin dephasing , one can recast the hamiltonian of the nuclear spin system as @xcite @xmath44 \right\}. \label{hqc}\end{aligned}\ ] ] the quadrupolar axis of the @xmath45-th nucleus generated by the growth induced strain @xcite is given by @xmath46 , and the additional vectors @xmath47 are chosen such that they complete an orthonormal basis with @xmath46 .
furthermore , @xmath48 are the quadrupole splitting strengths and @xmath49 denotes the parameter describing the in - plane anisotropy .
note that @xmath50 for the @xmath51ga , @xmath52ga and @xmath53as isotopes , while @xmath54 for the @xmath55 in isotope .
the latter is relevant for ( in , ga)as qds as studied here .
given the hamiltonian , one can , in principle , evaluate the spin dynamics and the spin noise in qds by making use of the fact that , in eq .
, the spin - spin correlation function can be expressed as @xcite @xmath56,\end{gathered}\ ] ] with @xmath57 being the equilibrium density operator . due to the huge number of involved nuclei , however , this problem is hardly tractable even numerically . on the other hand ,
the electron and nuclear spin systems possess drastically different time - scales of evolution @xcite .
therefore , certain simplifications and model approaches are usually engaged , which provide complementary information on spin noise . these approaches ,
their areas of validity and the corresponding results are briefly summarized below .
we note already at this point , that each model results in a characteristic set of parameters when applied for describing the recorded experimental data .
once this parameter set is fixed for a particular configuration , it is held constant then and used consistently also when applying the specific model to other configurations .
these parameter sets are given in tab .
[ tab:1 ] . in the framework of the semiclassical approach developed in ref .
@xcite , the spin dynamics of the charge carriers and nuclei in eq . are decoupled .
moreover , one neglects the dynamics of the nuclear spins . in this case
, the electron or hole spin simply precesses around the static effective field with the frequency @xmath58 where @xmath59 is given by eq . , and @xmath60 is the effective frequency of the precession in the field of the frozen nuclear fluctuations .
the latter obey a gaussian distribution @xmath61 here the parameter @xmath62 determines the dispersion of the nuclear fields acting on the charge carrier .
the superscript @xmath63 in @xmath64 has been omitted to shorten the notations .
this distribution is isotropic for electrons and anisotropic for heavy holes . from eq .
one can deduce that @xcite @xmath65 as a result , the individual charge carrier spin noise spectrum reads @xcite : @xmath66}{\left[1+(\omega-\omega_{{\mu}})^2{\tau_s^\mu}^2\right]\left[1+(\omega+\omega_{{\mu}})^2{\tau_s^\mu}^2\right]}\right\rbrace,\end{gathered}\ ] ] where @xmath67 , @xmath68 is the angle between @xmath69 and the @xmath4-axis , and @xmath70 is the phenomenological spin relaxation time unrelated to the hyperfine interaction , with @xmath71 as a rule . in qd ensembles one has to average eq . over the distributions of both the @xmath29-factors and the parameters @xmath62 as well as @xmath72 characterizing the hyperfine interaction , which are caused by the spread of qd sizes , shapes and compositions @xcite . the spin noise spectrum described by eq
. demonstrates two peaks at @xmath73 in general @xcite .
the peak at positive frequencies is related with the spin precession . at sufficiently small fields ( @xmath74 )
its position is determined by the characteristic frequency of spin precession in the field of nuclear spin fluctuations , @xmath75 , whereas at high fields ( @xmath76 ) it is determined by the larmor frequency @xmath77 of the spin precession about the external field .
the shape of the precession peak is determined by the distribution function of the random nuclear fluctuations .
the peak at @xmath78 in the noise spectrum is caused by the spin components conserved during precession @xcite . in the semiclassical model outlined above
this peak obeys a lorentzian shape with the half width at half maximum ( hwhm ) being @xmath79 .
in fact , the semiclassical treatment with frozen nuclear spins fails to accurately describe the low - frequency behavior of the spin noise , since at low frequencies @xmath80 the nuclear dynamics becomes important . generally , to describe the features of the spin noise spectrum at low frequencies , a quantum mechanical treatment as described in sec .
[ sec : quantum ] is needed . before explaining this treatment in detail we briefly address some situations in which the semiclassical approach can be used .
basically , the semiclassical treatment can be applied to study the spin noise at low frequencies @xmath80 if the electron ( hole ) feedback on the nuclear spin dynamics can be neglected . hence , the temporal evolution of electron ( hole ) and nuclear fluctuations are decoupled in eq . .
however , the dynamics of the nuclear spins can be included into consideration also . in this situation , the electron spin precesses , in addition to the static external field , in a time - dependent nuclear field @xmath81 .
this can be realized in a semiclassical treatment if the electron or the hole leaves its site of localization before nuclear spin precession takes place , i.e. due to hopping to a neighbouring empty qd or due to excitation to the wetting layer and consequent localization in another dot @xcite .
then the nuclear fields can be treated as gaussian random functions with correlation time @xmath82 , and analytical expressions for the spin noise spectrum can be derived , as shown in ref .
another regime where the semiclassical approach can be still employed is the case of strong quadrupolar splittings of the nuclear spin states , @xmath83 @xcite .
in this case one can find all temporal correlations of the nuclear fields , i.e. second order , @xmath84 , and higher order correlators , from the quadrupolar hamiltonian and then evaluate the electron ( hole ) spin noise spectra semiclassically from the bloch equations for the electron spin dynamics in a time - dependent magnetic field .
provided that the correlation time @xmath82 is finite but long enough , @xmath85 , the spin noise spectrum is described by eq . with some `` effective '' spin relaxation @xmath79 , which may be magnetic field dependent
particularly , at @xmath86 one has @xmath87
@xcite .
the inaccurate description of the zero - frequency peak in the semiclassical approach can be avoided using the quantum mechanical definition of the spin correlation functions , eq . , which we evaluate by applying the chebyshev polynomial expansion technique ( cet ) .
it was originally developed as a numerical tool to calculate the exact time evolution @xmath88 of an arbitrary , originally pure state @xmath89 in a well defined time interval @xmath90 @xcite .
as recently demonstrated @xcite , it can also be applied to evaluate correlation functions as @xmath91 by using the quantum typicality of random quantum states @xcite .
the spin noise spectrum due to precession of electrons in the overhauser field at zero external magnetic field , calculated by this quantum mechanical approach agrees well with the semiclassical result , eq . , and is only determined by the second moment @xmath62 of the distribution function for the @xmath37 , i. e. eq . , and , if applied , by the external magnetic field @xcite .
the zero - frequency peak or more precisely the low - frequency spin noise peak in the noise spectrum is governed by the interplay of the electron or hole feedback on the nuclear spin dynamics , the non - uniformity of the hyperfine interaction @xcite and the nuclear quadrupolar couplings @xcite .
this interplay between the parameters @xmath37 , @xmath72 , @xmath48 , @xmath92 and @xmath49 of the hamiltonian has significant impact on the shape of the spin noise spectrum at small frequencies .
a detailed discussion on the choice of the hyperfine coupling constants @xmath37 used in the cet can be found in ref .
the parameters associated with the quadrupolar interaction are derived from recent microscopic studies of the electric field gradients in ( in , ga)as quantum dots @xcite assuming an in concentration of @xmath93 .
although we only can include a small amount , @xmath94 , of spin-@xmath95 nuclei in the quantum mechanical simulation , we generate the coupling constants @xmath48 and the orientation vectors @xmath96 from a distribution function reproducing the average values found in the microscopic study by c. bulutay @xcite .
the average deviation angle @xmath97 between the orientation vectors @xmath96 is set to @xmath98 .
for coupling constants @xmath99 , the random variable @xmath100 is drawn from a uniform probability distribution in the interval @xmath101 $ ] , and the magnitude @xmath26 is determined by the ratio @xmath102 that will be set to a fixed value . in order to mimic larger amounts of nuclear spins ,
we have averaged over 50 different configurations of @xmath103 and @xmath48 distributions with a fixed @xmath62 and @xmath104 .
so far , we have only discussed the modeling of a single ( in , ga)as qd by the cet . since the experimental measurements are performed on an ensemble of qds , the averaging over the variations of the hyperfine field parameters should be implemented as noted in sec .
[ sec : semiclass ] . to that end
, we have assumed that the radius of the qds , which determines the parameters @xmath37 and the @xmath29-factors , is varying within the qd ensemble . for details concerning this averaging procedure
see ref .
we study @xmath1- and @xmath0-doped ensembles of self - assembled qds .
both samples have 20-layers of mbe - grown ( in , ga)as / gaas qds separated by 60 nm gaas barriers with a qd density of 10@xmath105 @xmath106 per layer .
the @xmath1-doped sample was annealed for 30 s at 960 @xmath107c , and the @xmath0-doped at 945 @xmath107c , shifting the emission spectra of the samples to the sensitivity range of silicon photodetectors and reducing the inhomogeneity of the qd ensemble .
although the @xmath1-doped qds were not intentionally doped , the sample has a background level of @xmath1-type doping due to residual carbon impurities @xcite .
the @xmath0-type sample doping was obtained by incorporating @xmath108-sheets of si - dopants 20 nm below each qd layer , with the dopant density roughly equal to the qd density @xcite .
the samples are mounted on the cold finger of a liquid helium flow cryostat , and cooled down to a temperature of 5 k. figure [ wav ] shows low excitation photoluminescence ( pl ) spectra ( solid curves ) for the @xmath0-type sample , panel ( a ) , and for the @xmath1-type sample , panel ( b ) , demonstrating mostly emission from the qd ground state transitions .
the pl has its maximum at 1.398 ev ( 887 nm ) for the @xmath0-type qds , and at 1.381 ev ( 898 nm ) for the @xmath1-type qds .
-doped qd sample ;
( b ) the same for the @xmath1-doped sample .
the slight red shift of the spin noise compared to the pl for the n - doped sample may result from a concentration of singly charged dots on the low energy side , while towards higher energies the dots become increasingly uncharged .
the blue shift in case of the p - doped sample , on the other hand , may result from a double charging of the low energy dots with a pair of holes forming a spin zero singlet state . towards the high energy side single resident hole spin charging occurs in the dots , resulting in an increase of spin noise .
this increased p - doping level is confirmed by the increased intensity of emission from excited states compared to the n - doped sample . ]
the linearly polarized probe light is taken from a tunable continuous wave ( cw ) titanium - sapphire ring laser emitting a single frequency laser line with a linewidth of < 10mhz .
the laser is then guided through a single mode optical fiber for spatial mode shaping ( @xmath109 ) .
after a glan - taylor polarizer providing a linear extinction ratio of 10@xmath110 , the beam is transmitted through the sample with a large area focus of 100 micrometers diameter , in order to minimize optical excitation effects in the qd ensemble @xcite .
detection of the faraday rotation noise is done with a standard polarimeter , consisting of a wollaston prism and a broadband optical balanced detector .
two different detectors are used to study slower and faster dynamics : one has a bandwidth of 40 khz @xmath111 650 mhz ( new focus 1607-ac ) requiring optical powers of typically a few milliwatts ( _ fast _ detector ) , while the other one ( femto hca - s ) covers a bandwidth of dc @xmath111 100 mhz , requiring powers of a few hundreds of microwatts ( _ slow _ detector ) .
the output of the detectors is amplified and low - pass filtered at variable cutoff frequencies to avoid undersampling , and then sent to the input of a digitizer .
the digitizer incorporates field programmable gate array ( fpga ) technology for real time computation of the fast fourier transform ( fft ) of the digitized voltage samples .
the maximum bandwidth is limited at 1ghz , providing 16384 spectral lines at a spectral resolution of 61.04 khz . if necessary , the bandwidth can be reduced down to 200 mhz with a resolution of 12.21 khz . for even higher resolution
as required , e.g. , in the case of zero and longitudinal magnetic field measurements , spectra could be also acquired with a pcie digitizer card installed in a computer , in combination with multicore fft processing and averaging . in this work , the signal accumulation times varied between 1 minute and 30 minutes , depending on the overall noise signal amplitude . figure [ wav ] shows , in addition to the pl spectra discussed above , the integrated spin noise power , i.e. the area under the spin noise power density spectrum ( sns ) , @xmath112 , as function of the probe wavelength for both samples .
the spin noise power nearly follows the pl line , with maxima at 889 nm for the @xmath0-type sample and at 896 nm for the @xmath1-type sample .
this is the typical behaviour for inhomogeneously broadened ensembles @xcite .
thus the laser is tuned into the absorption band to provide measurable noise signal . using electromagnets we are able to apply magnetic fields , in the longitudinal direction ( faraday geometry , along the growth direction of the sample ) up to @xmath113 = 120 mt and in the transverse direction ( voigt geometry ) up to @xmath114 = 350 mt . in order to remove the spin - independent contribution to @xmath115 from the overall photon shot noise and the intrinsic electronic noise
, the spin noise measurements were interleaved between two magnetic fields : one measurement with the desired strength and direction , and the other one applied in the voigt geometry at @xmath116 250 mt for shifting the precession peak out of the detectable bandwidth .
subtraction of both spectra then yields the pure spin noise contribution . in order to account for the frequency dependent sensitivity of the detectors ,
the noise power spectra were additionally normalized by the photon shot noise power spectra measured without the sample , with the laser radiation shone directly onto the detectors .
in this section we present the experimental data for the spin noise measurements in zero and finite transversal or longitudinal magnetic fields .
furthermore we provide the results of the simulations using the two theoretical approaches and discuss their limitations and advantages .
note that for convenient comparison with the experiment we present characteristic frequencies of the two models , such as @xmath117 and @xmath118 , as well as larmor frequencies and simulated spin noise spectra , in units of mhz , i.e. in units of an ordinary frequency , rather than by an angular frequency as used in the models . -doped and the ( b ) @xmath1-doped qd samples at zero external magnetic field using a probe power of 4 mw ( 50 w/@xmath119 ) with the fast detector .
the red arrow in panel ( a ) indicates the feature related with the precession of electron spins due to the hyperfine coupling to the nuclei . panels ( c ) and ( d ) show additionally fits ( red curve ) to the experimental data with the semiclassical model using the parameters @xmath120 mhz , @xmath121 ns , @xmath122 mhz , @xmath123 , @xmath124 ns ( c ) and @xmath125 ns ( d ) .
the spectra are calculated using eqs . and with @xmath126 for panel ( c ) , and eq . for ( d )
the dash - dotted red curves in panels ( e ) and ( f ) indicate fits obtained with the cet .
the cet results are calculated assuming @xmath127 , @xmath128 110 mhz and @xmath129 16 mhz .
the electron fraction for the @xmath0-doped sample is set to @xmath130 , and to @xmath131 for the @xmath1-doped sample . for the ensemble averaging a gaussian distribution of the qd radius @xmath132 with a relative standard deviation @xmath133
is assumed . the ratio @xmath104 is set to @xmath134 for electrons and to @xmath135 for holes since @xmath136 .
all further parameters are chosen as discussed in section [ sec : quantum ] . ]
figure [ zf ] compares the spin noise spectra for @xmath0- and @xmath1-doped qds , left and right column of panels , respectively , measured at zero external magnetic field .
for these experiments we used the fast detector and the fpga module to cover a broad frequency range . as shown in fig .
[ zf](b ) , the spin noise in the @xmath1-doped sample consists of a single peak centered around zero frequency . in the chosen log - log scale representation
we find that the curve closely follows a lorentzian in the frequency domain ( shown additionally by the dash - dotted line ) , which corresponds to an exponential decay in time @xcite .
the spin noise spectrum of the @xmath0-doped sample reveals , in addition to the zero - frequency peak , a pronounced `` shoulder '' ( termed in the following as precession peak or overhauser peak ) at a frequency of @xmath137mhz , as highlighted by the red arrow in fig .
[ zf](a ) .
this peak is related to the magnetic field caused by the intrinsic nuclear fluctuations , that act on the electrons as a constant ( frozen ) magnetic field pointing along a random direction .
the electron spin components parallel to the frozen nuclear field contribute to the zero - frequency peak in the signal with @xmath138 of the full spin polarization .
the other @xmath139 precess around the nuclear field and are thus seen as the precession peak @xcite .
the absence of such a shoulder in the hole spin ensemble signal is a direct indication of the anisotropic nature of the hole - nuclear interaction @xcite .
the relative amplitude of the precession peak and the zero - frequency peak is influenced by several factors : ( i ) finite correlation time of the electron in a quantum dot which might be caused , i.e. , by the probe - assisted transfer of the electrons into other qds , providing a redistribution of the spin noise power of the precession peak towards the zero - frequency peak @xcite , or ( ii ) presence of positively charged qds in the nominally @xmath0-doped sample , which would increase the relative weight of the zero - frequency peak .
in particular , the latter can be tested by applying a transverse magnetic field , as will be shown in the sec .
[ sec : trans : b ] .
it should be foreclosed at this point , that by applying this method we were able to evidence a percentage of about 20 @xmath140 of such non - intentionally @xmath1-doped qds , which is mostly due to impurities in the mbe machine .
however , this value slightly varies with the point at which the sample is probed by the laser .
thus , to reproduce the data in the semiclassical model , we assume that the ratio of electron and hole contributions to the measured sns is @xmath141 , see eq . .
the corresponding fit results for the @xmath0-doped sample are presented in fig .
[ zf](c ) , from which we can determine @xmath142 mhz and @xmath121 ns , which is consistent with estimations based on the hyperfine coupling constants , eq .
[ eq : def - delta - mu ] .
for the hole fraction in the same sample best agreement was achieved with @xmath122 mhz , @xmath123 and @xmath124 ns . the best agreement with the experimental data using the semiclassical model for the @xmath1-doped sample , see the red curve in fig . [ zf](d ) ,
is obtained by using the parameters @xmath122 mhz and @xmath123 in eq .
, as well as @xmath125 ns in eq .
( note that the curve is practically insensitive to a particular value of @xmath118 provided that @xmath143 mhz ) the panels ( e ) and ( f ) of fig . [ zf ]
show a direct comparison between the experimental data and the numerical results obtained via the cet approach .
the spin noise function obtained via the cet is highly sensitive to the choice of the model parameters introduced in sec .
[ sec : quantum ] . for electron spins ,
panel ( e ) , the agreement between theory and experiment in the numerically accessible regime is excellent .
the pronounced deviation at small frequencies stems from the finite frequency resolution of the cet above @xmath144 for the parameters used here .
the agreement between theory and experiment is less striking for the hole doped sample , where the theoretically determined gradient of the spin noise function in the intermediate frequency regime @xmath145 does not reproduce the experiment exactly .
we attribute that discrepancy to the small amount of nuclei in the cet simulations limiting the low frequency spectrum of the hamiltonian .
the obtained agreement between experiment and theory is still remarkably good within the finite size limitations of the cet and states the validity of the applied model .
all important parameters obtained from describing the experimental data using the two models are summarized in table [ tab:1 ] .
.sca and cet parameters determined from the zero magnetic field data which are consistently used also subsequently in describing the data in magnetic field . [ cols="^,^,^",options="header " , ] as a nearly non - perturbative method of measurement , spin noise should be able to unveil the intrinsic spin lifetimes in the studied systems .
however , the inhomogeneously broadened qd ensembles require the probe energy to be within the pl - emission band .
therefore , one has to pay special attention to the applied laser power and use the limit of lowest possible laser intensities to minimize excitation effects . as could be shown by yan li _ et al . _ for hole spins in a very similar sample , for the used large 100@xmath63 m laser spot and used probe powers in the range of 0.1mw , the qd spin noise shows only a weakly increasing linewidth of the zero - frequency peak with increasing probe power @xcite .
this shows , that probe induced effects are still present , but are minimized .
a linear fit of the linewidth power dependence should thus allow us to obtain a rather accurate value of the intrinsic spin relaxation time by extrapolating the fit to zero probe power .
this method was recently proven to give reliable results for the electron spin relaxation times in pump - probe studies @xcite . in our experiments , the zero - frequency peak in the spin noise of the @xmath0- and @xmath1-doped sample
is measured using the slow 100mhz detector .
its output is filtered for the frequency band from 0.01 mhz to 32mhz and then sent through a high frequency amplifier .
the probe laser powers range from 0.1mw to 0.6mw . since especially the lineshape of the sns from the @xmath0-doped qds deviates from lorentzian ( see above ) , we use a general definition of the linewidth , namely the geometrical half width at half maximum ( hwhm ) of the zero - frequency peak spectrum , denoted as @xmath146 , see also ref .
@xcite .
the effective linewidths extracted from our experimental data are depicted in fig .
we observe a weak increase of the linewidths with laser power in both samples .
however , the changes are much larger for the @xmath0-doped qds .
this partly originates from the much smaller spectral amplitudes of the zero - frequency peak as compared to the @xmath1-doped qds , for which the complete spin noise power is concentrated in the zero - frequency peak .
-doped sample the increase is a factor 2.5 stronger .
the larger error is caused by a much smaller spectral amplitude compared to that of the @xmath1-doped sample .
we determine the effective intrinsic spin lifetime in both samples by using the point of intersection of the linear fits with the vertical axis at zero laser power . ] from linear fits to the power dependences we obtain the following intrinsic linewidths : the @xmath1-doped sample obeys @xmath147mhz , whereas @xmath148mhz for the @xmath0-doped sample .
the linear increase of the linewidth is steeper for the n - doped dots : it grows by @xmath149 as the probe power is increased from 0.1mw to 0.6mw . for hole spins
we observe an increase of only @xmath150 within the same range of laser power .
this difference shall be studied in greater detail in future , here we point out only some related aspects .
first , as has been already mentioned above , there is a finite fraction of @xmath1-doped qds in the intentionally @xmath0-doped sample , which amounts to about 20 % . since at zero external field @xmath139 of the electron spin noise
is concentrated in the precession peak and only @xmath138 contributes to the zero - frequency peak , holes then may contribute up to 40 % to the spin noise of the zero - frequency peak , while the remaining 60 % stems from electrons .
thus , there is a significant hole contribution that may influence spectral shape and amplitude of the low - frequency sns from the @xmath0-doped sample .
second , both samples are measured under exactly the same probing conditions , especially with the same large laser spot of 100@xmath63 m that evidently produces only a weak perturbation of the spin system .
we prove this assumption by additionally looking at the fr noise amplitude , which is the square root of the fr noise power , obtained by integration of the sns from 0.012mhz to 32mhz . taking the dependence of the fr noise amplitude on the probe laser power
, we obtain an exponent of @xmath151 for both samples by fitting the dependence with a power law .
this value is quite close to unity , suggesting that the measurements were carried out close to thermal equilibrium , i.e. , close to the non - perturbative regime @xcite . as an example of the opposite case , ref
. @xcite showes signatures of strong optical perturbation , demonstrating the hole spin noise in a qd microcavity with the measurements carried out using @xmath152 times larger probe intensity than in our case .
the linewidth then is increased by a factor of 3 , i.e. by 200 % under a threefold increase of laser power .
this additionally corroborates that our measurements are performed largely non - perturbatively . using the well - known relation @xmath153
we derive the effective spin lifetimes .
we obtain @xmath154 = 0.46 @xmath63s for the @xmath1-doped sample , and @xmath155 = 0.41 @xmath63s for the @xmath0-doped sample .
these values are close to the known result of @xmath156 = 0.4 @xmath63s that was obtained in the refs .
@xcite . at first glance
it is quite surprising that electrons and holes have similar relaxation times , as they show a strong difference in the nuclear interaction strength .
otherwise , similar values for the spin decoherence time @xmath157 of qd electron and hole spins in magnetic field were revealed by the mode - locking technique applied to similar @xmath0- and @xmath1-doped heterostructures @xcite . as we will explain in the following ,
our observation of comparable timescales is related to the presence of additional quadrupolar interactions of the nuclei with electric field gradients in the dots , eq . .
without @xmath158 , the characteristic time scale for the electron spin noise is given by @xmath159 , and the one for hole spins by @xmath160 determined by the fluctuations of the overhauser field . while @xmath161 ns is found in the typical qd ensembles as investigated here ,
the @xmath1-wave nature of the hole - wave function reduces the hyperfine couplings @xmath162 by a factor of 10 compared to the electron spins @xcite .
the decay of @xmath163 is additionally suppressed by the anisotropy parameter @xmath164 , accounting for the anisotropic dipole - dipole coupling of hole spins to the nuclear spins @xcite , so that @xmath165 .
including a realistic modeling @xcite of the coupling between the nuclei s quadrupole moments and the electric field gradients ( efgs ) in the qds via @xmath158 is sufficient to explain the observed mismatch between the experiments and the prediction of the csm , see also ref .
this can be intuitively understood : adding an additional source of decoherence ( i.e. the quadrupole interaction ) to the csm hamiltonian results in a decrease of the coherence time of the considered electron / hole spins . without @xmath158 , the total spin component in @xmath4-direction ,
@xmath166 is a conserved operator in the csm provided that @xmath167 or @xmath168 , implying that each individual nucleus maintains its state until the next spin - flip process occurs , always involving a combined carrier and nuclear spin flip . adding @xmath158 breaks this conservation law for @xmath169 in the csm by allowing spin - flips in the nuclear spin bath without involving the electronic spin , and defines an additional long - time scale set by the coupling strength @xmath158 .
its effect onto the spin - noise function can be investigated in fourth order perturbation theory @xcite for fixed nuclear easy - axis vectors @xmath170 : a term linear in @xmath48 and cubic in @xmath37 accelerates the decay of @xmath171 but a @xmath172 term decelerates the spin decay .
we note that the influence of @xmath158 onto the decay of the electronic spin is indirect and only occurs in combination with the hyperfine interaction .
its effect is non - universal and can not be casted into a simple decay time or energy scale as defined by the fluctuation of the overhauser field .
the reason is related to its two opposite limits . for @xmath173 , typically relevant for the experiment , @xmath158 provides additional nuclear spin - flip processes leading to an additional spin decay at long time scales set by @xmath104 , eq . , as outlined above .
for the opposite limit , @xmath174 , the nuclear system must be diagonalized first , providing a set of two time - reversal doublets as the new eigenbasis for each nuclear spin energetically separated by @xmath175 . then the coupling to the electron / hole spin
is treated perturbatively , and the rigidity of the nuclear system can suppress additional dephasing for @xmath176 @xcite . since the @xmath177 have different strain induced orientations at each nuclei in a real qd , their angular distribution @xcite and the in - plane anisotropy @xmath49 , however , provide an additional source of randomness and , hence , spin dephasing even for stronger hyperfine coupling due to the misalignment of the local nuclear easy - axis and the growth orientation , defining the global @xmath4-axis .
crucial for the understanding of the comparable long - time spin decay of electron spins and hole spins is the fact that @xmath178 is caused by the growth induced strain in the qds , and its energy scale is consequently independent of the qd doping .
this additional source of decoherence counteracts the lifetime enhancement of a factor @xmath179 as predicted within the csm for hole spins .
the quantum mechanical cet predicts @xcite that the long - term lifetime is of the same order of magnitude for @xmath0-type and @xmath1-type qds grown under identical growth conditions , as demonstrated in fig .
[ zf](e ) and ( f ) . at values @xmath180 as used in fig .
[ zf](f ) , the nuclear quadrupole electric couplings become the limiting factor of the coherence time in zero magnetic field . on a semiclassical level , the influence of the additional quadrupolar couplings can be understood by investigating the equation of motion of a single nuclear spin in the hamiltonian coupled to a fictitious static carrier spin . within the csm
only , the nuclear spin would precess around this central spin with a larmor frequency proportional to the individual @xmath37 .
this leads to slow dynamics of the overhauser field @xcite and a long - time carrier spin decay described by a power law @xcite with some logarithmic corrections governing the low frequency part of the spin - noise spectra . adding the quadrupolar couplings
@xmath158 has two effects on the nuclear spin dynamics : firstly it enhances the nuclear larmor frequency , and , secondly , the breaking of the total spin conservation law translates into a change of an effective precession axis which is not longer determined by the carrier spin direction only . both effects yields to an addition dephasing of the overhauser field and consequently to a dephasing of the non - decaying fraction of the spin - correlation function @xcite on a time scale dominated by the quadrupolar coupling strength .
a magnetic field applied perpendicular to the probe beam propagation direction , @xmath181 ( voigt geometry ) , induces spin precession and shifts the corresponding noise peak along the frequency axis to the larmor frequency @xmath77 .
we have carried out measurements for electron and hole spins up to field strengths of @xmath182 mt , using the fast 650mhz photodetector with a probe power of 4mw .
the experimental results are presented in fig .
[ elpr ] panel ( a ) for the @xmath1-type sample and panel ( b ) for the @xmath0-type sample .
we start the analysis with the data on the @xmath1-type sample , shown in panel ( a ) . at @xmath183
the sns consists of two peaks : the first one remains centered at @xmath78 , but its amplitude decreases with increasing magnetic field , and the second one appears at @xmath184 , shifting with the field towards higher frequencies , in agreement with our expectations , see sec .
[ sec : semiclass ] , because the transverse magnetic field suppresses the role of the nuclear field fluctuations , therefore reducing the zero - peak contribution .
the experimental data are adequately described by the semiclassical model , eq . ,
using the following values of fit parameters : transverse hole @xmath29-factor @xmath185 , measured from the linear shift of the second peak maximum with magnetic field , @xmath29-factor spread of @xmath186 % , measured from the broadening of the peak , and @xmath187 ns for @xmath188 mt , see the solid curves in fig .
[ elpr](a ) for results .
to achieve better agreement between experiment and theory at @xmath189 the curve centered at zero frequency is calculated with a longer spin relaxation @xmath190 ns . this extension of the spin relaxation time by the magnetic field is beyond the scope of the semiclassical model and demonstrates its limitations . -doped qd sample in the transverse magnetic fields indicated in the legend .
the acronym `` psn '' means `` photon shot noise '' . fits with the semiclassical model are shown by the solid lines at @xmath191 = 40mt and 80mt .
the heavy hole @xmath29-factor is @xmath185 with a spread of 37% .
other parameters of calculations are the same as for fig .
[ zf](d ) except for @xmath192 ns at @xmath193 mt .
the dashed line at @xmath191 = 60mt indicates a fit with the cet model , for which a gaussian hole @xmath29-factor spread with a relative standard deviation of @xmath194 has been assumed ; ( b ) sns measured in the @xmath0-doped qd sample at the same transverse magnetic fields , showing also signatures of hole doped qds in the ensmble .
fits with the semiclassical model are again depicted by solid lines , with parameters of the hole spin noise being the same as in panel ( a ) .
electron @xmath29-factor is @xmath195 with a spread of 7% and the other parameters are the same as for fig .
[ zf](c ) .
the fit at @xmath191 = 60mt uses again the cet approach , with the @xmath29-factor being the same as for the sca approach .
all other cet parameters except the hole @xmath29-factor spread are the same as in fig . [ zf ] . ] to illustrate further the validity of the semiclassical model at non - zero fields and to analyze the interplay of the nuclear spin fluctuations and the external magnetic field in more detail , we plot the ratio of the area under the zero - frequency sns peak to the area under the whole sns for an extended range of field strengths up to @xmath191 = 120 mt , see fig .
[ fig : zero ] .
the area of the zero - frequency peak was extracted by fitting the sns by a linear combination of a gaussian function centered at @xmath196 and two lorentzians centered at @xmath78 with different widths .
the gaussian function is needed to account for the non - zero frequency spin precession peak and the two lorentzians reflecting exponential decays in time are needed to reproduce rather accurately the shape of the zero - frequency component whose precise calculation is beyond the scope of the semiclassical model .
the gaussian is characteristic for a behavior determined by an inhomogeneous distribution such as the @xmath29-factor variations @xcite .
figure [ fig : zero ] demonstrates good agreement between theory and experiment .
next we turn to the sns of the @xmath0-type qd sample in transverse magnetic field , shown in fig .
[ elpr](b ) . here , at @xmath183
one can see two features in the sns appearing at non - zero frequencies .
one of them correlates with the hole larmor precession , fig .
[ elpr](a ) , while the other one has much higher frequency and can be attributed to the electrons . as mentioned above , the ratio of the areas under these two peaks is slightly sensitive to the particular point of the sample that is probed by the laser .
this observation supports our conjecture that @xmath0- and @xmath1-doped quantum dots coexist in the nominally @xmath0-type sample , see sec .
[ sec : zerob ] .
hence , as in sec .
[ sec : zerob ] , to describe the experimental data within the semiclassical model we take into account contributions of electron and hole spins to the measured sns . the particular data in fig .
[ elpr](b ) are well described by eqs . and , see solid curves in fig .
[ elpr](b ) , calculated using a fraction of positively charged qds , @xmath197 .
other fit parameters are chosen as follows : @xmath198 , its spread is about @xmath199 % , and @xmath121 ns , taken independent on @xmath191 .
we note that such a short value of @xmath200 is introduced into the semiclassical model to approximate the relatively broad zero - frequency contribution in the sns .
an analysis of the `` intrinsic '' electron spin relaxation time is presented below in sec .
[ sec : longb ] . in fig .
[ elpr ] we demonstrate the agreement between the experimental results and the cet for transversal external fields .
the cet parameters are set to those of the zero - field calculation used in fig .
the comparison between the experimental results and the cet data ( dashed lines ) for @xmath201 serves as a proof of principle assuring comparable results for the remainder experimental data . for the @xmath1-type samples , the electron - spin contribution to the spectrum is negligible , and the cet reproduces the obtained hole - spin spectrum at all frequency ranges in fig .
[ elpr](a ) .
-doped sample using the semiclassical approach .
points are obtained by fitting the experimental data , see text for details .
the solid curve is calculated using eq . with the same set of parameters as fig .
[ elpr](a ) . ]
as noted above , the sns of @xmath0-type samples shows a shallow low - frequency feature tracing perfectly the spin - noise peak of the @xmath1-type samples with increasing magnetic field . due to the @xmath202 admixture of holes to the @xmath0-doped spectrum
, the hole sns provides an additional finite low - frequency contribution around the hole larmor frequency that is accurately reproduced by the cet approach as shown in fig .
[ elpr](b ) . as previously demonstrated @xcite , the width of the spin precession peak is mainly governed by the parameter @xmath62 at small and moderate fields , while at higher fields it is controlled additionally by the @xmath29-factor spread ; the peak shape is approximately gaussian for @xmath203 .
the transversal magnetic field reduces the non - decaying fraction of the spin - correlation function responsible for the zero - frequency @xmath108-peak in the spin - noise spectrum .
consequently spectral weight is transferred to finite frequencies to fulfil the spectral sum rule @xcite , which describes the conservation of a total number of fluctuating spins .
this is clearly visible in the @xmath1-type spectra shown in fig .
[ elpr](a ) . such a suppression
is also found in @xmath0-type samples .
synopsis of the transverse - field results shows that the data in figs .
[ elpr ] and [ fig : zero ] can be well described by both models , the semiclassical and the quantum - mechanical one .
under application of a magnetic field along the light propagation axis the role of transverse fluctuations of the overhauser field diminishes .
this leads to : ( i ) suppression of the precession peak , and ( ii ) modification of the zero - frequency peak amplitude and shape .
as noted above , the total area under the noise spectrum remains constant and does not depend on magnetic field due to conservation of the number of spins in the probed volume . in this section
we focus on the zero - frequency peak .
this allows / requires ( i ) usage of the highly sensitive slow detector with low probe laser powers to minimize optical excitation effects and ( ii ) detailed monitoring of zero - frequency peak modifications because of higher spectral resolution . in particular , we filtered the detector output voltage for the frequency band between 0.009 mhz and 17.5 mhz .
the suppression of the precession peak in the @xmath0-doped sample then becomes indirectly observable by changes of the integral spectral area of the zero - frequency peak , as for higher longitudinal fields the spins become stabilized and redistributed in frequency into the zero - peak component . figure [ lonspec ] shows the low - frequency sns of the @xmath0-type sample , panel ( a ) , and of the @xmath1-type sample , panel ( b ) , measured at 0.2 mw probe laser power for different values of the longitudinal magnetic field .
it was our aim to measure both samples under comparable probing conditions with the lowest possible laser power .
the chosen low probe laser power is dictated by the small noise amplitude in the selected frequency range for the @xmath0-doped sample at zero external field , because the nuclear field distributes 2/3 of the total spectral weight of the electron spin noise into the overhauser precession peak ( not shown ) .
-doped and ( b ) @xmath1-doped qd ensemble at different longitudinal magnetic fields using a probe laser power of 0.2 mw , see legend for the field strengths .
the black dashed lines illustrate the @xmath204-asymptotics . ]
the experimental data clearly demonstrate the enhancement of the zero - frequency peak with increasing magnetic field .
the spectra of both samples transform towards a @xmath204 behaviour of the sns at sufficiently high field strengths , as reported previously for a @xmath1-doped sample @xcite .
this indicates a crossover to a @xmath205 spin decay in time , which is especially noteworthy for the @xmath0-doped sample , since the longitudinal field redistributes electron spin fluctuations from the overhauser precession peak into the zero - frequency peak ( see below ) , such that at the maximum applied field strength the electron spin contribution to the zero - frequency peak clearly dominates , despite the finite admixture from @xmath1-doped qds .
thus , the crossover to @xmath204-noise can be termed a general property of carrier spins in qds , regardless of the type of resident carrier charge . to obtain deeper insight into the complex dynamics of interacting electron and nuclear spins we extract again the effective spin relaxation times @xmath206 from the half - width at half - maximum ( hwhm ) of the zero - frequency sns component ,
see fig .
[ arealifetime](a ) .
since the shape of the zero - frequency feature strongly deviates from a lorentzian , especially for sufficiently high magnetic fields where @xmath204 behavior dominates , such procedure only gives effective timescales for the spin relaxation , as explained above in sec .
[ sec : zerob : long ] .
the red dimonds in fig .
[ arealifetime](a ) show the effective relaxation times @xmath154 for the @xmath1-doped qd sample .
it abruptly increases with increasing magnetic field up to @xmath207mt , from @xmath208 @xmath63s ( consistent with the measurements shown in secs .
[ sec : zerob ] , [ sec : trans : b ] ) to 5.3@xmath63s .
the longitudinal field overwhelms the anisotropic nonzero hyperfine field experienced by the hole through the hole - nuclear coupling , and thus suppresses the hole spin dephasing due to hyperfine interaction . for the magnetic field range from 20 mt to 120 mt
, @xmath154 slowly increases further up to about @xmath209 @xmath63s , showing signatures of saturation at the highest field strengths .
we ascribe this behaviour to a small fraction of non - intentionally @xmath0-doped qds in this sample .
notably , the width of the dip in the magnetic field dependence of the electron spin lifetime around zero field should be considerably larger than that for holes .
this is because the hyperfine field acting on the electron spin is stronger .
hence , a larger external magnetic field is needed to overcome the hyperfine field .
this assumption is supported by the spin lifetime behaviour @xmath210 observed for the @xmath0-doped sample , see black dots in fig .
[ arealifetime](a ) .
the broad component is significantly more pronounced than in the @xmath1-doped sample .
the narrow dip for fields @xmath211 mt is ascribed to the presence of @xmath1-doped qds in the @xmath0-type sample .
the broad , electron - induced component demonstrates a further increase in the effective electron spin lifetime to 5.6 @xmath63s , showing also pronounced signatures of saturation at @xmath212mt .
additionally , with increasing magnetic field strength , the 2/3 of total spectral weight of electron spin noise contained at zero external field in the precession peak become continuously redistributed into the zero - frequency peak . in order to demonstrate this behaviour , we made use of two facts : ( i ) a simple measure for the spectral weight of the spin noise is provided by the area under the sns , which in turn can be obtained from an integration , @xmath213 ; ( ii ) for the analysis of the redistribution , it is sufficient to monitor the increase of the spectral area under the zero - frequency peak
. thus , we performed the integration of the sns in the frequency range depcited in fig .
[ lonspec ] , i.e. from 0.009 mhz to 10 mhz . for our @xmath0-doped sample
we finally obtain a dependence of the zero - frequency peak area on @xmath214 as depicted by the black dots in fig .
[ arealifetime](b ) .
the continuous redistribution of the spectral weight of the electron spin noise is clearly seen , whereupon saturation occurs for @xmath215 80 mt . in principle , the observed behaviour can be explained with the total longitudinal field overcoming the transverse components of the nuclear hyperfine field that is isotropic in the n - doped qd ensemble , such that finally an anisotropic total magnetic field acts on these qds , shuffling all spectral weight of the electron spin noise into the zero - frequency peak , as is the case for hole spins @xcite .
however , the total increase of the zero - frequency peak area in the @xmath0-doped sample amounts to a factor of @xmath2162.3 only , instead of the expected factor of @xmath217 @xcite .
this is in line with the presence of @xmath1-doped qds , and the spectral cutoff at the smallest frequencies ( @xmath218 mhz ) . to complete the discussion of the experimental observations ,
we address the dependence of the spin noise amplitude of the @xmath1-doped sample on the longitudinal field strength , see the red diamonds in fig .
[ arealifetime](b ) .
the absence of any pronounced dependence on @xmath214 indicates that the amount of @xmath0-type quantum dots in this sample is rather small .
indeed , for @xmath219 % of @xmath0-type dots their contribution to the zero - frequency peak amplitude can hardly be resolved within our experimental accuracy . summarizing the experimental findings
, we note that in sufficiently strong magnetic fields ( which suppress the frozen nuclear fields ) both types of carrier spins ( i ) demonstrate quite similar effective spin lifetimes , and ( ii ) show a trend towards a @xmath220 spin decay manifesting itself as a @xmath204 behavior of the zero - frequency peak sns .
the semiclassical model outlined in sec . [ sec : semiclass ] provides an increase of @xmath156 with increasing longitudinal magnetic field if a finite value of the correlation time @xmath82 is taken into account .
however , our estimations show that such an increase should be observed at larger fields than observed , @xmath221 mt for the parameters extracted from the fits in sec .
[ sec : trans : b ] .
hence , such an abrupt increase as well as the weak increase of the hole spin relaxation time at @xmath222 should be analyzed in terms of the interconnected hole and nuclear spin dynamics @xcite , quadrupolar splittings @xcite and , possibly , taking into account other spin relaxation mechanisms unrelated to the hyperfine interaction @xcite .
on the other hand , the semiclassical model provides a reasonable description of the redistribution of the spectral weight of electron spin noise into the zero - frequency peak under longitudinal magnetic fields : taking the value of @xmath120 mhz from the fitting of the zero - field sns , fig .
[ zf](c ) , as well as 20 % of @xmath1-doped qds , @xmath223 , we obtain the red solid curve in fig .
[ lonfit ] without any additional fitting parameters , giving fair accord with the measured data .
note , that the value of the electron spin lifetime @xmath200 only weakly affects the dependence of the zero - frequency peak area on the longitudinal magnetic field . within the csm , the suppression of the electron spin decoherence in a longitudinal field @xmath224 with field strength @xmath225 leads to a rapid increase of the long - time limit @xmath226 of the spin correlation function @xmath227 defined in eq . .
@xmath228 is proportional to the spectral weight of the zero - frequency peak contribution , and its dependency on the dimensionless magnetic field @xmath229 , with @xmath230 denoting the frequency of the larmor precession induced by the longitudinal field ( @xmath231 ) , is given by the interpolation formula @xmath232 for the @xmath0-type qds @xcite , with @xmath233 being the initial value of the spin - correlation function at time @xmath234 , and @xmath235 denoting the characteristic timescale of the electron spin fluctuations ( see above ) .
note , that eq .
holds in the sca as well with an accuracy higher than @xmath236 . within the quantum mechanical approach @xcite to the csm
it was found that ( i ) the spectral weight of the spin precession peak rapidly decays as @xmath237 @xcite , ( ii ) the low - frequency spectral weight ( @xmath238 ) is strongly reduced as well as ( iii ) collapsing into the zero - frequency peak @xcite leading to eq . for longitudinal magnetic fields .
( b ) using the semiclassical approach with @xmath239mhz and @xmath223 as obtained from the zero - field and transverse - field sns .
the blue line shows @xmath228 obtained from eq .
, under particular consideration of eq . , and with parameters @xmath117 = 110 mhz and @xmath240 = 2.14 consistent with the cet calculations added to fig
. [ zf ] , as well as @xmath241 = 0.08 . ] the sns shown in fig .
[ lonspec ] suggests that the high energy parts of the spectrum are not solemnly transferred to the zero - frequency @xmath108-peak as predicted by the csm , but are distributed over some finite low - frequency range obeying @xmath204 behaviour as shown in fig .
[ lonspec ] .
this experimental observation supports the claim that a simple csm is insufficient to explain the observed sns , and additional terms in the hamiltonian are needed for the proper description of low - frequency parts of the sns .
as mentioned , the @xmath0-type and @xmath1-type qds show similar low - frequency features in the sns spectrum , leading to the estimated long - time scales in fig .
[ arealifetime](a ) that are quite comparable for both spin species and only weakly field dependent once @xmath242 . therefore
, the mechanism governing the low - frequency part of the spectrum must be the same for both types of qds and nearly independent of the longitudinal magnetic field .
this serves as a piece of evidence that the nuclear quadrupolar couplings , eq .
and refs .
@xcite provide this mechanism : the quadrupolar splittings defined by the hamiltonian @xmath243 are induced during the growth process @xcite and are independent of doping . moreover , since the nuclear magnetic moment is about 2000 times smaller than the bohr magneton , the nuclear system remains almost unaffected by the comparably weak longitudinal magnetic fields applied in our experiments .
replacing the non - decaying fraction of the spin - correlation function @xcite by the integral over the low - frequency part of the spectrum yields the same functional dependence of this signal area on the longitudinal magnetic field as in eq . for @xmath0-type qds .
in fact , identifying @xmath228 with the zero - frequency peak area in the sns and setting @xmath244 using the sca estimate @xcite precisely describes the increase of the zero - frequency peak area as function of the magnetic field , where @xmath241 takes the role of a prefactor fixed by the experiment . for the particular case of admixture from non -
intentionally @xmath1-doped qds as observed in our studies , eq .
should also remain suitable .
then the parameters @xmath245 as well as @xmath241 include the same ( field - independent ) contributions from hole spins , respectively , such that the field - dependent part describes the redistribution of electron spin noise from the overhauser peak into the zero - frequency peak , regardless of the amount of @xmath1-doped qds in the sample . in order to find out whether the expression in eq .
is sufficient to describe the data depicted in fig .
[ arealifetime](b ) under utilization of the cet parameters derived from the zero - field and transverse - field measurements ( @xmath117 = 110 mhz , @xmath21620 % of hole doped qds ) ,
we firstly derived a relation for the dimensionless magnetic field @xmath246 that contains both @xmath117 and @xmath214 . therein , we put @xmath247 .
furthermore , by taking @xmath248 to denote the larmor frequency of the precession peak in the zero - field sns , we can adopt ( i ) @xmath249 as a major cet result @xcite , and ( ii ) @xmath250 from the sca @xcite , to obtain @xmath251 . in combination ,
the relation @xmath252 is yielded , which we then included into eq . .
next , we accounted for 20 % of @xmath1-doped qds by incorporating the ratio @xmath253 , but left @xmath241 as a free parameter to optimize the fit between experiment and theory .
finally , @xmath254 is included . under all aforementioned preconditions
, we obtained the best agreement for @xmath255 .
the corresponding plot is depicted by the blue curve in fig .
[ lonfit ] .
as can be derived from this plot , the cet is well suited to describe our data .
especially , the good consistency of the model parameters obtained previously from the zero - field and the transverse - field measurements deserves particular attention . and @xmath256 in a varying longitudinal external field @xmath224 , and for @xmath257 mhz .
for the @xmath86 spectrum we supplemented a lorentzian ( dash - dotted line ) with a half width at half maximum of @xmath258 . , width=321 ] with this specific conclusion
we turn the analysis to the @xmath1-type qds .
the situation here is considerably different from the @xmath0-type qds : due to the significant anisotropy factor @xmath259 , the spin decay is suppressed in the csm : the gaussian peak from the overhauser field fluctuation is shifted to lower frequencies approximately by a factor of @xmath260 @xcite compared to @xmath0-type qds and , therefore , merges with the rest of the low frequency spectrum , as was also shown in fig .
11 of ref .
in addition , its area is reduced due to the suppression of the spin decay via the hyperfine coupling . to this end , the depicted frequency range in fig .
[ lonspec](a ) contains almost the complete spin noise , and , consequently , the area remains constant within the experimental restrictions as reported in fig .
[ arealifetime](b ) and in accordance with the spectral sum rule . for a detailed comparison between the experiment and the cet approach for the full hamiltonian , the frequency resolution for the sns must be matched .
since we have typically used a fixed number of chebyshev polynomials @xmath261 , and the energy spectrum of the hamiltonian increases with increasing magnetic field , the lowest accessible frequency in the cet approach also increases with the applied magnetic field .
therefore , the cet has very reliable access to the high - frequency part of the spectrum above 1 mhz , while that part on the experimental side is unfortunately already disappearing in the detector noise . on the other hand
, the cet approach lacks the necessary low - frequency resolution for @xmath262 1 mhz with increasing longitudinal magnetic field , where the crossover to a @xmath204-type spin decay has been detected in experiment as reported above .
nevertheless , such a crossover has also been found in a longitudinal magnetic field when applying the cet approach to the full hamiltonian for @xmath1-typed qds .
for the results presented in fig .
[ fig - holes ] , however , we have used slightly different parameters than in fig .
[ zf ] enabling us to reach lower frequencies . setting aside the increased experimental noise , the qualitative and quantitative agreement remains remarkable between theory and experiment : ( i ) with increasing magnetic field , the spectral weight is reduced above 0.1 mhz and transferred to frequencies below 0.1 mhz so that the sns approaches the @xmath204 asymptotic in the depicted interval , and ( ii ) all spectra cross each other at about @xmath263 0.1 - 0.15 mhz . due to the incompatibility of the experimental and theoretical frequency resolution in longitudinal magnetic fields we provide fig .
[ fig - holes ] as a proof of principle only and leave a more detailed parameter fit of the hamiltonian to the experimental data depicted in fig .
[ lonspec ] to the next generation of high power computers .
we have applied spin noise spectroscopy to measure the spin lifetimes of qd electrons and holes in zero and finite magnetic fields .
both types of charge carriers demonstrate a similar change of the sns shape in longitudinal magnetic field , from about lorentzian to @xmath204 dependence , as they become stabilized along the field .
our findings demonstrate that the spin noise in longitudinal fields allows for a clear separation of electron and hole spin contributions in qd ensembles and , correspondingly , the determination of their spin lifetimes due to the difference of their interaction strength with the nuclear spin fluctuations .
transverse magnetic fields allow one to define the fractions of electron and hole spin subensembles within a given sample .
the experimental data have been discussed in the framework of two complementary theoretical approaches : the chebyshev polynomial expansion technique ( cet ) which provides a fully quantum treatment of the spin noise problem and the semi - classical approximation ( sca ) which is based on the separation of the electron and nuclear spin dynamics .
the applicability of these approaches is evaluated for the different studied experimental configurations .
comparison of the experimental data with the modeling shows that the cet approach is a powerful method to describe the spectral sns shape in the intermediate and high frequency ranges .
the limitations of this method are given only by the computational power that apply in our case to frequencies below 0.1mhz .
an additional advantage of the cet is given by the rigorous inclusion of quadrupolar splittings of the nuclei which provides a quantitative explanation of the similar long - time / low - frequency spin dynamics for both types of carriers .
on the other hand , the sca enables one to elaborate an intuitive and consistent qualitative picture of the spin dynamics and spin fluctuations .
this approach has its limitations at very low frequencies to describe the shape of the zero - frequency peak .
however , it gives a good qualitative description of the overall spin noise spectra using simplified assumptions and a high quantitative precision for the high frequency part of the spectra .
this method can serve to estimate the parameters of the hyperfine interaction and the @xmath29-factors of electrons and holes from the experimental data which then can serve as input into a more advanced quantum mechanical approach .
we acknowledge the financial support by the deutsche forschungsgemeinschaft and the russian foundation of basic research through the transregio trr 160 , the rf president grants md-5726.2015.2 , and sp-643.2015.5 , the rf government grant 14.z50.31.0021 ( leading scientist m. bayer ) , the dynasty foundation , and a st .- petersburg government grant .
s. a. crooker , d. g. rickel , a. v. balatsky and d. l. smith , _ spectroscopy of spontaneous spin noise as a probe of spin dynamics and magnetic resonance _ , nature * 431 * , 49 - 52 ( 2004 ) s. a. crooker , l. cheng and d. l. smith , _ spin noise of conduction electrons in n - type bulk gaas _
, physical review b * 79 * , 035208 ( 2009 ) m. oestreich , m. rmer , r. j. haug and d. hgele , _ spin noise spectroscopy in gaas _ , physical review letters * 95 * , 216603 ( 2005 ) h. horn , a. balocchi , x. marie , a. bakin , a. waag , m. oestreich and j. hbner , _ spin noise spectroscopy of donor - bound electrons in zno _ , physical review b * 87 * , 045312 ( 2013 ) f. berski , j. hbner , m. oestreich , a. ludwig , a. d. wieck , and m. glazov , _ interplay of electron and nuclear spin noise in n - type gaas _ , physical review letters * 115 * , 176601 ( 2015 ) .
g. m. mller , m. rmer , d. schuh , w. wegscheider , j. hbner and m. oestreich , _ spin noise spectroscopy in gaas ( 110 ) quantum wells : access to intrinsic spin lifetimes and equilibrium electron dynamics _
, physical review letters * 101 * , 206601 ( 2008 ) s. v. poltavtsev , i. i. rhyzov , m. m. glazov , g. g. kozlov , v. s. zapasskii , a. v. kavokin , p. g. lagoudakis , d. s. smirnov and e. l. ivchenko , _ spin noise spectroscopy of a single quantum well microcavity _ , physical review b * 89 * , 081304(r ) ( 2014 ) n. h. bonadeo , j. erland , d. gammon , d. park , d. s. katzer and d. g. steel , _ coherent optical control of the quantum state of a single quantum dot _ , science * 282 * , 1473 - 1476 ( 1998 ) a. greilich , r. oulton , e. a. zhukov , i. a. yugova , d. r. yakovlev , m. bayer , a. shabaev , al .
l. efros , i. a. merkulov , v. stavarache , d. reuter and a. d. wieck , _ optical control of spin coherence in singly charged ( in , ga)as / gaas quantum dots _ , physical review letters * 96 * , 227401 ( 2006 ) a. greilich , d. r. yakovlev , a. shabaev , al . l. efros , i. a. yugova , r. oulton , v. stavarache , d. reuter , a. d. wieck and m. bayer , _ mode locking of electron spin coherences in singly charged quantum dots _
, science * 313 * , 341 ( 2006 ) s. a. crooker , j. brandt , c. sandforth , a. greilich , d. r. yakovlev , d. reuter , a. d. wieck and m. bayer , _ spin noise of electrons and holes in self - assembled quantum dots _ , physical review letters * 104 * , 036601 ( 2010 ) yan li , n. sinitsyn , d. l. smith , d. reuter , a. d. wieck , d. r. yakovlev , m. bayer and s. a. crooker , _ intrinsic spin fluctuations reveal the dynamical response function of holes coupled to nuclear spin baths in ( in , ga)as quantum dots _ , physical review letters * 108 * , 186603 ( 2012 ) luyi yang , p. glasenapp , a. greilich , d. reuter , a. d. wieck , d. r. yakovlev , m. bayer , and s. a. crooker , _ two - colour spin noise spectroscopy and fluctuation correlations reveal homogeneous linewidths within quantum - dot ensembles _ , nature communications * 5 * , 4949 ( 2014 ) m. m. glazov and e. l. ivchenko , _ spin noise in quantum dot ensembles _ , physical review b * 86 * , 115308 ( 2012 ) j. hackmann and f. b. anders _ spin noise in the anisotropic central spin model _
, physical review b * 89 * , 045317 ( 2014 ) j. hackmann , d. s. smirnov , m. m. glazov and f. b. anders , _ spin noise in a quantum dot ensemble : from a quantum mechanical to a semi - classical description _ , physica status solidi ( b ) * 251 * , 1270 - 1275 ( 2014 ) n. a. sinitsyn , yan li , s. a. crooker , a. saxena , and d. l. smith , _ role of nuclear quadrupole coupling on decoherence and relaxation of central spins in quantum dots _ , physical review letters * 109 * , 166605 ( 2012 ) . j. hackmann , p. glasenapp , a. greilich , m. bayer and f. b. anders , _ influence of the nuclear electric quadrupolar interaction on the coherence time of hole and electron spins confined in semiconductor quantum dots _ , physical review letters * 115 * , 207401 ( 2015 ) f. heisterkamp , e. a. zhukov , a. greilich , d. r. yakovlev , v. l. korenev , a. pawlis , and m. bayer , _ longitudinal and transverse spin dynamics of donor - bound electrons in fluorine - doped znse : spin inertia versus hanle effect _
, physical review b * 91 * , 235432 ( 2015 ) j. hackmann , _ spin dynamics in doped semiconductor quantum dots _
, phd thesis , technische universitt dortmund ( 2015 ) .
g. s. uhrig , j. hackmann , d. stanek , j. stolze , and f. b. anders , _ conservation laws protect dynamic spin correlations from decay : limited role of integrability in the central spin model _ , physical review b * 90 * , 060301 ( 2014 ) .
s. varwig , a. ren , a. greilich , d. r. yakovlev , d. reuter , a. d. wieck , and m. bayer , _ temperature dependence of hole spin coherence in ( in , ga)as quantum dots measured by mode - locking and echo techniques _
, physical review b * 87 * , 115307 ( 2013 ) .
d. s. smirnov and m. m. glazov , _ exciton spin noise in quantum wells _
, physical review b * 90 * , 085303 ( 2014 ) l. d. landau and e. m. lifshitz , _ statistical physics , part 1 _ ( butterworth - heinemann , oxford , 2000 ) v. s. zapasskii , a. greilich , s. a. crooker , yan li , g. g. kozlov , d. r. yakovlev , d. reuter , a. d. wieck and m. bayer _ optical spectroscopy of spin noise _ , physical review letters * 110 * , 176601 ( 2013 ) e. l. ivchenko and a. a. kiselev , _ electron @xmath29 factor of quantum wells and superlattices , _ sov .
* 26 * , 827 ( 1992 ) .
x. marie , t. amand , p. le jeune , m. paillard , p. renucci , l. e. golub , v. d. dymnikov , and e. l. ivchenko , _ hole spin quantum beats in quantum - well structures , _ physical review b * 60 * , 5811 ( 1999 ) . i. toft and r. t. phillips , _ hole @xmath29 factors in gaas quantum dots from the angular dependence of the spin fine structure _ , physical review b * 76 * , 033301 ( 2007 ) .
e. i. gryncharova , v. i. perel , _ relaxation of nuclear spins interacting with holes in semiconductors , _ sov .
. semicond . * 11 * , 997 ( 1977 ) .
j. fischer , w. a. coish , d. v. bulaev , and d. loss , _ spin decoherence of a heavy hole coupled to nuclear spins in a quantum dot , _ physical review b * 78 * , 155329 ( 2008 ) .
b. eble , c. testelin , p. desfonds , f. bernardot , a. balocchi , t. amand , a. miard , a. lematre , x. marie , and m. chamarro , _ hole - nuclear spin interaction in quantum dots , _ physical review letters * 102 * , 146601 ( 2009 ) .
e. a. chekhovich , m. m. glazov , a. b. krysa , m. hopkinson , p. senellart , a. lematre , m. s. skolnick , and a. i. tartakovskii , _ element - sensitive measurement of the hole - nuclear spin interaction in quantum dots _
, nature physics * 9 * , 74 ( 2013 ) .
m. i. dyakonov ( ed . ) , _ spin physics in semiconductors _ ( springer , berlin , 2008 ) r. i. dzhioev and v. l. korenev , _ stabilization of the electron - nuclear spin orientation in quantum dots by the nuclear quadrupole interaction _ , physical review letters * 99 * , 037401 ( 2007 ) .
w. a. coish and d. loss , phys .
b * 70 * , 195340 ( 2004 ) .
m. s. kuznetsova , k. flisinski , i. ya .
gerlovin , m. yu .
petrov , i. v. ignatiev , s. yu .
verbin , d. r. yakovlev , d. reuter , a. d. wieck , and m. bayer , _ nuclear magnetic resonances in ( in , ga)as / gaas quantum dots studied by resonant optical pumping _
, physical review b * 89 * , 125304 ( 2014 ) . v.
g. fleisher and i. a. merkulov , ch .
5 in `` _ _ optical orientation _ _ '' , eds .
f. meier and b. p. zakharchenya , north holland , amsterdam ( 1984 ) .
landau and e.m .
lifshitz , _ quantum mechanics : non - relativistic theory , _ butterworth - heinemann , oxford , uk ( 1977 ) .
i. a. merkulov , a. l. efros , and m. rosen , _ electron spin relaxation by nuclei in semiconductor quantum dots _ , physical review b * 65 * , 205309 ( 2002 ) m. m. glazov , _ spin noise of localized electrons : interplay of hopping and hyperfine interaction _ , physical review b * 91 * , 195301 ( 2015 ) a. v. khaetskii , d. loss , and l. glazman , _ electron spin decoherence in quantum dots due to interaction with nuclei _ , physical review letters * 88 * , 186802 ( 2002 ) .
k. a. al - hassanieh , v. v. dobrovitski , e. dagotto , and b. n. harmon , _ numerical modeling of the central spin problem using the spin - coherent - state @xmath1 representation _ , physical review letters * 97 * , 037204 ( 2006 ) .
l. cywinski , v. v. dobrovitski , and s. das sarma , _ spin echo decay at low magnetic fields in a nuclear spin bath _ , physical review b * 82 * , 035315 ( 2010 ) . supplemental material to reference @xcite ( 2012 ) r. dahbashi , j. hbner , f. berski , j. wiegand , x. marie , k. pierz , h. w. schumacher and m. oestreich , _ measurement of heavy - hole spin dephasing in ( inga)as quantum dots _ ,
applied physics letters * 100 * , 031906 ( 2012 ) c. bulutay , _ quadrupolar spectra of nuclear spins in strained _
quantum dots _ , physical review b * 85 * , 115313 ( 2012 ) .
e. fermi , _ ber die magnetischen momente der atomkerne _
, zeitschrift fr physik * 60 * , 320 ( 1930 ) .
j. fischer , w. a. coish , d. v. bulaev , and d. loss , _ spin decoherence of a heavy hole coupled to nuclear spins in a quantum dot _ , physical review b 78 , 155329 ( 2008 ) . c. testelin , f. bernadot , b. eble and m. chamarro , _ hole - spin dephasing time associated with hyperfine interaction in quantum dots _ , physical review b * 79 * , 195440 ( 2009 ) .
n. wu , n. froehling , x. xing , j. hackmann , a. nanduri , f. b. anders and h. rabitz , _ decoherence of a single spin coupled to an interacting spin bath _
, arxiv:1507.04514 ( _ submitted to _ physical review b ) | the spin fluctuations of electron and hole doped self - assembled quantum dot ensembles are measured optically in the low - intensity limit of a probe laser in absence and presence of longitudinal or transverse static magnetic fields .
the experimental results are modeled by two complementary approaches based either on semiclassical or quantum mechanical descriptions .
this allows us to characterize the hyperfine interaction of electron and hole spins with the surrounding bath of nuclei on time scales covering several orders of magnitude .
our results demonstrate ( i ) the intrinsic precession of the electron spin fluctuations around the effective nuclear overhauser field caused by the host lattice nuclear spins , ( ii ) the comparably long time scales for electron and hole spin decoherence , as well as ( iii ) the dramatic enhancement of the spin lifetimes induced by a longitudinal magnetic field due to the decoupling of nuclear and charge carrier spins . |
recently diffractive phenomena in qcd have attracted considerable theoretical and experimental interests .
in particular , diffractive photo- and leptoproductions of vector mesons off the proton , @xmath4 , provide crucial constraints on properties of the qcd pomeron and the vector meson wave functions@xcite . in this report , we focus on the heavy quarkonium production , where the heavy quark mass plays a role of the hard scale . due to the regge kinematics
, the vector meson production can be understood as the scattering of the color dipole with the target nucleon . in this case
we expect the amplitude to be dominated by the four diagrams illustrated in fig.1 .
the upper part of each diagram , where the heavy quarks interact with two - gluons , is assumed to be calculated within perturbative qcd even for the photoproduction . on the other hand , the lower blob represents both non - perturbative structure ( impact factor ) of the proton and complicated gluon dynamics at small-@xmath5 which is identified with the qcd pomeron .
the color dipole approach is adopted to evaluate the @xmath0 production in refs .
@xcite , and gives reasonable results in agreement with the data .
however , as pointed out by hoyer and peign@xcite very recently , the ratio of @xmath1 to @xmath0 photoproduction cross sections calculated in the dipole approximation is much smaller than the measured values@xcite at hera , when one uses typical non - relativistic quark models to obtain the charmonium wave functions .
the formula used in the dipole approach@xcite is obtained by assuming that both transverse momentum of exchanged gluons and the transverse size of the @xmath6 pair are small .
however , the non - relativistic quark model indicates that the typical size of @xmath7 is twice as large as the radius of @xmath0 .
hence , the validity of the small - size dipole approximation is questionable for @xmath8 and other mesons with the large radii .
we may expect considerable contributions from the scattering of the color dipole with a large transverse fluctuation , if the @xmath9 and @xmath6 wave functions allow such a large separation between the quarks .
above argument motivates us to calculate the diffractive @xmath10 and @xmath1 productions without resorting to the dipole approximation .
let us briefly introduce the framework to calculate the diffractive heavy quarkonium production , which was developed in ref .
we consider the forward heavy vector meson production @xmath11 shown in fig.1 with the kinematical condition @xmath12 . in this
small-@xmath5 kinematics , the lifetime of a color - singlet quark - antiquark fluctuation is large compared to the typical interaction time scale@xcite .
then the production amplitude can be expressed as a convolution of the @xmath13-proton cross section with the wave functions of initial ( @xmath9 ) and final ( @xmath14 ) states . in terms of light - cone wave functions of the photon @xmath15 and vector meson @xmath16 ,
the production amplitude at @xmath17 is proportional to @xmath18 \ ,
\psi_{v } ( z , b ) \ ; , \label{amp1}\end{aligned}\ ] ] where @xmath19 and @xmath20 are longitudinal momentum fraction of a quark and transverse distance between the @xmath6 pair , respectively .
@xmath5 is the longitudinal momentum fraction of the gluon @xmath21 with @xmath22 being the heavy quark mass , and @xmath23 the gluon s transverse momentum .
@xmath24 denotes the unintegrated gluon number density of the proton , which contains a complicated non - perturbative gluon ladder . @xmath25 and
@xmath26 represent the qcd coupling constant and the ordinary bessel function of the first kind . after multiplying suitable functions of @xmath19 and kinematical factors , and carrying out @xmath19-integration
, we obtain both longitudinal and transverse differential cross sections for the heavy quarkonium production@xcite ; @xmath27 \ ; \ ; , \label{cross}\end{aligned}\ ] ] where @xmath28 is the mass of the heavy quarkonium . in eq .
( [ amp1 ] ) the photon wave function @xmath15 with 4-momentum square @xmath29 is calculated perturbatively , while @xmath16 is a non - perturbative object .
we shall use the non - relativistic potential model which successfully reproduces the properties of the heavy quarkonium to evaluate @xmath30 . in the following calculations , we use two types of the potential ; the qcd motivated potential which essentially consists of linear and coulomb potentials@xcite , and a phenomenological logarithmic potential @xcite .
our procedure to evaluate the light - cone wave functions is in line with those of refs .
@xcite . solving the schrdinger equation with a potential @xmath31
, we obtain non - relativistic wave functions @xmath32 , and also subsequently those in the momentum space @xmath33 through the fourier transformation .
we make a simple kinematical identification between light - cone variables @xmath34 and @xmath35 as @xmath36 where @xmath37 are the transverse momenta of the quarks .
this relation gives the light - cone wave function ; @xmath38 ^ 3}}\right)^{1/4 } \cdot \psi_v \left ( |\vec k| = \sqrt{\frac{k_t^2+(2z -1)^2 m^2}{4 z ( 1-z ) } } \right ) \ ; \ ; .
\label{wf}\end{aligned}\ ] ] finally , performing the two - dimensional fourier transformation in the @xmath39 space , we obtain the light - cone wave function @xmath40 .
if we take the infinitely heavy quark mass limit , where the fermi motion of heavy quarks in the quarkonium can be entirely neglected , _ i.e_. @xmath41 , the production cross section is simply related to the leptonic decay width of the heavy quarkonium@xcite . however , for the charmonium , there exist substantial fermi motion corrections at small @xmath29@xcite . assuming that @xmath42 and keeping terms up to second order of @xmath43 @xcite
, one can rewrite eq .
( [ amp1 ] ) as @xmath44 \ ; \ ; , \label{dipole}\end{aligned}\ ] ] where @xmath45 . in eq .
( [ dipole ] ) the unintegrated gluon density has been identified with the gluon distribution function measured in dis via@xcite @xmath46 the approximation ( [ dipole ] ) is often called the color dipole approximation .
it gives a simple physical interpretation of this process , in which the scattering amplitude of the @xmath13 dipole with the target has a geometrical expression , @xmath47 .
in addition , the factorized formula eq .
( [ dipole ] ) makes it possible to constrain the gluon distribution function of the proton from the diffractive vector meson production .
typical results for @xmath48 production cross section obtained by using eq .
( [ dipole ] ) are available in refs .
@xcite , and in agreement with the data .
however , if we focus on @xmath1 , which is a radially excited @xmath49-state of the charmonium , the calculated ratio of @xmath50 to @xmath51 cross sections at the photoproduction point @xmath52 is much smaller than the experimental data as pointed out in @xcite .
the ratio calculated using eq .
( [ dipole ] ) is shown by the dashed curve in fig.2 .
although the theoretical prediction agrees with the data at larger @xmath29 , it has a factor 3 discrepancy at @xmath52 .
we have tried to evaluate it with several potential models , but none of them can describe the data as far as the parameters of the potential are fixed to reproduce the charmonium properties@xcite .
let us now examine carefully the dipole approximation which leads us to eq .
( [ dipole ] ) . for the leptoproduction with large @xmath29 eq .
( [ dipole ] ) works well , since the photon wave function projects out the @xmath6 fluctuation with the small transverse size .
this approximation is also reasonable for the photoproduction of the @xmath53 heavy quarkonium due to the small transverse size of the @xmath0 .
however , for the photoproduction of the radially excited vector mesons , it is not clear whether the small-@xmath43 expansion of eq .
( [ dipole ] ) is valid .
since the size of @xmath8 turns out to be about @xmath54 , twice as large as that of @xmath51 , in the typical non - relativistic quark model , the color dipole expansion may not work in this case .
namely , there may exist non - negligible contributions to the integral eq .
( [ amp1 ] ) > from an overlap of the large - size @xmath6 configuration and the @xmath1 wave function . in this work
we explicitly evaluate eq .
( [ amp1 ] ) instead of using the simplified formula ( [ dipole ] ) .
to do this , we need the explicit form of the unintegrated gluon density @xmath55 , which is not precisely known . here
, we follow the prescription of refs .
@xcite to carry out the @xmath23-integration in eq .
( [ amp1 ] ) by relating @xmath24 to the derivative of the gluon distribution ; @xmath56 where @xmath57 is a given function of @xmath23 in eq .
( [ amp1 ] ) , and @xmath58 plays a role of the infrared separation scale .
using the cteq4lq@xcite parametrization for the gluon distribution and taking @xmath59 , we obtain the result shown in fig.2 . calculation without ( with ) the dipole approximation
is depicted by the solid ( dashed ) curve . for large @xmath29
the difference between them becomes small as expected .
however , around @xmath52 , the calculation of eq .
( [ amp1 ] ) with eq.([app ] ) clearly larger than the result obtained by eq .
( [ dipole ] ) by factor about 2@xcite , and shows a reasonable agreement with the experimental data@xcite .
we demonstrate the major improvement of the ratio is caused by the contribution from the large - size @xmath13 fluctuation
. there may be several other mechanisms to enhance the @xmath1 to @xmath48 ratio further , which will be discussed elsewhere@xcite .
if we vary the value of @xmath60 from @xmath61 to @xmath62 , resulting ratio changes within a few @xmath63 .
our results are also found to be insensitive to the choice of the potential model .
we show in fig.3 results with grv98lo@xcite parametrization for the gluon distribution function ( thin - solid curve ) in comparison with those with the cteq4lq distribution ( thick - solid curve ) .
calculated ratio is slightly modified due to the difference of the gluon distributions between cteq4 and grv98 at low @xmath29 . for completeness , sensitivity of the results to @xmath64 is demonstrated in fig.3 , because several choices of @xmath64 are adopted in the literature@xcite . in fig.3 results with @xmath65 as an extreme case are shown by the thick - dashed ( cteq ) and thin - dashed ( grv ) curves , while those with @xmath66 by the solid curves .
although increasing the upper limit of the @xmath23-integration improves the result , significant enhancement at @xmath52 comes from higher order terms of @xmath67 in eq .
( [ amp1 ] ) .
we also consider the production of @xmath68-states .
one can naively expect that the transverse separation @xmath20 is small enough to ensure the validity of the dipole expansion for the bottom quark case .
results for @xmath69 and @xmath70 are shown in fig.4 .
they are consistent with those obtained in ref .
@xcite by the dipole formula .
there are only slight differences between the calculations without ( solid ) and with ( dashed ) the dipole approximation even for the @xmath71 case , whose radius is about @xmath72 . in conclusion , we have studied the diffractive photo- and leptoproduction of radially excited heavy quarkonia , and discussed the validity of the small - size color dipole approximation , which gives the smaller value of the @xmath1 to @xmath0 ratio compared with the experimental data .
it is pointed out that the contribution from the higher order terms of @xmath67 in the @xmath13-nucleon scattering amplitude , which is inherent in the original formula eq .
( [ amp1 ] ) , is important for the radially excited @xmath1 case .
we have demonstrated that the calculation of eq .
( [ amp1 ] ) without the dipole approximation provides a much larger value for the elastic photoproduction ratio @xmath73 than the dipole result , and is consistent with the hera data . as claimed in ref .
@xcite , relativistic treatment of the charmonium system may be of some importance to construct the @xmath1 wave function .
however , what we have shown here is that , even within our non - relativistic framework where there are some ambiguities of choosing the kinematics _
e.g_. in eq .
( [ momentum ] ) , substantial improvement for the @xmath1 production can be achieved by evaluating eq .
( [ amp1 ] ) instead of eq .
( [ dipole ] ) .
our results suggest that a careful treatment is necessary to study the radially excited light vector meson leptoproduction , _
e.g_. @xmath74 at moderate @xmath29 .
we have also discussed the @xmath75 production processes , and found that eq .
( [ dipole ] ) is a good approximation for the @xmath76 system because of its heavier quark mass .
we thank dr .
d. jido for useful discussions on the early stage of the present work .
a.h . are supported by jsps research fellowships for young scientists .
t. h. are partially supported by jsps grant no .
m. ryskin , z. phys . * c57 * , 89 ( 1993 ) s. brodsky _ et al_. , phys
. rev . * d50
* , 3134 ( 1994 ) l. frankfurt , w. koepf and m. strikman , phys . rev .
* d57 * , 512 ( 1998 ) m. ryskin , r. roberts , a. martin , e. levin , z. phys . *
c76 * , 231 ( 1997 ) a. martin , m. ryskin , t. teubner , phys .
rev . * d55 * , 4329 ( 1997 ) j. nemchik _
et al_. , jetp * 86 * , 1054 ( 1998 ) a. donnachie and p. v. landshoff , phys .
* b470 * , 243 ( 1999 ) m. wsthoff and a. martin , j. phys . * g25 * , r309 ( 1999 ) and references therein p. hoyer and s. peign , phys . rev .
* d61 * , 031501(r ) ( 2000 ) c. adolf _ et al_. ( h1 collaboration ) , eur .
j. * c10 * , 373 ( 1999 ) j. r. forshaw and d. a. ross , _ `` quantum chromodynamics and the pomeron '' _ , ( cambridge university press , london , 1997 ) w. buchmller and s .- h .
tye , phys .
* d24 * , 132 ( 1981 ) c. quigg and j.l .
rosner , phys .
* b71 * , 153 ( 1977 ) a. hayashigaki _
et al_. , in preparation e. levin _ et al_. , z. phys .
* c74 * , 671 ( 1997 ) h.l .
et al_. , phys .
* d55 * , 1280 ( 1997 ) we have also calculated eq .
( [ amp1 ] ) using the ` linear ' approximation@xcite for the unintegrated gluon density , instead of the approximation adopted in eq .
( [ app ] ) . resulting @xmath1 to @xmath48 ratio at @xmath77 increases about @xmath78 compared with the result shown in fig.2 , and comes closer to the data .
more realistic approximation for @xmath24 may improve our result .
the result discussed in this paper should be considered as a lower bound of the ratio obtained by the calculation of eq .
( [ amp1 ] ) without the dipole approximation .
m. glck , e. reya and a. vogt , eur .
j. * c5 * , 461 ( 1998 ) l. frankfurt , m. mcdermott and m. strikman , jhep * 9902 * , 002 ( 1999 ) | we study the diffractive photo- and leptoproductions of @xmath0 and @xmath1 on the proton , and examine the validity of the small - size color dipole approximation to the production of radially excited heavy quarkonium .
the dipole model predicts a small ratio of @xmath1 to @xmath0 photoproduction cross sections , which does not agree with experimental data .
we show that this discrepancy originates from a large transverse size of @xmath1 which makes the convergence of the transverse size expansion questionable , and the calculation without the dipole approximation turns out to be consistent with the data .
productions of @xmath2 are also discussed , and the dipole approximation is found to be reasonable for the @xmath3-family . |
we assume familiarity with the basic concepts about graphs . + as usual , we denote by @xmath5 and @xmath6 the _ complete graph on @xmath7 vertices _ and the _
complete @xmath8-partite graph with parts of size @xmath9 _ , respectively . also , let @xmath10 , @xmath11 , be the cycle on @xmath12 vertices and let @xmath13 , @xmath14 , be the path on @xmath8 vertices .
graphs of the form @xmath15 can be viewed as grids on cylinders and they are bipartite if and only if @xmath12 is even . if @xmath16 , @xmath17 is nothing but the prism @xmath18 on @xmath19 vertices . for any graph @xmath20
we write @xmath21 for the set of its vertices and @xmath22 for the set of its edges .
if @xmath23 , we say that @xmath20 has _ size _ @xmath24 . given a subgraph @xmath20 of a graph @xmath25 , a @xmath20-_decomposition of _
@xmath25 is a set of graphs , called _ blocks _ , isomorphic to @xmath20 whose edges partition the edge - set of @xmath25 .
such a decomposition is said to be _ cyclic _ when it is invariant under a cyclic permutation of all vertices of @xmath25 . in the case
that @xmath26 one also speaks of a @xmath20-_system of order _
the problem of establishing the set of values of @xmath7 for which such a system exists is in general quite difficult . for a survey on graph decompositions
see @xcite .
the concept of a _ graceful labeling _ of @xmath20 , introduced by a. rosa @xcite , is quite related to the existence problem of cyclic @xmath20-systems . a _ graceful labeling _ of a graph @xmath20 of size @xmath24 is an injective function @xmath27 such that @xmath28\in e({\gamma})\}=\{1,2,\ldots , e\}.\ ] ] in the case that @xmath20 is bipartite and @xmath29 has the additional property that its maximum value on one of the two bipartite sets does not reach its minimum on the other one , one says that @xmath29 is an @xmath1-labeling . in @xcite
, rosa proved that if a graph @xmath20 of size @xmath24 admits a graceful labeling then there exists a cyclic @xmath20-system of order @xmath30 and that if it admits an @xmath1-labeling then there exists a cyclic @xmath20-system of order @xmath31 for any positive integer @xmath9 . for a very rich survey on graceful labelings
we refer to @xcite .
many variations of graceful labelings have been considered .
in particular gnana jothi @xcite defines an _ odd graceful labeling _ of a graph @xmath20 of size @xmath24 as an injective function @xmath32 such that @xmath28\in e({\gamma})\}=\{1,3,5,\ldots,2e-1\}.\ ] ] in a recent paper , see @xcite , we have introduced the following new definition which is , at the same time , a generalization of the concepts of a graceful labeling ( when @xmath33 ) and of an odd graceful labeling ( when @xmath34 ) .
let @xmath20 be a graph of size @xmath35 . a @xmath0-_divisible graceful labeling _
of @xmath20 is an injective function @xmath36 such that @xmath37\in e({\gamma})\ } & = \{1,2,3,\ldots , d(m+1)\}\\ & \quad - \{m+1,2(m+1),\ldots , d(m+1)\}.\end{aligned}\ ] ] namely the set @xmath38\in e({\gamma})\ } $ ] can be divided into @xmath0 parts @xmath39 where @xmath40 for any @xmath41 .
the @xmath1-labelings can be generalized in a similar way .
a @xmath0-_divisible graceful _ @xmath1-_labeling _ of a bipartite graph @xmath20 is a @xmath0-divisible graceful labeling of @xmath20 having the property that its maximum value on one of the two bipartite sets does not reach its minimum value on the other one .
we have to point out that in @xcite the above labelings have been called `` @xmath0-graceful ( @xmath1-)labelings '' , but the author was unaware that this name is already used in the literature whit a different meaning , see @xcite and @xcite .
+ + it is known that there is a close relationship between graceful labelings and difference families , see @xcite . in @xcite
we established relations between @xmath0-divisible graceful ( @xmath1-)labelings and a generalization of difference families introduced in @xcite , proving the following theorems .
if there exists a @xmath0-divisible graceful labeling of a graph @xmath20 of size @xmath24 then there exists a cyclic @xmath20-decomposition of @xmath42 .
[ decomp ] if there exists a @xmath0-divisible graceful @xmath1-labeling of a graph @xmath20 of size @xmath24 then there exists a cyclic @xmath20-decomposition of @xmath43 for any integer @xmath44 . in this paper
we determine the existence of @xmath0-divisible graceful @xmath1-labelings of @xmath2 for several values of @xmath0 . in order to obtain these results ,
first of all we will find @xmath0-divisible graceful @xmath1-labelings of prisms , which correspond to the case @xmath16 , and then by induction on @xmath8 we will be able to construct @xmath0-divisible graceful @xmath1-labelings of @xmath2 for any @xmath45 . for what said above , these results allow us to obtain new infinite classes of cyclic decompositions of the complete multipartite graph in copies of @xmath2 .
in this section we will investigate the existence of @xmath0-divisible graceful @xmath1-labelings of prisms . from now on , given two integers @xmath46 and @xmath47 , by @xmath48 $ ] we will denote the set of integers @xmath49 such that @xmath50 . + for convenience , we denote the @xmath19 vertices of @xmath18 by @xmath51 ; @xmath52 @xmath53 where the @xmath54 s are the consecutive vertices of one @xmath12-cycle and the @xmath55 s are consecutive vertices of the other @xmath12-cycle and @xmath54 is connected to @xmath55 .
clearly @xmath18 has size @xmath56 and it is bipartite if and only if @xmath12 is even . in @xcite frucht and gallian
proved that @xmath18 admits an @xmath1-labeling if and only if @xmath12 is even .
[ prism3 ] the prism @xmath57 admits a @xmath58-divisible graceful @xmath1-labeling for every @xmath3 .
we set @xmath59 , @xmath60 , @xmath61 @xmath62 , @xmath63 @xmath64 .
clearly @xmath65 and @xmath66 are the two bipartite sets of @xmath67 .
+ consider the map @xmath68 defined as follows : @xmath69\\ \textrm{for}\ i \in \left [ k+1,2k-1 \right ] \end{array}\\ f(x_{2i})&=4k+i & & \begin{array}{l } \textrm{for}\ i\in [ 1,2k ] .
\end{array}\\ f(y_{2i+1})&=i & & \begin{array}{l } \textrm{for}\ i\in [ 0,2k-1 ] \end{array}\\ f(y_{2i})&=\left\ { \begin{array}{l } 12k+3-i\\ 12k+2-i \end{array } \right . & & \begin{array}{l } \textrm{for}\ i \in \left [ 1,k \right]\\ \textrm{for}\ i \in \left [ k+1,2k \right ] .
\end{array}\end{aligned}\ ] ] we have @xmath70\cup[4k+1,6k]\\ f({{\cal o}_x}\cup{{\cal e}_y})&=[6k+1,7k]\cup[7k+2,8k+1]\cup[10k+2,11k+1]\cup\\ & \cup[11k+3,12k+2].\end{aligned}\ ] ] hence @xmath29 is injective and @xmath71 .
now for @xmath72 set @xmath73 where the indices are understood modulo @xmath74 . by a direct calculation
, one can see that @xmath75\\ \{\sigma_i\ |\
i=2k+2,\ldots,4k\ } & = [ 1,2k-1]\\ \rho_1 & = 6k+1\\ \{\rho_i\ |\ i=2,\ldots,2k+1\ } & = [ 6k+2,8k+1],\\ \{\rho_i\ |\ i=2k+2,\ldots,4k\ } & = [ 4k+2,6k],\\ \{{\varepsilon}_i\ |\ i=1,\ldots,2k\ } & = [ 10k+3,12k+2],\\ \{{\varepsilon}_i\ |\ i=2k+1,\ldots,4k-1\ } & = [ 8k+3,10k+1],\\ { \varepsilon}_{4k } & = 10k+2.\end{aligned}\ ] ] hence @xmath76 $ ] , @xmath77 $ ] and @xmath78 $ ] . this concludes the proof .
@xmath79 [ prism6 ] the prism @xmath57 admits a @xmath80-divisible graceful @xmath1-labeling for every @xmath3 . proof . set @xmath81 as in the proof of previous theorem .
consider the map @xmath82 defined as follows : @xmath83\\ \textrm{for}\ i \in \left [ k+1,2k-1 \right ] \end{array}\\ f(x_{2i})&=4k+1+i & & \begin{array}{l } \textrm{for}\ i\in [ 1,2k ] .
\end{array}\\ f(y_{2i+1})&=i & & \begin{array}{l } \textrm{for}\ i\in [ 0,2k-1 ] \end{array}\\ f(y_{2i})&=\left\ { \begin{array}{l } 12k+6-i\\ 12k+4-i \end{array } \right .
& & \begin{array}{l } \textrm{for}\ i \in \left [ 1,k \right]\\ \textrm{for}\ i \in \left [ k+1,2k \right ] .
\end{array}\\\end{aligned}\ ] ] it results @xmath70\cup[4k+2,6k+1]\\ f({{\cal o}_x}\cup{{\cal e}_y})&=[6k+2,7k+1]\cup[7k+4,8k+3]\cup[10k+4,11k+3]\cup\\ & \cup[11k+6,12k+5].\end{aligned}\ ] ] hence @xmath29 is injective and @xmath71 .
let @xmath84 , for @xmath72 , be as in ( [ ers ] ) .
it is not hard to see that @xmath85\cup[2k+2,4k+1]\\ \{\rho_i\ |\ i=1,\ldots,4k\ } & = [ 4k+3,6k+2]\cup[6k+4,8k+3]\\ \{{\varepsilon}_i\ |\ i=1,\ldots,4k\ } & = [ 8k+5,10k+4]\cup[10k+6,12k+5].\end{aligned}\ ] ] hence @xmath29 is a @xmath80-divisible graceful @xmath1-labeling of @xmath57
. @xmath79 [ prism12 ] the prism @xmath57 admits a @xmath86-divisible graceful @xmath1-labeling for every @xmath3 . proof . also here
we set @xmath81 as in the proof of theorem [ prism3 ] .
we are able to prove the existence of a @xmath86-divisible graceful @xmath1-labeling of @xmath57 by means of two direct constructions where we distinguish the two cases : @xmath12 even and @xmath12 odd .
+ + case 1 : @xmath12 even .
+ consider the map @xmath87 defined as follows : @xmath88 8k+8-i\\[2pt ] 8k+7-i\\[2pt ] 8k+5-i \end{array } \right .
& & \begin{array}{l } \textrm{for}\ i=0\\[2pt ] \textrm{for}\ i \in \left [ 1,\frac{k}{2 } \right]\\[2pt ] \textrm{for}\ i \in \left [ \frac{k}{2}+1,k \right]\\[2pt ] \textrm{for}\
i \in \left[k+1,2k-1 \right ] \end{array}\\ f(x_{2i})&=\left\ { \begin{array}{l } 4k+3+i\\[2pt ] 4k+4+i \end{array } \right . & & \begin{array}{l } \textrm{for}\ i\in \left[1,\frac{3k}{2}\right]\\[2pt ] \textrm{for}\ i\in \left[\frac{3k}{2}+1,2k\right ] .
\end{array}\\ f(y_{2i+1})&=\left\ { \begin{array}{l } i\\[2pt ] i+1 \end{array } \right . & & \begin{array}{l } \textrm{for}\ i\in \left[0,\frac{3k}{2}-1\right]\\[2pt ] \textrm{for}\ i\in \left[\frac{3k}{2},2k-1\right ] \end{array}\\ f(y_{2i})&=\left\ { \begin{array}{l } 12k+12-i\\[2pt ] 12k+11-i\\[2pt ] 12k+9-i \end{array } \right . & & \begin{array}{l } \textrm{for}\ i \in \left [ 1,\frac{k}{2 } \right]\\[2pt ] \textrm{for}\ i \in \left [ \frac{k}{2}+1,k \right]\\[2pt ] \textrm{for}\ i \in \left[k+1,2k \right ] \end{array}\end{aligned}\ ] ] it is easy to see that @xmath89 \cup \left[\frac{3k}{2}+1,2k\right]\\ f({{\cal e}_x } ) & = \left[4k+4,\frac{11k}{2}+3\right ] \cup \left[\frac{11k}{2}+5,6k+4\right]\\ f({{\cal o}_x } ) & = \left[6k+5,7k+4\right ] \cup \left[7k+7,\frac{15k}{2}+6\right ] \cup \left[\frac{15k}{2}+8,8k+7\right]\\ f({{\cal e}_y } ) & = \left[10k+9,11k+8\right ] \cup \left[11k+11,\frac{23k}{2}+10\right ] \cup \left[\frac{23k}{2}+12,12k+11\right].\end{aligned}\ ] ] hence @xmath29 is injective and @xmath90 . set @xmath91 , for @xmath72 , as in ( [ ers ] ) . by a long and tedious calculation , one can see that @xmath92-\{k+1,2k+2,3k+3\}\\
\{\rho_i\ |\ i=1,\ldots,4k\ } & = [ 4k+5,8k+7]-\{5k+5,6k+6,7k+7\}\\ \{{\varepsilon}_i\ |\ i=1,\ldots,4k\ } & = [ 8k+9,12k+11]-\{9k+9,10k+10,11k+11\}.\end{aligned}\ ] ] this concludes the proof of case 1 . + + case 2 : @xmath12 odd .
+ let now @xmath87 defined as follows : @xmath88 8k+8-i\\[2pt ] 8k+6-i\\[2pt ] 8k+5-i \end{array } \right .
& & \begin{array}{l } \textrm{for}\ i=0\\[2pt ] \textrm{for}\ i \in \left [ 1,k \right]\\[2pt ]
\textrm{for}\ i \in \left [ k+1,\frac{3k-1}{2 } \right]\\[2pt ] \textrm{for}\ i \in \left[\frac{3k+1}{2},2k-1 \right ] \end{array}\\ f(x_{2i})&=\left\ { \begin{array}{l } 4k+3+i\\[2pt ] 4k+4+i \end{array } \right . & & \begin{array}{l } \textrm{for}\ i\in \left[1,\frac{k+1}{2}\right]\\[2pt ] \textrm{for}\ i\in \left[\frac{k+3}{2},2k\right ] .
\end{array}\\ f(y_{2i+1})&=\left\ { \begin{array}{l } i\\[2pt ] i+1 \end{array } \right . & & \begin{array}{l } \textrm{for}\ i\in \left[0,\frac{k-1}{2}\right]\\[2pt ]
\textrm{for}\ i\in \left[\frac{k+1}{2},2k-1\right ] \end{array}\\ f(y_{2i})&=\left\ { \begin{array}{l } 12k+12-i\\[2pt ] 12k+10-i\\[2pt ] 12k+9-i \end{array } \right . & & \begin{array}{l } \textrm{for}\ i \in \left [ 1,k \right]\\[2pt ] \textrm{for}\ i \in \left[k+1,\frac{3k-1}{2 } \right]\\[2pt ] \textrm{for}\ i \in \left[\frac{3k+1}{2},2k \right ] \end{array}\end{aligned}\ ] ] arguing exactly as in case 1 , one can check that @xmath29 is a @xmath86-divisible graceful @xmath1-labeling of @xmath57 .
@xmath79 + the three graphs in figure [ t24 ] show the @xmath58-divisible graceful @xmath1-labeling , the @xmath80-divisible graceful @xmath1-labeling and the @xmath86-divisible graceful @xmath1-labeling of @xmath93 provided by previous theorems . ,
title="fig:",scaledwidth=27.0% ] , title="fig:",scaledwidth=27.0% ] , title="fig:",scaledwidth=27.0% ]
in this section using the results of the previous one we will construct @xmath0-divisible graceful @xmath1-labelings of @xmath2 . in particular , since @xmath94 we consider @xmath95 . in @xcite
jungreis and reid proved that for any @xmath96 not both odd there exists an @xmath1-labeling of @xmath97 .
+ for convenience , we denote the vertices of @xmath2 as illustrated in figure [ c4p4ij ] and we set @xmath98 for any @xmath99 . proof .
we will prove the result by induction on @xmath8 .
if @xmath16 the thesis follows from theorem [ prism3 ] .
let now @xmath4 .
suppose that there exists a @xmath100-divisible graceful @xmath1-labeling @xmath29 of @xmath2 with vertices of @xmath101 so labeled : @xmath102 note that the @xmath58-divisible graceful @xmath1-labeling of @xmath103 constructed in theorem [ prism3 ] has this property , in fact @xmath104 .
so in order to obtain the thesis it is sufficient to construct a @xmath105-divisible graceful @xmath1-labeling @xmath106 of @xmath107 satisfying the same property , namely such that @xmath108 we set @xmath109 by the hypothesis on @xmath110 it results @xmath111 so there exists @xmath112 $ ] such that @xmath113 .
we set @xmath114 as in ( [ gc(m+1 ) ] ) where @xmath115 .
+ now we will see that @xmath116 defined as above is indeed a @xmath105-divisible graceful @xmath1-labeling of @xmath107 .
since @xmath117 $ ] , by the definition of @xmath106 , it follows that @xmath118.\ ] ] also we have @xmath119\cup[(4k+1)(2m+1)-(2k+1),(4k+1)(2m+1)-1].\ ] ] hence @xmath106 is an injective function .
since , by hypothesis @xmath29 is a @xmath100-divisible graceful @xmath1-labeling of @xmath2 , we have @xmath120 with @xmath121 .
let @xmath122 . by the construction
, it follows that @xmath123\\ g(d)&\subseteq(f(b)+(4k+1))\cup[(4k+1)(2m+1)-(2k+1),(4k+1)(2m+1)-1]\end{aligned}\ ] ] hence @xmath124 .
now we have to consider the differences between adjacent vertices . since @xmath29 is a @xmath100-divisible graceful @xmath1-labeling of @xmath2 , by the construction of @xmath106 , it results @xmath125\\ j\in[1,4k]\end{subarray}}|f((i , j))-f((i+1,j))|\ \cup \bigcup_{\begin{subarray}{c } i\in[1,m]\\ j\in[1,4k]\end{subarray } } |f((i , j))-f((i , j+1))|=\ ] ] @xmath126-\{\beta(4k+1)\ |\ \beta\in[1,2m-1]\}\ ] ] where the index @xmath127 is taken modulo @xmath74 . finally it is not hard to check that @xmath128\}=\\ & [ ( 4k+1)2m-4k,(4k+1)2m-1]\end{aligned}\ ] ] and @xmath129\}=\\ & [ ( 4k+1)(2m+1)-4k,(4k+1)(2m+1)-1].\end{aligned}\ ] ] this concludes the proof .
@xmath79 in figure [ c4 ]
we will show the @xmath130-divisible graceful @xmath1-labeling of @xmath131 , the @xmath132-divisible graceful @xmath1-labeling of @xmath133 and the @xmath134-divisible graceful @xmath1-labeling of @xmath135 obtained starting from the @xmath58-divisible graceful @xmath1-labeling of @xmath136 and following the construction illustrated in the proof of theorem @xmath137 . proof
. we will prove the result by induction on @xmath8 .
if @xmath16 the thesis follows from theorem [ prism6 ] .
let now @xmath4 .
suppose that there exists a @xmath138-divisible graceful @xmath1-labeling @xmath29 of @xmath2 with vertices of @xmath101 so labeled : @xmath139 we want to show the existence of a @xmath140-divisible graceful @xmath1-labeling @xmath106 of @xmath107 satisfying the same property , namely such that @xmath141 first of all set @xmath142 this implies that there exists @xmath112 $ ] such that @xmath143 .
we set @xmath114 as in ( [ 2gc(m+1 ) ] ) where @xmath115 . arguing exactly as in the previous proof one can prove that @xmath106 is a @xmath140-divisible graceful @xmath1-labeling @xmath106 of @xmath107 .
@xmath79 proof .
we will prove the result by induction on @xmath8 .
if @xmath16 the thesis follows from theorem [ prism12 ] . let now @xmath4 .
we have to distinguish two cases : @xmath12 even and @xmath12 odd .
+ let @xmath12 be even .
suppose that there exists a @xmath144-divisible graceful @xmath1-labeling @xmath29 of @xmath2 with vertices of @xmath101 so labeled : @xmath145 we want to show the existence of a @xmath146-divisible graceful @xmath1-labeling @xmath106 of @xmath107 satisfying the same property , namely such that @xmath147 first of all set @xmath148 this implies that there exists @xmath112 $ ] such that @xmath149 .
we set @xmath114 as in ( [ 4gc(m+1 ) ] ) where @xmath115 .
+ let now @xmath12 be odd .
suppose that there exists a @xmath144-divisible graceful @xmath1-labeling @xmath29 of @xmath2 with vertices of @xmath101 so labeled : @xmath150 we want to show the existence of a @xmath146-divisible graceful @xmath1-labeling @xmath106 of @xmath107 satisfying the same property , namely such that @xmath151 first of all set @xmath148 this implies that there exists @xmath112 $ ] such that @xmath149 .
we set @xmath114 as in ( [ 4gc(m+1)odd ] ) where @xmath115 .
+ arguing exactly in the proof of theorem [ d=2m-1 ] one can prove that , in both cases , @xmath106 is a @xmath146-divisible graceful @xmath1-labeling @xmath106 of @xmath107 .
@xmath79 + in figure [ c12p3 ] we have the @xmath152-divisible graceful @xmath1-labeling of @xmath153 obtained starting from the @xmath80-divisible graceful @xmath1-labeling of @xmath154 shown in figure [ t24 ] and following the construction explained in the proof of theorem @xmath155 and the @xmath156-divisible graceful @xmath1-labeling of @xmath153 obtained starting from the @xmath86-divisible graceful @xmath1-labeling of @xmath93 shown in figure [ t24 ] and following the construction illustrated in the proof of theorem @xmath157 .
-divisible graceful @xmath1-labeling of @xmath153 and a @xmath156-divisible graceful @xmath1-labeling of @xmath153 , respectively.,title="fig:",scaledwidth=40.0%]-divisible graceful @xmath1-labeling of @xmath153 and a @xmath156-divisible graceful @xmath1-labeling of @xmath153 , respectively.,title="fig:",scaledwidth=40.0% ] 999 r.j.r .
abel and m. buratti , _ difference families _ , in : crc handbook of combinatorial designs ( c.j .
colbourn and j.h .
dinitz eds . ) , crc press , boca raton , fl ( 2006 ) , 392409 .
d. bryant and s. el - zanati , _ graph decompositions _ , in : crc handbook of combinatorial designs ( c.j .
colbourn and j.h .
dinitz eds . ) , crc press , boca raton , fl ( 2006 ) , 477486 .
m. buratti and a. pasotti , _ graph decompositions with the use of difference matrices _ , bull .
* 47 * ( 2006 ) , 2332 .
r. frucht and j.a .
gallian , _ labeling prisms _ , ars combin . , * 26 * ( 1988 ) , 6982 . | in @xcite the concept of a @xmath0-divisible graceful @xmath1-labeling has been introduced as a generalization of classical @xmath1-labelings and it has been shown how it is useful to obtain certain cyclic graph decompositions . in the present paper
it is proved the existence of @xmath0-divisible graceful @xmath1-labelings of @xmath2 for any integers @xmath3 , @xmath4 for several values of @xmath0 .
* keywords : * graceful labeling ; @xmath1-labeling ; graph decomposition . + * msc(2010 ) : * 05c78 . |
there are several independent ways to determine the global cosmological parameters that describe the current state of the universe .
it is important to have several complementary and independent methods that yield consistent results since any given method could be plagued by unknown systematic errors .
a particularly useful way to determine global cosmological parameters is through measurements of the coordinate distance to high - redshift sources .
this method is particularly useful because the coordinate distance depends only upon global cosmological parameters , and is independent of the clustering properties of the mass - energy components that control the expansion rate of the universe , as long as each component is homogeneous and isotropic on very large scales .
the coordinate distance is also independent of whether different components cluster differently , known as biasing , and on whether the dark matter is cold , warm , or hot ( though it does depend on the equation of state of each component ) .
two cosmological tools that are particularly sensitive to the coordinate distance as a function of redshift are friib radio galaxies , which provide a modified standard yardstick ( daly 1994 , 2002a ; guerra 1997 ; guerra & daly 1998 [ gd98 ] ; guerra , daly , & wan 2000 [ gdw00 ] ) , and type ia supernovae , which provide a modified standard candle ( e.g. riess et al . 1998 , perlmutter et al .
. some aspects of the methods are compared in
[ compare ] .
the coordinate distance to a source at a given redshift depends on the present value of the mean mass - energy density of each component , and on the redshift evolution of the mass - energy density of each component .
non - relativistic matter , including baryonic matter and clustered dark matter , have a mean mass - energy density that evolves as @xmath5 .
there is a contribution from radiation and neutrinos left over from the big bang , currently negligible , which has an energy - density that evolves as @xmath6 .
there may also be a cosmological constant , which constant energy density , and so evolves as @xmath7 .
the redshift evolution of the mean mass - energy density of a component depends on the equation of state @xmath8 of the component .
there could exist a dynamical vaccum energy density called quintessence ( caldwell , dave , & steinhardt 1998 ) .
quintessence would have an equation of state @xmath1 such that the density evolves as @xmath9 , where @xmath10 ( see turner & white 1997 ; bludman & roos 2001 , wang et al .
2000 ; or the appendix of this paper ) . for example
, pressure - less dust ( i.e. , non - relativistic matter ) is described by @xmath11 and @xmath12 ; a relativistic fluid is described by @xmath13 and @xmath14 , and a cosmological constant is described by @xmath15 and @xmath16 .
two components of the universe are known to exist with certainty : there is a non - relativistic component ( including baryons and clustered dark matter ) , with zero - redshift , normalized mean mass - energy density @xmath2 , and a primordial relativistic component that is negligible for the epochs of interest here ( including microwave background photons and neutrinos ) .
there must be at least one more component , which remains to be determined .
this third component could be a cosmological constant , quintessence , space curvature , or something else . in the simplest case ,
there is only one additional component that is significant and has a measurable impact over the redshift interval where the coordinate distance is used to constrain cosmological parameters and the properties of the unknown component . here
, the determination of the coordinate distances to 20 radio galaxies with redshifts between zero and two are used to constrain the equation of state and current mean mass - energy density of quintessence assuming a spatially flat universe as indicated by measurements of the microwave background radiation ( de bernardis et al .
2000 , balbi et al . 2000 ) .
the radio galaxy method is briefly reviewed in [ radgals ] .
constraints on cosmological parameters and the equation of state of quintessence obtained using friib radio galaxies as a modified standard yardstick are presented in [ quint ] .
the supernovae and radio galaxy methods and results are compared in detail in [ compare ] . the relation between beam power and total energy for friib radio jets
is discussed in [ beta ] .
here it is shown that the underlying hypotheses of the radio galaxy method is consistent with current models of large - scale jet production in agn .
the results are summarized in [ summary ] .
the use of friib radio galaxies as a modified standard yardstic to determine global cosmological parameters has been presented in detail elsewhere ; see daly ( 1994 , 2002a ) , gd98 , and gdw00 .
radio images of the 20 sources used here can be found in leahy , muxlow , & stephens ( 1989 ) , liu , pooley , & riley ( 1992 ) , and gdw00 .
the radio galaxy method is based on the following premises .
first , it is assumed that the forward region of the radio bridge of an friib source represents a strong shock front , and the equations of strong shock physics are applied at this boundary .
second , it is assumed that there is a power - law relation between the average beam power of the source @xmath17 and the total lifetime of the outflow @xmath18 : @xmath19 ( see [ beta ] ) .
third , it is assumed that , at a given redshift , the sources studied will have an average source size similar to that of the parent population of friib sources at the same redshift , where the source size is defined as the separation between the radio hot spots .
all of these points are discussed in detail in previous work ( e.g. daly 2002a , b ; gdw00 ) .
in addition , in its simplest interpretation , the model works in a straight - forward way if the radio power of a given friib radio galaxy does not decrease monotonically and measurably with time .
there is substantial evidence that this is the case for radio galaxies that fall into the friib category , as do all of the radio galaxies considered here ( daly 2002b ) .
for example , neeser et al .
( 1995 ) find that there is no correlation between radio size and radio power .
the low frequency radio surface brightness of these sources can be explained using adiabatic expansion alone if the radio powers of the sources are roughly time - independent ( wellman , daly , & wan 1997 ) .
gopal - krishna , kulkarni , & wiita ( 1996 ) find that a roughly constant radio power followed by an exponential drop in radio power would explain the relative sizes of radio galaxies and radio loud quasars in the context of the orientation unified model . and
, it is shown in [ beta ] that the beam power is independent of core - hot spot separation , which would not be expected if the radio power of a given source decreases with time .
thus , there is no empirical evidence that the radio power of a given friib radio galaxy decreases with time , and substantial evidence that it is roughly constant over the lifetime of a given source , followed by a precipitous drop .
the mach numbers of friib radio galaxies are about 2.5 to 10 ( wellman , daly , & wan 1997 ) , implying overpressures of 10 to 200 , so the lobe will rapidly expand and drop in radio power when the beam power shuts off , as suggested by neeser et al .
some authors ( e.g. , blundell , rawlings , & willcot 1999 ) have studied large heterogeneous samples that include sources with all types of radio structure and power and both radio loud quasars and radio galaxies .
to explain the properties of this heterogenous sample , a model is proposed that allows for the decrease in the radio power of a given source as a function of time .
there is , however , no evidence that this applies to the select subset of sources ( the most powerful frii radio galaxies ) studied here , and substantial evidence that the radio powers of friib radio galaxies do not monotonically decline over the lifetime of a given source .
since friib radio galaxies form a remarkably homogenous population , and all the sources at a given redshift are quite similar , the average size of a given source @xmath20 will equal the average size of the full population @xmath21 of friib radio galaxies with similar redshifts .
the dependence of @xmath20 and @xmath21 on the coordinate distance , and hence on cosmological parameters , is quite different , and @xmath22 since @xmath23 ( see gd98 eq .
6 ) . by construction ,
the ratio @xmath24 should be a constant , independent of redshift . because the model is based upon a comparison of individual source characteristics with the characteristics of the parent population at the same redshift ,
selection effects , such as radio power selection effects , are unlikely to affect the constraints placed on cosmological parameters .
all of the sources for which values of @xmath20 are determined belong to the parent population .
the sources for which @xmath20 is determined are subject to selection effects that are identical to those of the parent population , used to determine @xmath21 .
thus , this method is one of the few methods that is unlikely to be affected by observational selection effects .
equation ( 1 ) allows constraints to be placed on the model parameter @xmath4 , and the cosmological parameters that enter into the determination of the coordinate distance .
recent measurements of the cosmic microwave background anisotropy suggest that the universe has zero space curvature ( de bernardis et al .
2000 , balbi et al . 2000 ) . here
, a universe with zero space curvature ( @xmath25 ) and quintessence is considered .
the coordinate distance to a source at redshift @xmath26 in a spatially flat universe is @xmath27 where @xmath28 .
the coordinate distance is obtained by substituting this into eq .
[ eqcoor1 ] and solving for @xmath29 ( see , for example , turner & white 1997 ; bludman & roos 2001 ; the appendix of this paper ) .
the equation of state @xmath1 of quintessence is assumed to be time - independent over the redshift interval from zero to two ( wang et al .
thus , the mean mass - energy density of quintessence will evolve as @xmath9 , where @xmath30 . for a present value of the normalized mass - energy density of quintessence @xmath0 , and that of non - relativistic matter @xmath2 , we have @xmath31 at @xmath32 , since space curvature is taken to be zero .
this implicitly assumes that there is only one unidentified component , @xmath0 , that contributes significantly over the redshift interval where the coordinate distance will be determined .
the minimization of the chi - squared of @xmath33 constant leads to constraints on @xmath2 , @xmath1,and the model parameter @xmath4 .
the one and two - dimensional confidence intervals have been determined in the usual way ( see , for example , press , teukolsky , & flannery 1992 ) .
figure 1 shows the 68% , 90% , and 95% one - dimensional confidence contours in the @xmath3 plane . at 90% confidence
, @xmath2 is less than about 0.5 for any value of @xmath1 .
a cosmological constant , represented by @xmath15 is clearly consistent with the radio data .
this figure differs slightly from the two - dimensional confidence intervals presented in figure 2 of daly & guerra 2001 . at 68 % confidence ,
@xmath34 , and @xmath35 . at 90 % confidence ,
@xmath36 and @xmath37 .
note , some models for the dark energy invoke mechanisms that behave like quintessence with w less than -1 ( e.g.frampton 2002 ) .
the fifty - four supernovae identified by perlmutter et al .
( 1999 ) and used to obtain their `` primary fit '' c have been similarly fitted , and imply strong contraints on @xmath2 , though the supernovae do not place a lower bound on w , as shown in figure 2 .
as described by turner & white ( 1997 ) and presented in some detail in the appendix , the expansion rate of the universe is accelerating at the present epoch if @xmath38 this line is drawn on the @xmath3 plane ( see figures 1 and 2 ) ; points below the line represent solutions for which the universe is currently accelerating , while those above the line represent solutions for which the universe is currently decelerating .
radio galaxies alone ( fig 1 ) indicate that the expansion of the universe is accelerating ( @xmath39 ) ; this is an 84% confidence result .
it is very important to consider whether significant covariance exists between the different parameters determined by the fit .
figure 3 shows the 68% , 95% , and 99% one - dimensional confidence contours in the @xmath40 plane . clearly , there is no covariance between the equation of state @xmath1 of quintessence and the model parameter @xmath4 : @xmath1 and @xmath4 are independent .
figure 4 shows the 68% , 95% , and 99% one - dimensional confidence contours in the @xmath41 plane .
clearly the two parameters are independent .
the data suggest that @xmath42 independent of the cosmological parameters @xmath2 , @xmath0 , or the equation of state @xmath1 of quintessence .
the observation of the acceleration of the universe through measurements of the coordinate distance to high redshift sources is so important that it needs to be independently verified using different methods . in
4 , it was shown that the radio galaxy method alone indicates at about 84% confidence that the universe is accelerating in its expansion at the present epoch if the universe has zero space curvature .
the supernovae teams report similar results at stronger confidence levels ( riess et al .
1998 , perlmutter et al . 1999 ) , as illustrated in figure 2 .
it is important to consider the similarities and differences between the methods to insure that the results of each are independent , and to compare the quantities that are actually determined in each method .
the two methods are compared in table 1 .
the two methods are similar is their dependence on the coordinate distance , and in the sense that both provide a modified standard rather than an absolute standard .
they are somewhat similar in their redshift interval of coverage , and in the number of sources studied .
the radio galaxy and supernova methods differ in the way the `` modification '' relation is derived ; one is empirical and one is based on physical arguments .
they also differ in that the radio galaxy method is not normalized at zero redshift , and is independent of the properties of local sources and of the local distance scale ; in fact , nearly identical results obtain when the low redshift bin is excluded from the analysis ( gdw00 ) .
the supernova relation is derived at low redshift .
thus , the radio galaxy method provides independent verification of the supernova results .
current constraints on cosmological parameters determined using the radio galaxy method are consistent with those obtained using the supernova method . clearly , any selection effects , or unknown systematic errors , must be completely different for these two methods .
the supernova method involves short - lived optical events , and the radio galaxy method involves much more long - lived radio agn events .
the methods are derived in vastly different ways , and are normalized differently .
however , similar constraints obtain , suggesting each method is reliable .
it is interesting to compare the constraints placed on cosmological parameters assuming that three main terms control the expansion of the universe at the present epoch : non - relativistic matter with contribution @xmath2 , a cosmological constant with contribution @xmath43 , and space curvature .
figures 5 and 6 show constraints placed by the microwave background radiation ( bond et al .
2000 ) , and the high - redshift supernovae project ( riess et al . 1998 ) ; the results obtained by the supernovae cosmology project ( perlmutter et al .
1999 ) are similar and are presented by daly & guerra ( 2001 ) .
each method carves out a different part of parameter space because each method covers a different redshift range with the supernovae being at the lowest redshift , the radio galaxies at higher redshift , and the microwave background radiation at still higher redshift .
the three methods are thus complementary in their redshift coverage .
the radio galaxy and supernovae methods are independent of the initial power spectrum of density fluctuations , and hence are independent of the index of the primordial fluctuation spectrum .
these methods are also independent of the hubble constant , and of the baryon fraction .
the radio galaxy and supernovae methods depend only on the global cosmological parameters @xmath2 and @xmath43 allowing for space curvature , or @xmath2 and @xmath1 allowing for quintessence , where @xmath44 , and assuming zero space curvature .
they are therefore particularly important to study since most other methods depend on several other parameters such as the initial spectrum of fluctuations , the hubble constant , and/or the baryon fraction .
the application of the radio galaxy cosmology method indicates that @xmath42 irrespective of cosmological model , as discussed here for quintessence , by gdw00 in cosmologies that include space curvature , and summarized by daly ( 2002a ) . the total energy released over the lifetime of the friib in the form of a highly collimated outflow is @xmath45 assuming that @xmath17 is roughly constant over the lifetime of a given source , which is supported by lack of correlation between @xmath17 and the hotspot to core distances in these sources ( see wan , daly , & guerra 2000 ; guerra & daly 2001 ) .
values of @xmath17 and @xmath18 are obtained independently , using different aspects of the radio data for a given source .
two values are determined for each source , one on each side of each source .
an indicator of possible problems with the model assumptions or selection biases is whether the beam powers , total source lifetimes , or total energies depend strongly on the distance of the radio hot spot from the location of the agn .
each of these quantities is plotted as a function of hot spot distance from the agn in figures 7 , 8 , and 9 .
there are 40 data points on each figure , two for each of the 20 radio galaxies studied here .
it is interesting to note that the total energy processed through large - scale jets has a rather small range , from @xmath46 to @xmath47 , though it does increase systematically with redshift , as shown in figure 10 . because @xmath19 and @xmath45 , the relation between beam power and energy released by an friib is @xmath48 , where @xmath49 to 3 for @xmath50 to 2.0 . in terms of the total lifetime of the source , @xmath51 for values of @xmath4 of 1.5 , 1.75 , and 2.0 respectively .
clearly , @xmath4 remains far from the limiting cases of @xmath4 = 0 or 3 .
blandford ( 1990 ) gives a summary of the energy and beam power of outflows associated with the electromagnetic extraction of the spin energy of a rotating black hole .
the beam power and total energy available are @xmath52 and @xmath53 for @xmath54 , where @xmath55 is the mass of the black hole , @xmath56 is the mass in units of @xmath57 , @xmath58 is the spin angular momentum @xmath59 per unit mass @xmath55 : @xmath60 , c is the speed of light , @xmath61 is the gravitational radius @xmath62 , @xmath63 is the magnetic field strength , and @xmath64 is the magnetic field strength in units of @xmath65 g ( blandford 1990 ) .
empirically , @xmath66 with @xmath67
. this is consistent with equations ( 4 ) and ( 5 ) when the magnetic field strength satisfies @xmath68 this is particularly simple when @xmath50 ; in this case @xmath69 . for the cases @xmath70 and 2 , the empirical constraint is consistent with equations ( 4 ) and ( 5 ) when @xmath71 and @xmath72 respectively .
typical values for the beam power , total energy processed by a jet , and lifetime are @xmath73 , @xmath74 , and @xmath75 yr ( see figures 7 , 8 , and 9 ) .
equations [ ljeq ] and [ estareq ] imply a total source lifetime of @xmath76 , indicating a lifetime of about @xmath77 yr for @xmath78 .
this is satisfied for @xmath79 , and @xmath80 . in this case , the beam power is @xmath81 for @xmath82 , and the total energy is @xmath83 .
these values for @xmath56 , @xmath64 , and @xmath84 seem quite reasonable .
the scaling between variables required by the empirically determine relation between total energy and beam power are discussed above .
type iib radio galaxies and type ia supernovae are particularly important methods to develop to constrain the global cosmological parameters @xmath2 , @xmath43 , and space curvature , or @xmath2 , @xmath0 , and the equation of state of quintessence @xmath1 .
these methods depend only upon the properties of global cosmological parameters , and are independent of other factors such as the index of the primordial power spectrum , the hubble constant , the baryon fraction , the properties of the dark matter that clusters around galaxies and clusters of galaxies , any biasing of dark relative to luminous matter , etc .
the radio galaxy and supernova methods are completely independent and have completely different potential systematic errors , as discussed in
[ compare ] .
radio galaxies may be used to constrain the mass - energy density and equation of state of quintessence .
these results are presented here assuming a spatially flat universe , which is supported by recent measurements of fluctuations of the cosmic microwave background .
radio galaxies alone suggest that the universe is accelerating in its expansion at present ( see s [ quint ] ) , consistent with results obtained by the supernovae teams , indicating that both methods are working well and probably are not plagued by unknown systematic errors .
the implications for models of energy extraction for the cases of an outflow related to accretion and the eddington luminosity , and electromagnetic energy extraction of rotational energy are considered .
if the outflow is produced by the electromagnetic extraction of energy from a rotating black hole , then the magnetic field strength must be related to the spin angular momentum of the rotating black hole @xmath59 , the mass of the black hole @xmath55 , and the gravitational radius of the black hole @xmath61 , as described in 5 .
the relation is particularly simple if @xmath85 , and implies that the magnetic field strength satisfies @xmath69 , where @xmath86 .
we are grateful to lin wan and greg wellman who have contributed significantly our studies of radio galaxies .
we also thank megan donahue , paddy leahy , chris odea , adam riess , max tegmark , and an anonymous referee for helpful comments .
this work was supported in part by a national young investigator award ast-0096077 from the national science foundation , and by penn state university .
research at rowan university was supported in part by the college of liberal arts and sciences and national science foundation grant ast-9905652 .
balbi , a. et al . 2000 , apj , 545 , l1 begelman , m. c. , blandford , r. d. , & rees , m. j. 1984 , rev .
phys . , 56 , 255 blandford , r. d. 1990 , in active galactic nuclei , ed . t. j. l. courvoisier & m. mayor , springer - verlag , 161 bludman , s. , & roos , m. 2001 , apj , 547 , 77 blundell , k. m. , rawlings , s. , & willott , c. j. 1999 , aj , 117 , 677 bond et al .
2000 , preprint ( astro - ph/0011378 ) caldwell , r. r. , dave , r. , & steinhardt , p. j. 1998 , phys .
, 80 , 1582 daly , r. a. 1994 , apj , 426 , 38 daly , r. a. 2002a , to appear in `` the dark universe : matter , energy , and gravity , '' in press ( available at www.bk.psu.edu/faculty/daly ) daly , r. a. 2002b , to appear in `` lifecycles of radio galaxies , '' in press ( available at www.bk.psu.edu/faculty/daly ) daly , r. a. , & guerra , e. j. 2001 , in iau symposium no .
201 : new cosmological data and the values of the fundamental parameters , ed .
a. lasenby , a. w. jones , & a. wilkinson ( san francisco : asp conference series ) , in press de bernardis , p. et al .
2000 , nature , 404 , 995 fanaroff , b. l. , & riley , j. m. 1974 , mnras , 167 , 31p frampton , p. h. 2002 , in the coral gables conference proceeding gopal - krishna , kulkarni , v. k. , & wiita , p. j. 1996 , apj , 463 , l1 guerra , e. j. 1997 , phd thesis , princeton university guerra , e. j. , & daly , r. a. 1998 , apj , 493 , 536 ( gd98 ) guerra , e. j. & daly , r. a. 2001 , in particles and fields in radio galaxies , ed .
r. laing & k. blundell , ( san francisco : asp conference series ) , in press .
guerra , e. j. , daly , r. a. , & wan , l. 2000 , apj , 544 , 659 ( gdw00 ) leahy , j. p. , muxlow , t. w. , & stephens , p. w. 1989 , mnras , 239 , 401 liu , r. , poole , g. , & riley , j. m. 1992 , mnras , 257 , 545 neeser , m. j. , eales , s. a. , law - green , j. d. , leahy , j. p. , & rawlings , s. 1995 , apj , 451 , 76 perlmutter et al . 1999 ,
apj , 517 , 565 press , w. h. , teukolsky , s. a. , & flannery , b. p 1992 , numerical recipes in c , cambridge university press riess et al .
1998 , aj , 116 , 1009 turner & white 1997 , phys .
d , 56 , 4439 wan , l. , daly , r. a. , & guerra , e. j. 2000 , apj , 544 , 671 wang , l. , caldwell , r. r. , ostriker , j. p. , & steinhardt , p. j. 2000 , apj , 530 , 17 wellman , g. f. , daly , r. a. , & wan , l. 1997 , apj , 480 , 79 crrrrrrrrrrr type snia & type friib + @xmath87 & @xmath88 + @xmath89 & @xmath90 + @xmath91 sources&20 sources ( 70 in parent pop . ) + modified standard candle & modified standard yardstick + light curve @xmath92 peak luminosity&radio bridge @xmath92 average length + empirical relation & physical relation + written in terms of observables&written in terms of physical variables + model parameter determined at @xmath32 & model parameter not determined at @xmath32 + may depend on local distance scale & independent of local distance scale + universe is accelerating & @xmath2 is low + & universe is acclelerating if k=0 + some theoretical understanding & good theoretical understanding + well tested empirically&needs more empirical testing +
the deceleration parameter is defined to be @xmath93 evaluated at z=0 , where @xmath58 is the cosmic scale factor .
this may also be written @xmath94 , where quantities evaluated at z=0 have a subscript o , and @xmath95 .
it is straight - forward to show that ( e.g. peebles [ 1993 ] eq .
5.15 ) @xmath96 where @xmath97 is the pressure , @xmath98 is the mean mass - energy density , @xmath99 is the equation of state of @xmath100 component , @xmath101 , and the quintessence or cosmological constant term is included in the summation . for each component , mass - energy conservation implies @xmath102 ( e.g. peebles [ 1993 ] eqs .
5.16 ) . by definition ,
the equation of state is @xmath103 , so the eq .
[ conseq ] implies @xmath104 when the equation of state @xmath99 is time independent , the solution to this equation is @xmath105 , where @xmath106 .
thus , a component with equation of state @xmath99 and present mean mass - energy density @xmath107 will have a mean mass - energy density at redshift @xmath26 of @xmath108 , where @xmath109 . when @xmath25 , @xmath110 where @xmath111 .
it follows that @xmath112 . for @xmath25 , @xmath113 where @xmath114 is the critical density at redshift zero .
thus , @xmath115 for @xmath25 .
the deceleration parameter @xmath116 then becomes @xmath117 where @xmath118 and equations [ acceq ] and [ hnot ] have been used .
the universe is accelerating rather than decelerating when @xmath119 .
this can only occur if @xmath120 , or @xmath121 , which is a necessary but not sufficient condition to have an accelerating universe at the present epoch .
if there are only two significant types of mass - energy controlling the expansion rate of the universe at the present epoch , quintessence and non - relativistic matter , then the deceleration parameter is @xmath122 .
the universe will be accelerating in its expansion when @xmath123 which follows since @xmath124 .
thus , a curve can be drawn on the @xmath3 plane ; points below the curve represent solutions for which the universe is currently accelerating , while those above the curve represent solutions for which the universe is currently decelerating ( e.g. see figure 1 ) .
for example , this curve is shown on figure 1 , which illustrates the constraints obtained using radio galaxies alone , and is discussed in more detail below .
note that radio galaxies alone place interesting constraints on @xmath2 and @xmath1 , and these results are consistent with those obtained using other methods ( e.g wang et al .
2000 ) .
the coordinate distance to a source at redshift @xmath26 follows from the equation @xmath125 ( weinberg 1972 ) . for a spatially flat universe
, the left hand side of the equation reduces to the coordinate distance @xmath126 .
equations [ eqadot ] and [ hnot ] imply that @xmath127 or @xmath128 following the notation of peebles ( 1993 ) , this reads @xmath129 , where @xmath130 thus , the coordinate distance to a source at redshift @xmath26 in a spatially flat universe is ( see equation [ eqcoor ] ) @xmath27 a spatially flat universe has @xmath131 .
the components that must be included in equation [ eqeofz ] are those that contribute from redshift zero out to the redshift that the coordinate distance is being determined .
two components are considered here : non - relativistic matter with normalized mean mass - energy density at @xmath32 of @xmath2 , and quintessence with normalized mean mass - energy density at @xmath32 of @xmath0 .
thus , @xmath132 , and equation [ eqeofz ] becomes @xmath28 .
the coordinate distance is obtained by substituting this into eq .
[ eqcoor1 ] and solving for @xmath29 . | friib radio galaxies provide a modified standard yardstick that allows constraints to be placed on global cosmological parameters .
this modified standard yardstick is analogous to the modified standard candle provided by type ia supernovae .
the radio galaxy and supernova methods provide a measure of the coordinate distance to high - redshift sources , and the coordinate distance is a function of global cosmological parameters .
a sample of 20 friib radio galaxies with redshifts between zero and two are compared with the parent population of 70 radio galaxies to determine the coordinate distance to each source .
the coordinate distance determinations are used to constrain the current mean mass - energy density of quintessence @xmath0 , the equation of state of the quintessence @xmath1 , and the current mean mass - energy density of non - relativistic matter @xmath2 ; zero space curvature is assumed .
radio galaxies alone indicate that the the universe is currently accelerating in its expansion ( with 84% confidence ) ; most of the allowed parameter space falls within the accelerating universe region on the @xmath3 plane .
this provides verification of the acceleration of the universe indicated by high - redshift supernovae , and suggests that neither method is plagued by systematic errors .
it is found that @xmath2 must be less than about 0.5 and the equation of state @xmath1 of the quintessence must lie between -0.25 and -2.5 at about 90% confidence .
fits of the radio galaxy data constrain the model parameter @xmath4 , which describes a relation between the beam power of the agn and the total energy expelled through large - scale jets .
it is shown that the empirically determined model parameter is consistent with models in which the outflow results from the electromagnetic extraction of rotational energy from the central compact object . a specific relation between the strength of the magnetic field near the agn , and the spin angular momentum per unit mass of the central compact object is predicted . |
low - energy hadron - hadron scattering plays an important role in the understanding of non - perturbative physics of strong interaction .
due to its genuine non - perturbative nature , such problems can only be studied from first principles using non - perturbative methods like lattice qcd .
lscher has outlined a finite - size formalism which enables us to calculate the elastic scattering phase shifts using lattice simulations @xcite . over the years
, extensive numerical simulations have been carried out to the study on hadron - hadron scattering using lscher s formalism , both within the quenched approximations and using gauge field configurations with dynamical quarks @xcite . in lattice study on hadron spectroscopy and hadron - hadron scattering ,
the most important physical quantity is the energy of the system which is obtained via the measurements of various correlation functions .
however , since a quantum field theory does not conserve particle numbers in general , the distinction between single- and multi - particle states becomes an important and delicate issue in lattice calculations . in the infinite volume ,
the difference is obvious since they have different kinematic behaviors : single - particle states have discrete energy eigenvalues when viewed in their rest frame while multi - particle states usually have continuous spectrum starting from the corresponding threshold .
however , when performing a lattice simulation in a finite volume , all energy eigenvalues in the finite box become discrete . therefore , other means have to to be applied in order to identify the particle nature of a corresponding state . in principle
, differences between single- and multi - particle states still persist in a finite volume .
for example , although both have discrete spectra , the level spacing between neighboring multi - particle scattering states becomes infinitesimally small while the level spacing between the neighboring single - particle states remains finite as the volume goes to infinity .
however , it is difficult to utilize this difference as a practical criteria since this requires the computation of excited energy eigenvalues in monte carlo simulations which is usually quite challenging .
another method suggested by various authors is to use the so - called spectral weight as the identifier .
this is the quantity which can be measured directly ( and relatively easily ) from monte carlo simulations , together with the corresponding energy eigenvalue . in a finite volume
, the volume dependence of the spectral weight for a eigenstate is expected to show different behavior for single- and multi - particle states .
for example , one expects the following empirical rule : the spectral weight to show little volume dependence for a single particle state ( if properly normalized ) , while for a two - particle state , it is expected to show a @xmath0 dependence where @xmath1 being the size of the cubic box .
this expected difference in volume dependence can be measured in lattice simulations by performing the same calculation in two distinct volumes . as an example , this strategy has been used in ref .
@xcite to study the possible penta - quark state . using this technique
, the authors concluded that the expected penta - quark ( single - particle ) states measured in their lattice calculations are in fact kaon - nucleon two - particle scattering states .
however , this conclusion is not so settled even in the first - principle lattice qcd calculations @xcite .
therefore , the volume dependence of the spectral weight indeed can provide us useful information about the particle nature of the corresponding state . in a previous model study
, we have shown that the above mentioned empirical rule to distinguish single- and multi - particle states are in fact only valid for stable particles and narrow resonances . using a solvable model , the lee model
, we showed that this rule is violated for broad resonances @xcite .
a general formula for the spectral weight was obtained which can show either single- or two - particle volume behavior depending whether the width of the resonance is narrow or broad . in this paper , we attempt to generalize this conclusion that we obtained in the lee model , to the case of general massive quantum field theory . for this purpose , the general lscher s formalism is adopted . in previous studies
, people have been focusing mainly on the energy eigenvalue ( which directly enters the famous lscher s formula ) of the system within lscher s formalism
. however , since the spectral weight @xmath2 of a given state is intimately related to the overlap of the exact energy eigenfunction with the free scattering states , we have to study the wavefunction of a energy eigenstate in a finite volume . in this paper ,
our study focuses on the wavefunction in the @xmath3 sector and a formula for the spectral weight is thus obtained within the non - relativistic quantum mechanics model . by studying the volume dependence of the spectral weight in the large volume limit , we arrive at the same conclusion as we drew from the previous lee model study .
then , following lscher s arguments , this result is generalized to massive quantum field theory .
our results also show a possibility of extracting the resonance parameters from the spectral weight function on various volumes . this paper is organized as follows . in sec .
[ sec : model ] , we briefly review the quantum - mechanical model in the infinite volume . in sec .
[ sec : model_torus ] , the quantum - mechanical model is studied on a three - dimensional torus of size @xmath1 . in this section ,
we derive the relevant formulae for the spectral weight function and study its volume dependence .
it is found that similar conclusion is reached as in our previous study using the lee model .
we then argue that , under the same restrictions as in lscher s formula , our results found in the quantum - mechanical model can be generalized to massive quantum field theory .
the possibility of extracting resonance parameters from spectral weight is also discussed . in sec .
[ sec : conclude ] , we will conclude with some general remarks .
details on the evaluation of a function @xmath4 are listed in the appendix .
consider a quantum mechanical model whose hamiltonian is given by : [ eq : hamiltonian ] h= -12m^2+v(r ) , where the potential @xmath5 is zero for @xmath6 with some @xmath7 .
we now discuss the energy eigenstates satisfying : @xmath8 .
one can expand the eigenfunction in terms spherical harmonics : ( r)=_lm(r)y_lm(n ) .
with : @xmath9 and @xmath10 is the radial wave - function satisfying the radial schrdinger equation : [ eq : radial ] ( d^2dr^2 + 2rddr - l(l+1)r^2 + k^2 - 2mv(r))_lm(r)=0 .
where @xmath11 being the energy eigenvalue of the state .
it is well - known that , there exist only one solution to the radial schrdinger equation that is bounded near the origin .
this solution will be denoted as : @xmath12 . to fix the normalization
, we impose the condition : [ eq : normalize_u ] _
r0 r^-lu_l(r;k)=1 , and the solution to the radial schrdinger equation then has the form : _
lm(r)=b_lmu_l(r;k ) , with some constant @xmath13 to be fixed by other conditions ( normalization , boundary conditions , etc . ) . in the region @xmath6 where the interaction vanishes
, the solution @xmath12 are expanded in terms of spherical bessel functions : [ eq : ul_expand ] u_l(r;k ) = _ l(k)j_l(kr ) + _ l(k)n_l(kr ) .
the coefficients @xmath14 and @xmath15 have simple relation with the scattering phase shift : e^2i_l(k)=_l(k)+i_l(k ) _
l(k)-i_l(k ) , _
l(k)=_l(k)_l(k ) . in the low - energy limit : @xmath16 , one normally defines : ^0_l=_k0 k^l_l(k ) , ^0_l=_k0 k^-l-1_l(k ) , and the threshold parameters : which is usually the case . ]
in particular , @xmath17 for @xmath18 is referred to as the @xmath19-wave scattering length .
other @xmath20 s for @xmath21 are sometimes also called scattering lengths in the corresponding channel , although they do not have the dimension of a length . ) , it is easy to verify that the spectral parameters @xmath20 has the length dimension of @xmath22 . ] the threshold parameters @xmath20 are important because they characterize the behaviors in low - energy scattering processes .
for example , we have : _
l(k ) a_l k^2l+1+o(k^2l+3 ) , ( ) .
we now enclose the system we discussed in the previous section in a large cubic box and impose the periodic boundary condition in all three spatial directions .
the potential itself is also modified to @xmath23 by periodically extending over the whole space : @xmath24 . for later convenience ,
we define the the so - called outer region " as : = \{r : |r+nl|>a , n^3}. this is the region where the potential vanishes identically . we assume the size of the box is @xmath1 which is much larger than any of the physical scale in the system .
in particular , we need to have @xmath25 so that the outer region admits free spherical wave solutions ( asymptotic states ) .
we now would like to study the change in the energy eigenvalues , the corresponding wave - functions and their possible connections with the scattering phase shifts in the infinite volume .
our discussion here will focus on the case of a cubic box whose relevant symmetry group being the cubic group @xmath26 .
generalization to an arbitrary rectangular box can be performed easily by changing the symmetry group to the corresponding ones ( @xmath27 or @xmath28 , etc . ) .
since the boundary condition breaks rotational symmetry explicitly , we anticipate that energy eigenstates of the system will not have a definite angular momentum in general . to be specific
, the original eigenstate in the @xmath19-wave will acquire mixtures from higher angular momentum modes ( mainly @xmath29 for a cubic box ) .
however , since the original radial wave - function @xmath12 and the spherical harmonics forms a complete set in the functional space , we may still expand the true eigenfunction in the box in terms of them : [ eq : eigenfunction_exact ] ( r;k)=_lm b_lmu_l(r;k ) y_lm(n ) . where the coefficients are to be determined by boundary conditions and normalization . in the outer region @xmath30 ,
the solution are those singular , periodic solutions for the helmholtz equation .
thus we may write : ( r;k)|_r=_lmv_lmg_lm(r;k^2 ) . in the meantime , the outer solution can also be expanded in terms of spherical harmonics and the spherical bessel functions @xmath31 and @xmath32 : [ eq : expand_ylm ] g_lm(r;k^2)=(-)^lk^l+14 , the explicit expression for @xmath33 is given in ref .
@xcite which we quote here : [ eq : calm - zeta ] m_lm;js(k^2 ) & = & _ lm(-)^si^j - lz_lm(1,q^2 ) ^3/2q^l+1 + & & ( ccc l & l & j + 0 & 0 & 0 ) ( ccc l & l & j + m & m & -s ) .
here we have used the wigner s @xmath34-symbols and @xmath35 . the zeta function @xmath36 is defined as : [ eq : zeta_def ] z_lm(s , q^2)= _
n y_lm(n ) ( n^2-q^2)^s . from the analytically continued formula ,
it is obvious from the symmetry of @xmath26 that , for @xmath37 , the only non - vanishing zeta functions at @xmath38 are : @xmath39 , and @xmath40 .
this is in accordance with the fact that @xmath19-wave and @xmath41-wave mixes with each other in a cubic box . in what follows
, we will focus on the @xmath19-wave eigenfunction . in the remaining part of this paper
, we will be only concerned with the energy eigen - functions in the @xmath3 sector , which is the analogue of @xmath19-wave in a cubic box .
a good approximation for the @xmath19-wave dominated eigenfunction can be written as a superposition of @xmath18 and @xmath29 spherical harmonics with the @xmath19-wave component much larger than that of @xmath41-wave . to explicitly construct this type of wave - functions
, we notice that the eigen - function in @xmath3 sector has to be invariant under cubic symmetries .
it is easy to verify that , there are only two homogeneous harmonic polynomials which are invariant under cubic symmetry up to @xmath42
. they can be conveniently expressed as : y_00=1 , y_40 + 14(y_4,4+y_4,-4 ) = 154(x^4+y^4+z^4 - 35 r^4 ) .
so , we may write the eigen - function in @xmath3 sector as : [ eq : psi_a1 ] ^(a^+_1)(r;k ) = b_00u_0(r;k)y_00 + b_40u_4(r;k)(y_40 + 14(y_4,4+y_4,-4))+ , with @xmath43 in the large volume limit .
in other words , to ensure cubic symmetry , the general coefficients @xmath13 at @xmath29 with different @xmath44 values must have definite ratios . in the outer region , using relation ( [ eq : ul_expand ] ) , we have : [ eq : psi_in_out ] ^(a^+_1)(r;k)|_r & = & b_00[_0j_0(kr)+_0n_0(kr)]y_00(_r ) + & + & b_40[_4j_4(kr)+_4n_4(kr ) ] ( y_40 + 14(y_4,4+y_4,-4 ) ) + .
on the other hand , we know that , in the outer region @xmath30 , the eigen - function can also be expanded into singular periodic solutions of helmholtz equation . since @xmath45 with @xmath46 being rotationally invariant , we see that in order to keep the eigen - function invariant under cubic symmetry , we must have the combination : @xmath47 in the expansion .
thus we may write : and an extra factor of @xmath48 for the coefficient of @xmath49 . ]
[ eq : psi_into_glm ] ^(a^+_1)(r;k)|_r= ( 4k ) v_00 .
the fact that such a combination respects cubic symmetry can also be checked explicitly . using the expressions ( [ eq : expand_ylm ] ) and ( [ eq : calm - zeta ] ) , we may write the expansion for @xmath50 as : [ eq : g00 ] g_00 = k4 , where we have introduced : @xmath51 and @xmath52 for later convenience ( see ref . @xcite for the notation ) .
similarly , for the higher angular momentum functions , we have : [ eq : g_expand ] g_40 & = & k^54 , + g_4,4+g_4,-4 & = & k^54 , in the above expansions , we have also utilized the following properties of the matrix elements @xmath53 : m_lm;lm=m_lm;lm = m_l ,- m;l,-m. note that in the expansion of @xmath49 and @xmath54 in eq .
( [ eq : g_expand ] ) , there are terms with @xmath55 , @xmath56 spherical harmonics . however ,
when we construct the combination @xmath57 , the terms with @xmath55 cancel out explicitly since : @xmath58 which can be checked by looking into table e.1 in ref .
therefore we finally have : [ eq : g40 ] g_40 + 14(g_44+g_4,-4)&= & k^54 where @xmath59 . at this stage , it is worthwhile to point out that , @xmath60 , @xmath61 and @xmath62 that we introduced here are exactly those reduced matrix elements of @xmath63 in the @xmath3 sector .
please refer to ref .
@xcite for further detailed explanations ( especially table e.1 and table e.2 in the reference ) .
collecting relevant information from the expansions obtained thus far , i.e. eq . ( [ eq : psi_into_glm ] ) , eq .
( [ eq : g00 ] ) and eq .
( [ eq : g40 ] ) , we have : [ eq : psi_out_out ] & & ^(a^+_1)(r;k)|_r = v_00 + , we should now match the two solutions given by eq .
( [ eq : psi_in_out ] ) and eq .
( [ eq : psi_out_out ] ) in the outer region @xmath30 .
this yields the following set of linear equations : v_00 & = & b_00_0 , v_00(m_00+v_40m_04)=b_00_0 , + v_00v_40&=&b_40_4 , v_00(m_04+v_40m_44)=b_40_4 .
these four equations can be viewed as a set of homogeneous linear equations for the four coefficients : @xmath64 , @xmath65 , @xmath66 and @xmath67 .
demanding a non - trivial solution to exist requires the corresponding determinant of the @xmath68 matrix to vanish .
another simple way to proceed is to divide the second equation by the first and similarly divide the fourth one by the third .
this will eliminate all coefficients except for @xmath69 .
we then arrive at : ^(0)=m_00+v_40m_04 , ^(4)=m_44+m_04/v_40 . eliminating @xmath69 from the above two equations then yields : [ eq : match2 ] ( ^(0)-m_00 ) ( ^(4)-m_44)= m_04m_04 .
this is exactly the equation obtained by general lscher s method when we only consider the mixing between @xmath18 and @xmath29 waves @xcite .
therefore , using more explicit construction , not only have we recovered lscher s formula , we also obtained an explicit approximate expression for the energy eigen - function in the @xmath3 channel which is given by eq .
( [ eq : psi_a1 ] ) in general and given by eq .
( [ eq : psi_into_glm ] ) in the outer region .
now we would like to derive a formula for the spectral weight function which can be measured in a monte carlo simulation . instead of working with general states
, we will focus on the single- and two - particle states .
these states naturally arise in the lattice study of hadron - hadron scattering and hadron spectrum . in such simulations ,
one constructs an operator ( also known as the interpolating field operator ) , or operators if more than one is needed , within a specific symmetry sector of the theory .
the correlation matrix among these operators are then computed by ensemble averaging over different gauge field configurations that is generated in a monte carlo simulation . for this purpose , we pass over to the second - quantized version of our quantum mechanical scattering model .
in this model , two distinguishable particles scatter via a potential @xmath5 where @xmath70 being the distance between them .
the center - of - mass coordinate of the two - particle system is separated out and the mass parameter @xmath44 in the hamiltonian ( [ eq : hamiltonian ] ) refers to the reduced mass of the two - particle system .
for each type of particle , a local scalar field operator @xmath71 , with @xmath72 designating different types of particles , is introduced together with its momentum space counterpart : [ eq : fourier_def ] _
i(x , t)=1_p_i(p , t)e^ ipx , _ i(p ,
t)=1d^3x_i(x , t)e^-ipx they satisfy the usual equal - time commutation relations : @xmath73=\delta_{{{\mathbf p}}{{\mathbf k}}}\delta_{ij}$ ] . using free states made up of two particles , one from each type
, one can form a state : |=o^(0)|0=1l^3/2_p(p ) ^_1(p,0)^_2(-p,0 ) |0 , with the interpolating operator @xmath74 defined by : o(t)=1_p ^*(p)_1(p , t)_2(-p , t ) .
requiring such a state to be normalized as : @xmath75 yields the condition : _ p|(p)|^2=1 .
if such a state were a bound state of two particles , @xmath76 would be the corresponding momentum - space wavefunction normalized according to the above equation .
we can now define the corresponding correlation function : c(t)=0|o(t)o^(0)|0 = _ e |e|o^(0)|0|^2 e^-et , where @xmath77 and @xmath78 represents the eigenvalue and eigenstate of the full hamiltonian , respectively . by fitting the time - dependence of the correlation function obtained from monte carlo simulations , the exact eigenvalue @xmath77 , and the corresponding spectral weight function @xmath79 , which is the coefficient in front of the exponential ,
is obtained .
if we denote the overlap of two wavefunctions : [ eq : overlap ] o(e)=e|o^(0)|0 = d^3r_1d^3r_2e|r_1,r_2 r_1,r_2|o^(0 ) |0 , the spectral weight function is simply given by : [ eq : w_def ] w(e)=|e|o^(0)|0|^2=|o(e)|^2 . at this point
, it is worthwhile to point out that the spectral weight function @xmath79 defined above depends explicitly on the normalization of @xmath80 . due to translational symmetry ,
the exact wave - function @xmath81 will only depend on the relative coordinate @xmath82 .
it is independent of the center - of - mass coordinate @xmath83 .
this means that , if the eigenstate @xmath78 is normalized according to @xmath84 as it should , the wave - function @xmath85 should be normalized according to : [ eq : a1_normalize ]
_ t_3 d^3r|r|e|^2= _
t_3 d^3r|^(a^+_1)(r;k)|^2=1l^3 . therefore , in order to compute the volume dependence of the spectral weight function , we first have to fix the normalization of @xmath86 according to this convention . as discussed in the previous subsection , the wavefunction in the @xmath3 sector in eq .
( [ eq : psi_a1 ] ) must be normalized properly on the torus @xmath87 according to eq .
( [ eq : a1_normalize ] ) .
the integral of the eigen - function on the torus runs over two regions : the inner region where the explicit form of the wavefunction is not known and the outer region @xmath30 where an approximate form of the function is given by eq .
( [ eq : psi_into_glm ] ) .
although we do not know the exact form of the eigen - function in the inner region , we do know that the eigenfunction is bounded in this region .
since it is assumed that the interaction region is of size @xmath88 with @xmath89 , therefore the integral in the normalization condition ( [ eq : a1_normalize ] ) is dominated by the integral of the function in the outer region @xmath30 . therefore , we may modify the normalization condition to : _
d^3r|^(a^+_1)(r;k)|^2 . since in the large volume limit
, the eigen - function is dominated by the @xmath19-wave contribution , we may use the first term in eq .
( [ eq : psi_into_glm ] ) and write : [ eq : norm_approx ] ( 4k)^2|v_00|^2(_t_3 d^3r|g_00(r;k)|^2 -_b d^3r|g_00(r;k)|^2 ) , where the second integral is over the interaction ball region : @xmath90 .
we now use the definition for @xmath50 : [ eq : g00_explicit ] g_00(r;k)=1l^3_pe^ip^2-k^2 , where the summation of @xmath91 is for all three - dimensional integers : @xmath92 . substituting this expression into the first term and eq .
( [ eq : psi_out_out ] ) into the second integral in eq .
( [ eq : norm_approx ] ) we get : _
p1(p^2-k^2)^2 - k^216 ^ 2 ^a_0 r^2dr ( n_0(kr)+m_00j_0(kr))^2 .
the integral in the second term maybe evaluated directly within @xmath93 .
we thus obtain : [ eq : v00_two_terms ] 1|v_00|^2l^3_p1(p^2-k^2)^2 - a2k^2 ^ 2 , where we have utilized the definition : @xmath94 . in the large volume limit , the first term on the r.h.s . of the above equation
is much larger than the second ( see appendix [ app : f ] for the explanation of this assertion ) .
if we drop the second term , we then arrive at : [ eq : v00_f ] ( 4k)^2|v_00|^2l^3 4(1l^3_p1(p^2-k^2)^2)^-1 , where we have defined the function : f(k^2)=1l^3_pf(p^2)^2-k^2 , where we have introduced a cutoff function @xmath95 to regulate possible ultra - violet divergences .
the property of this function in the large volume limit is addressed in appendix [ app : f ] .
the relevant formula for us is given by eq .
( [ eq : fprime ] ) .
we now evaluate the spectral weight using eq .
( [ eq : w_def ] ) with the exact energy eigen - function given approximately by : @xmath96 . the overlap of the two wave - function is approximately given by : o=(4k)v^*_001 _ p(p)^2-k^2 . using the expression ( [ eq : v00_f ] ) and
the expression in eq .
( [ eq : fprime ] ) , we finally obtain @xmath97 as : [ eq : w_formula ] w(e)=8k |_l(k^2)|^2 _ 0(k ) + 2k^2p^2 ^ 2_0(k ) = 8k |_l(k^2)|^2 _ 0(k ) + 2ee^2_0(k ) , where the function @xmath98 is defined as : _
l(k^2)=1l^3_p(p ) ^2-k^2 . in the large volume limit ,
following similar derivation as in our discussion of function @xmath4 , this function goes over to : _
( k^2)=p+ k(k^2)4_0(k ) .
thus the function @xmath98 has little volume dependence in the large volume limit .
therefore , the explicit volume dependence of the spectral weight function @xmath79 comes mainly from the denominator in eq .
( [ eq : w_formula ] ) . normally , if @xmath99 is not changing rapidly , the second term in the denominator of eq .
( [ eq : w_formula ] ) , which is proportional to @xmath100 , dominates the result and one finds that the spectral weight is proportional to @xmath0 .
this is the typical two - particle spectral weight function .
however , if there exists a rather narrow resonance at energy @xmath101 , then close to this resonance energy , one has approximately : ( e ) , where @xmath102 is the physical width of the resonance . in this case , we obtain : [ eq : resonance ] w(e ) .
if @xmath103 , then the quantity in the denominator is dominated by the first term and the spectral weight shows a typical single - particle behavior .
this means that an extremely narrow resonance behaves like a stable particle .
if on the other hand @xmath104 , which is always true for an extremely large volume ( assuming the width of the resonance remains finite ) , the denominator is dominated by the second term and the spectral weight itself is roughly proportional to @xmath0 which is typical for a two - particle scattering state .
we therefore arrive at the conclusion that the volume dependence of the spectral weight near a resonance is controlled by the ratio @xmath105 .
our results on the volume dependence of the spectral weight is obtained within a quantum mechanical model . in this subsection
, we would like to generalize these results to massive quantum field theory , following the line of arguments in lscher s formalism @xcite . using an effective schrdinger equation ( derived from the bethe - salpeper equation ) @xcite
, lscher has argued that , if the size of the box is large enough such that all quantum field theory effects are suppressed exponentially , the results obtained within the quantum - mechanical model can be carried over to the case of massive quantum field theory literally @xcite . here
, we will assume that the same conditions are satisfied and thus our results obtained within the quantum - mechanical model are expected to be valid for massive quantum field theory .
the relation established in eq .
( [ eq : resonance ] ) opens up a possibility for extracting the width of a resonance if the spectral weight can be measured in the simulation . assuming that there exists a single resonance in the energy region that we are interested in , and the contribution from this single resonance dominates the scattering , we simply rewrite eq .
( [ eq : resonance ] ) as : ( e_-e + e_e ) .
therefore , by fitting the function @xmath106 for different @xmath77 and @xmath1 ( hence different @xmath107 as well ) , it is possible to extract the width parameter @xmath102 together with the resonance position @xmath108 of the resonance .
note that in previous lattice calculations , focus has been mainly put on the energy levels , i.e. the values of @xmath77 , only .
no attention is paid to the associated spectral weight function @xmath2 which in fact can be obtained from the fitting procedure of the corresponding correlation functions _ with almost no extra costs_. the study in this paper indicates that , the spectral weight function at various volumes also contains valuable information about the scattering and might also be utilized in some way .
in fact , it can be used as an cross - check for the scattering phase obtained from the energy levels .
of course , this is only a possibility at this stage .
the feasibility of this method has to be check in realistic simulations .
in this paper , we have studied the volume dependence of the spectral weight function which is accessible in monte carlo lattice simulations . motivated by our previous study in the lee model
, it is expected that the spectral weight function shows little volume dependence for a stable or narrow resonance while for a broad resonance , it exhibits a typical @xmath0 dependence , the same as a two - particle scattering state . to verify this scenario , lscher s formalism is adopted .
it is first shown in a quantum mechanical model and then generalized to any massive quantum field theory , assuming that the polarization effects are exponentially suppressed following lscher s arguments .
in particular , we expect this scenario to be true also for qcd which governs the scattering of hadrons and therefore our result is relevant for lattice qcd simulations .
our final result for the spectral weight is summarized in eq .
( [ eq : w_formula ] ) which exhibits either single- or two - particle volume dependence depending on the value of @xmath109 where @xmath102 is the physical width of the resonance and @xmath107 is the typical level spacing near the resonance in the finite volume .
possibilities of using the formula to extract the width of a resonance is also discussed .
the author c. liu would like to thank prof .
liu from university of kentucky , dr . j.p.ma from itp , academia sinica , dr .
y. chen from ihep , academia sinica , prof .
b . liu from nankai university , prof . j .-
b . zhang from zhejiang university , prof .
h. q. zheng , prof .
s. h. zhu and prof .
s. l. zhu from peking university for valuable discussions .
to study the normalization of the wavefunction @xmath86 in the large volume limit , we define the function : f(k^2)=1l^3_pf(p^2)^2-k^2 , where we have introduced a cutoff function @xmath95 .
the relevant function appearing in the normalization condition ( [ eq : v00_f ] ) is given by the derivative of @xmath4 with respect to @xmath110 : f(k^2)=1l^3_pf(p^2)(p^2-k^2)^2 .
we now follow the argument in ref .
@xcite to estimate the value of @xmath4 for arbitrary value of @xmath110 in the large @xmath1 limit .
we separate the summation into two parts with : @xmath111 and @xmath112 .
the first part goes smoothly to the principle - valued integral @xmath113 while the second summation may be written as : _
p,|p^2-k^2|<1p^2-k^2 & = & 1l^3^_n=-1 p^2_+np^2-k^2 + & = & -l^3p^2 , where @xmath114 is the value of @xmath115 that is closest to @xmath110 ; @xmath116 is the typical level spacing between neighboring @xmath115 values which can be estimated by : 2p^2 = 1 l^3p^2=(2)^2 therefore we obtain : f(k^2)=(k^2)-k4 .
since it is easy to verify that : f(k^2)=z_00(1;q^2)2 ^ 3/2 l _ 0(k ) , where we have utilized the approximate relation ( lscher s formula ) : _
0(k)=z_00(1;q^2)^3/2 q. we therefore seem to have : @xmath117 in which case we recover the dewitt s formula : _ 0(k)=-(k^2-p^2_p^2 ) .
if one evaluate @xmath113 explicitly , one gets : ( k^2)=p = 4 + 2k|-k+k| , with a sharp momentum cutoff @xmath118 .
this expression indeed goes to zero if we drop the constant term and taking @xmath119 .
consequently we have for the function @xmath120 : [ eq : fprime ] f(k^2 ) & = & -18k+ k4p^2 ^2 + & = & 18k_0(k ) + k4p^2 ^ 2_0(k ) , where in the second line we have used dewitt s formula . since @xmath121 , we find that @xmath122 in the large volume limit .
this justifies the assertion made after eq .
( [ eq : v00_two_terms ] ) in the main text . | it has been suggested that the volume dependence of the spectral weight could be utilized to distinguish single and multi - particle states in monte carlo simulations . in a recent study using a solvable model , the lee model
, we found that this criteria is applicable only for stable particles and narrow resonances , not for the broad resonances . in this paper , the same question is addressed within the finite size formalism outlined by lscher . using a quantum mechanical scattering model , the conclusion that was found in previous lee model study
is recovered .
then , following similar arguments as in lscher s , it is argued that the result is valid for a general massive quantum field theory under the same conditions as the lscher s formulae . using the spectral weight function ,
a possibility of extracting resonance parameters is also pointed out . , ,
, spectral weight , finite - size technique , lattice qcd . |
recent observations ( adelberger 2003 , shapley 2003 ( s03 ) , steidel 2003 , adelberger 2005 ) based on a combination of quasar absorption - line and faint - galaxy techniques applied to the same cosmic volumes have revealed that uv - selected , star forming galaxies at redshift @xmath4 ( in brief , lyman break galaxies , lbgs ) show spectral absorption lines due to heavy elements .
these absorption lines , which can be very strong ( @xmath5 @xmath6 ) , are often called interstellar " because they resemble the absorption lines produced by interstellar material in local galaxies .
nevertheless , these lines appear ( without exception ) to be _ blueshifted _ with respect to the galaxy systemic velocity deduced from nebular lines .
the typical velocity difference is @xmath7 km s@xmath8 , but a non - negligible fraction of lbgs show differences in excess of 300 km s@xmath8 ( pettini 2002 ) .
these absorbing systems are often interpreted as arising in cool and/or neutral structures ( maybe a shell ) part of an outflow powered by supernovae associated with the detected galaxy star formation activity .
an additional element in support of this hypothesis comes from the detection of _ redshifted _
ly@xmath0 photons backscattered from gas behind the stars .
quite interestingly , this emission could trace the receding part of the outflowing shell .
finally , strong ly@xmath0 absorption is produced by the intergalactic gas within @xmath9 comoving mpc of most lbgs , but for about 1/3 the absorption is weak or absent .
the transparency of the medium around some lbgs could arise from the gas collisional ionization / heating by the wind ( croft 2002 , adelberger 2003 , kollmeier 2003 , bruscoli 2003 , bertone 2005 , kollmeier 2006 ) , from a galaxy proximity ( photoionization ) effect , as pointed out by maselli ( 2004 ) , or simply reflect cosmic variance and artifacts of the data reduction ( desjacques 2004 , desjacques , haehnelt & nusser 2005 ) . in general
, one would expect to observe also infalling gas in the surroundings of lbgs arising either from gas accretion along the cosmic web filaments or because at least part of the outflowing gas could rain back onto the galaxy if its velocity does not exceed the escape speed of the system ( a process known as galactic fountain ) .
in fact , the infall ram pressure might be of primary importance in confining and reverting the wind expansion ( fujita 2004 ) .
this situation is clearly appreciated from the inspection of the results of recent , high resolution simulation of lbg outflows ( ferrara , scannapieco & bergeron ; see fig . 2 of that paper ) .
however , observations so far have not found any experimental evidence of such physically - based expectation .
the central aim of this paper is to understand why outflows are routinely observed in lbgs whereas gas infall seem to escape detection .
this puzzle is possibly related to the following scenario . during the first phase of a starburst the galactic wind produces a cavity filled with hot gas , enclosed by a cold and dense shell . afterwards the shell fragments and the velocity of the cold debris starts decreasing while the hot gas escapes from the galaxy . whether or not they survive for some time , the uv light from the galaxy should become progressively less absorbed by the debris since their covering factor decreases with time as they travel to larger distances from the galaxy .
observations may show a time evolution of the equivalent width ( ew ) of metal absorption lines and dust reddening produced by the decrease of the screening effect of the cold debris on the uv light from the galaxy . if the debris are not disrupted or ejected from the galaxy we should also observe the gas return ( infall ) process .
the relevant question to pose then seems : can we observe galaxies during all these different evolutionary stages ?
a large statistical galactic sample ( more than 800 galaxies at intermediate redshifts ) that can be used to test this model has been compiled by s03 . from the analysis of this data set a very interesting number of relationships emerge ( some of which already noticed by s03 and others obtained by correlating their data in a different way ) , which are listed in the following : * the wind velocity , @xmath10 , is proportional to the star formation rate ( sfr ) : @xmath11 * the ly@xmath0 emission line ew is inversely proportional to the wind velocity : @xmath12 * the ew of metal absorption lines is proportional to the wind velocity : @xmath13 * the reddening is proportional to the wind velocity : @xmath14 should we interpret these properties as produced by intrinsic variations of the physical properties ( _ eg .
_ , sfr , mass , size ) of lbgs ? or do they rather reflect the different ages of a single ( broadly speaking ) galactic population with uniform properties ?
the relationships ( 3 ) and ( 4 ) naturally descend from the properties of a time evolving uniform population : as the cold debris move outward they are slowed down by gravity / energy losses , their covering factor decreases , and the ew and reddening also decrease .
property ( 2 ) can easily be understood in the same framework .
contrary to the metal absorption lines , the ew of the ly@xmath0 emission line increases with decreasing wind velocity ( _ ie .
_ , increasing time from the starburst ) since the scattering by the outflowing neutral gas is reduced and a larger fraction of the blue wing of the ly@xmath0 line is transmitted .
naively , relationship ( 1 ) can be interpreted as due to a variation of the energy input of the wind for galaxies with different star formation rate .
however , a closer look at that relationship ( figure [ fig : v_sfr ] ) shows that the dependence of @xmath10 on the sfr is not as the one expected under this assumption .
assuming , for simplicity , a constant density profile for the gas in the galaxy core , the velocity is a very weak function of the sfr : @xmath15 with @xmath16 ( shown by the solid line ) . even assuming a more general gas density profile of the galaxy core , @xmath17 , would produce @xmath18 that is too flat to fit the observations for any realistic choice of @xmath19 . a very steep slope @xmath20 ( shown by the dashed line ) fits the observations , but would produce unrealistically large outflow velocities incompatible with the data .
moreover , we will show later that the cold shell becomes unstable when the slope of the density profile becomes steeper than @xmath21 : at that point fragments decouple from the hot wind .
thus , the steepest physical dependence of @xmath10 on the sfr is @xmath22 when @xmath23 , while the best fit to the data requires @xmath24 . guided by the aforementioned arguments , in this paper we propose that the different observed
wind velocities correspond to different starburst evolutionary times , and are not directly related to variations of their mean sfr or other internal properties .
the proportionality of the wind velocity and sfr is incidental .
it reflects the fact that both quantities decrease as the time elapsed from the starburst increases .
we have organized the paper as follows . in [
sec : sca ] we describe our model analytically assuming a power law density profile for the gas . in
[ sec : res ] we assume a more realistic density profile of the gas and we compare the numerical results to the observations . in [ sec : dis ] we discuss the implications of our model and provide an explanation for the lack of evidences for infalling gas in lbgs .
to understand some basic features of lbg outflows we first resort to scaling arguments .
we closely follow the method outlined in ferrara & tolstoy ( 2000 ) and madau , ferrara & rees ( 2001 ) to which we refer the interested reader for the details .
assume that ambient gas pressure , gravity , and cooling can be neglected ; moreover , we take a constant mechanical luminosity for the starburst , @xmath25 ( the case @xmath26 can be explored but there is no analytical solution ) .
then the outflowing ( thin ) shell radius evolution is described by : @xmath27 we further make the hypothesis that the halo gas density profile can be approximated ( at least piece - wise ) by a power - law profile : @xmath28 with @xmath29 and @xmath30 is the central density . as for a navarro , frenk & white ( 1997 , nfw ) dark matter halo baryons have a core radius @xmath31 , where @xmath32 is the concentration parameter , we have used such relation in eq . [ rho ] .
note that @xmath33 in the case of a central core , i.e. if @xmath34 . by substituting @xmath35 into eq .
[ rs1 ] we find @xmath36 with @xmath37 ; we use the standard definition of @xmath38 @xmath39^{-1/3 } \left ( \frac{1+z}{4 } \right ) ^{-1 } h^{-1 } { \rm kpc}.\ ] ] we also take @xmath40 . in the following ,
we adopt the value @xmath41 , the appropriate overdensity of the core in a nfw profile . if we further define ] @xmath42^{-{1
/3}}= 2.5~{\rm gyr}\left({0.22\over c}\right)^{\beta/3}\left({r_{vir}\over 24.7 ~{\rm kpc}}\right)^{5/3 } \left({1+z
\over 4}\right ) l_{w,40}^{-1/3 } , \label{tw}\ ] ] eq .
( [ rs2 ] ) can be casted in the very simple form : @xmath43 the velocity , time to @xmath44 , velocities to @xmath45 , masses to @xmath46 of the shell is then given by @xmath47 the dense shell is pushing on the rarefied interstellar medium ( ism ) gas . the ( external ) shell interface becomes rayleigh - taylor ( rt ) unstable if the shell is _
accelerating_. in this case @xmath48 is parallel to @xmath49 , and the vector goes from the heavy ( shell ) to the light ( ism ) fluid and the rt instability develops . from maclow & mccray ( 1988 ) simulations we note that the rt fingers are beyond the thin shell solution , i.e. the instability develops outside the bubble .
the nonlinear evolution leads to the formation of clumps which detach from the shell ; the hot gas wraps around them and escapes ; at the same time , the ram pressure on the clumps decreases so they lag behind the shock .
hence , the onset of the rt instability is fixed by the condition that the acceleration of the shell becomes positive .
the shell acceleration is @xmath50 the rt condition is satisfied when @xmath51 ; as @xmath52 for any reasonable density profile slope @xmath19 , this implies @xmath53 .
hence the instability starts to develop at the time @xmath54 when the shell enters in the @xmath55 part of the gas distribution .
before that it evolves in a @xmath56 density profile ( e.g. the baryonic core in a nfw profile or the inner part of a disk ) .
the linear growth rate of the instability on spatial scale @xmath57 is given by @xmath58^{1/2 } \simeq \left[{2\pi \vert g_s(\tau ) \vert \over \lambda } \right]^{1 \over 2 } , \label{rtgrow1}\ ] ] the second equality descending by the fact that the density contrast between the shell and the ism gas is @xmath59 .
note also that the instability grows first on the shortest scales and it diverges as @xmath60 . at these small scales , the rt is stabilized by viscosity and/or magnetic fields . as we are interested in the shell fragmentation , we estimate the growth time on a scale @xmath61 such that @xmath62 , which by eq .
( [ rs3 ] ) is equal to @xmath63 .
the maximum growth rate occurs when the acceleration is largest and scale is smallest , _
i.e. _ , at @xmath64 . then @xmath65^{1/2 } = \sqrt{6\pi q\vert ( 3q-1)\vert } ~\tau_i^{-1 } = { \sqrt{6\pi \vert ( 2-\beta)\vert } \over \vert 5-\beta \vert } ~\tau_i^{-1 } , \label{rtgrow2}\ ] ] where @xmath66 is the value appropriate for the stable region
. given this range we find that the longest time corresponds to @xmath34 , for which @xmath67 .
so what is @xmath54 ? as already mentioned , the baryonic core in a nfw halo has size @xmath68 , hence for the @xmath34 ( @xmath16 ) solution and for @xmath69 we find @xmath70 at the time @xmath71 , the instability becomes nonlinear and the shell fragments in dense clumps the move almost ballistically in the gravitational field of the dark matter halo from there on .
their precise mass / number is difficult to calculate , but to a first approximation @xmath72 , where @xmath73 is the shell thickness calculated at @xmath74 ; as now , we fix @xmath75 with @xmath76 .
the mass of the shell is approximately equal to the swept - up mass in the halo . to write the equation of motion for the fragments ( note that this is _ independent _ of the mass of fragments ) we first derive the nondimensional gravitational acceleration : @xmath77 where @xmath78^{-1/2 } \approx 0.32[(1+z)/4]^{-3/2}$ ]
gyr is proportional to the halo free fall time .
we have assumed a nfw profile with @xmath79 @xmath80 , and @xmath81 it follows that fragment velocity is governed by @xmath82 as @xmath83 , @xmath84 , then this approximation has been dropped . ]
@xmath85 where @xmath86 , @xmath87 and the constant @xmath88 has value @xmath89 note that @xmath90 ( @xmath34 ) before @xmath54 , but it increases to @xmath91 ( @xmath23 ) after that , due to the steeper profile initiating the instability .
we neglect that complication , by noting that the time - weighted mean is @xmath92 when computing @xmath93 and @xmath94 .
thus , we obtain @xmath95 and @xmath96 .
the resulting expression for the velocity is @xmath97 substituting the expression for @xmath98 above , and using @xmath99 we find @xmath100 the fragments are decelerated and they stop at @xmath101 at a time @xmath102 obtained by solving @xmath103;\ ] ] evaluating this expression at @xmath104 , yields @xmath105 . for @xmath106 ,
the fragments start to infall onto the galaxy unless , @xmath107 , in which case they are lost to the gravitational potential and ejected in the igm . using the value of @xmath44 from eq .
[ tw ] , with @xmath23 and @xmath69 and the other fiducial values there , one obtains @xmath108 myr and @xmath109 .
then we find @xmath110 @xmath111 we conclude that mechanical luminosities @xmath112 are required to eject the fragments ( @xmath113 ) from a typical lbg galaxy .
this conclusion is approximate as our formalism is valid in @xmath83 ( see eq .
[ vel ] ) and it actually overestimates the velocity ; nevertheless , it gives a first rough estimate of the escape of fragments .
are these mechanical luminosities consistent with the observed lbgs ?
the mechanical luminosity ( stellar winds and sne ) for a continuous star formation rate of @xmath114 yr@xmath8 , with @xmath115 , is @xmath116 ( this slightly top - heavy imf seems to be favored by the results obtained for high - redshift galaxies by malhotra & rhoads 2003 ) . assuming that a fraction @xmath117 ( veilleux2005 ) of this energy goes into kinetic energy , and the rest is radiated away , we find that @xmath118 .
thus an enormous sfr of about @xmath119 solar masses / yr would be needed ( even more , given the overestimate of the velocity from eq .
[ [ vel ] ] ) to eject the fragment .
the fact that cold , dense gas is very difficult to eject does not exclude that the hot gas can instead be expelled , as already noticed by mac low & ferrara ( 1999 ) .
suppose that after shell fragmentation the clumps move radially and keep constant size ( this is a reasonable assumption as long as they are embedded in and pressure - confined by the hot bubble interior gas flowing into the halo ) .
then their covering factor of the source ( which is independent of their mass ) decreases with time as @xmath120}^2 = { \left[1.8 \tau_i\over x_s(\tau)\right]}^2 = 10^{-4 } \tau^{-2 } , \label{eq : c}\ ] ] that is , the covering factor has already dropped to 1% when @xmath121 , which occurs at @xmath122 , or in dimensional units , after @xmath123 myr .
finally , the hot shocked gas continues its expansion , leaving the clumps behind .
where does the shock stop ?
it stops when it reaches pressure balance with the igm surrounding the halo , i.e. when its velocity has dropped to the igm effective sound speed , which is in the formula below is assumed to be @xmath124 km s@xmath8 .
this happens approximately when the internal pressure @xmath125 is equal to @xmath126 or , at @xmath127 , @xmath128^{1/3 } \approx \eta^{1/3 } \left({\dot m_{\star } \over { \rm m}_\odot { \rm yr}^{-1}}\right)^{1/3 } \left({t_{burst}\over 10 ^ 8{\rm yr}}\right)^{1/3 } { \rm mpc}.\ ] ] the previous argument neglects radiative losses , gravity and the possible infall ram - pressure ; hence , such estimate represents an upper limit to @xmath129 ( physical units ) .
the aim of this section is to show quantitatively that the s03 data can be interpreted according to our model . by fitting the model to the data we attempt to infer some underlying properties of the lbg population .
hereafter , we adopt a more realistic description of the gas density profile with respect to the idealized case adopted in [ sec : sca ] .
we also drop most of the simplifying approximations adopted previously .
we solve numerically eq . ( [ rs1 ] ) to calculate the evolution of the radius , velocity and acceleration of the shock front during the wind - driven phase .
when the slope of the density profile becomes steeper than @xmath23 the cold shell becomes unstable to rt instability and fragments .
the subsequent evolution of the cold fragments is calculated by solving eq .
( [ eq : fra ] ) and the time scale for shell fragmentation is calculated from eq .
( [ rtgrow1 ] ) .
the quantitative results found in this section are robust as they are insensitive to the poorly known properties of the gaseous disk .
we will show that our model is independent of the particular choice of the gas density profile , at least for `` beta models '' or exponential density profiles .
the results are also insensitive to the values of physical parameters such as the disk scale height , core density etc . , provided that they obey a given relationship .
a gas in hydrostatic equilibrium in a nfw dark halo with virial radius @xmath130 and concentration @xmath32 , has a density profile that is well fitted by a beta model ( makino , sasaki & suto 1998 ) @xmath131^{1.4}},\ ] ] where , for a gas at the virial temperature , the core radius is @xmath132 and @xmath133 is the central gas density .
it is likely that , due to gas cooling , the temperature of the gas is lower than the virial temperature . in this case
the core radius will be smaller and the central density higher .
independently of the assumed value of @xmath134 , if the gas mass is a cosmic fraction of the halo mass , @xmath135 , by imposing mass conservation and disk geometry we find @xmath136 we have assumed the following fiducial values for lbgs : total mass @xmath137 m@xmath138 , concentration @xmath69 ( @xmath139 kpc ) .
the data fitting is insensitive to the functional form of the density profile .
we have found that adopting an exponential density profile @xmath140 $ ] ( where now @xmath134 is the scale height ) the model fits the data equally well .
the time scale for shell fragmentation depends on the shape of the profile .
the evolution of the cold shell before fragmentation depends on the details of the profile but the subsequent evolution of the cold fragments is independent of the assumed profile .
this happens because the fragmentation time - scale varies in such a way to keep the fragment velocity the same .
we find that @xmath141 for the `` beta model '' , @xmath142 for the exponential profile with @xmath143 and @xmath144 for @xmath145 . during the acceleration and fragmentation phases , the velocity of the shell increases only by a factor @xmath146 of a few percent with respect to its initial value @xmath147 at zero acceleration .
this is consistent with eq .
( [ rtgrow1 ] ) that can be recast in terms of the variables of interest : @xmath148 and @xmath149 .
it follows that @xmath150 .
the velocity evolution of the cold fragments is degenerate for choices of the mechanical luminosity of the starburst @xmath25 that obey @xmath151 , where @xmath152 is a constant and @xmath153 is the fraction of mechanical luminosity converted into kinetic energy of the cold shell ( the rest being radiated away by the hot gas ) .
thus , further assuming eq .
( [ eq : loc ] ) , it follows that the velocity of the fragments is proportional to @xmath154 .
the cold shell driven by the starburst is stable for only a few myrs before it fragments and the clumps move ballistically .
thus the appropriate value of @xmath25 to use is that corresponding to this early phase , when the kinetic energy input produced by the starburst is dominated by winds from massive stars and is @xmath155 , for stellar metallicities @xmath156 ( leitherer 1999 ) .
the value of @xmath153 , the fraction of mechanical luminosity converted into kinetic energy of the supershell , is not well known from observations .
we adopt the fiducial value @xmath157% ( vielleux 2005 ) .
the lbg sample of s03 is grouped in four bins according to their ly@xmath0 equivalent width .
each bin contains about the same number of galaxies ( about 200 ) . in our interpretation ,
different bins describe a statistically steady - state time sequence and correspond to different times from the beginning of the starburst with typical time scale @xmath158 .
thus , if we neglect differential selection bias effects in the data , we expect that equally time - spaced bins contain an equal number of galaxies of age : @xmath159 for bins @xmath160 .
the observed star formation rate of the galaxies in each of the four bins is well fitted by an exponential function @xmath161 the numerical value of @xmath158 will be derived later by fitting the observations of the fragment velocities .
it also follows from eq .
( [ eq : sfr ] ) that the mean sfr integrated over the time scale of the burst is @xmath162 m@xmath138 yr@xmath8 and the total stellar mass produced during the burst is @xmath163 . in figure [ fig:3 ]
we show the fit of our fiducial model ( the `` beta model '' with @xmath164 kpc and @xmath165 @xmath166 ) to the data from s03 , shown by the symbols with errorbars .
the panels from top to bottom show the time evolution of the star formation rate ( fit by eq .
[ [ eq : sfr ] ] ) , the velocity of the hot gas ( dashed line ) and cold shell / fragments ( solid line ) , the equivalent width of low ionization metal absorption lines , and the distance from the galaxy of the hot ( dashed line ) and cold fragments ( solid line ) .
the symbols with errorbars in the third panel refer to the normalized ew of siii ( circles ) , oi ( open circles ) , cii ( squares ) and feii ( open squares ) calculated using eq .
( [ eq : c ] ) .
good fits to the observed velocity and ews of the metal lines are obtained for values of the starburst time scale @xmath167 and for @xmath168 ( appropriate for low stellar metallicity ) we find @xmath169 where the second equality is derived by using eq .
( [ eq : loc ] ) that express mass conservation for the gas . as an example , if we instead assume an efficiency @xmath170% we find @xmath171 pc , comparable to the disk scale height of the milky way .
we also find that , independently of the assumed details of the gas density profile , the fragments start to fall back into the galaxy at @xmath172 myr when they reach a distance from the disk of @xmath173 kpc .
although on larger scales the outflow approaches spherical geometry , the fragmentation point is located close enough to the galactic center that the effects of the disk density stratification are important .
we have already established that the cold clumps do not escape from lbgs .
therefore only two scenarios remain : either they disintegrate during the outflow or they eventually fall back onto the galaxy . in the second case
the reason why we do not see infalling clumps is simple : we do not observe this evolutionary phase because it happens too late when the uv emission from the galaxy has faded and the galaxy can not be found using the lyman break selection criteria ( selection techniques based on infrared colors might instead detect these galaxies long after the process has come to an end ) .
more quantitatively , after about @xmath174 myr the fragments will reach a maximum distance from the galaxy of about @xmath173 kpc and than start falling back . according to our model the typical duration of the starburst phase
is @xmath175 myr .
although ages since the most recent episode of star formation are found to be @xmath176 myr for 25% of the lbg sample ( shapley 2001 ) , the derived median value is 320 myr . a possible explananation for this tension could be that the stellar light is affected by the integrated star formation history of the galaxy ( which could be more complex than assumed in population synthesis models used for the interpretation of the data ) , while the kinematic properties of the wind are primarily affected by the very last burst episode . to clarify this point
additional study is required . during this time
scale the sfr declines as @xmath177}$ ] from 30 to 10 m@xmath138 yr@xmath8 . during the starburst
the fragments slow down from about @xmath178 km s@xmath8 to @xmath179 km s@xmath8 .
the uv burst is not long enough to allow the observation of galaxies showing the infall phase of the gas fragments . in our model
we have assumed that the fragments are stable and survive for at least @xmath180 myr , that is approximately the age of lbgs grouped in the last bin . in other words
the galaxies in the last bin show the low ionization metal absorption lines that arise from dense gas in the fragments .
let s now estimate the time scales for the destruction of the fragments .
the two main processes that may disintegrate the fragments are shear instability ( kelvin - helmholtz , hereafter kh ) and conductive evaporation .
the time scale for the growth of the kh instability , neglecting saturation effects , is approximately @xmath181 where @xmath182 is the fragment - hot gas density contrast , @xmath183 is the size of the fragments and @xmath184 their velocity relative to the flow of hot gas .
we have estimated the fragment size assuming that is comparable to the thickness of the cold shell .
when @xmath185 , the relevant case for our estimate , the instability growth rate saturates and eq .
( [ eq : kh ] ) ceases to be valid ( vietri , ferrara & miniati 1997 ) . in this case a good estimate of the minimum @xmath186 is derived by plugging the value @xmath187 km s@xmath8 in eq .
( [ eq : kh ] ) . at least for the largest fragments we find that @xmath188 .
magnetic fields and cooling can increase the stability of the clumps , increasing our estimate of @xmath186 . a dimensional estimate of the evaporation time scale of the cold fragments embedded in the hot wind as a result of conductive heating gives @xmath189 where @xmath190 is the mean column density of the fragments ( assumed to be equal to the column density of the cold shell ) .
we have assumed spitzer conductivity @xmath191 , neglecting saturation effects and magnetic field ( spitzer 1953 ) .
the survival time to kh instability and thermal evaporation are about @xmath179 myr , a factor of three larger than the life - time of the star burst .
the time scale for thermal evaporation is proportional to @xmath192 and for kh instability to @xmath183 .
we expect that the larger clumps may survive long enough to start falling back after @xmath172 myr but at this time the starburst has faded and the galaxy will not appear as a lbg with strong uv and ly - alpha emission
. it may be possible to test our model by selecting the faintest lbgs and searching for a subsample that has weak ly@xmath0 emission .
this subsample may show evidence of either low velocity outflows or gas infall .
adelberger , k. l. , steidel , c. c. , shapley , a. e. , pettini , m. 2003 , apj , 584 , 45 adelberger , k. l. , shapley , a. e. , steidel , c. c. , pettini , m. , erb , d. k. , reddy , n. a. 2005 , apj , 629 , 636 bertone , s , white , s. d. m. 2005 , astro - ph/0511028 bruscoli et al .
2003 , 2003 , mnras , 343 , 41 croft , r. a. c. , hernquist , l. , springel , v. , westover , m. , white , m. 2002 , apj , 580 , 634 desjacques , v. , nusser , a. , haehnelt , m. g. , stoehr , f. 2004 , mnras , 350 , 879 desjacques , v. , haehnelt , m. g. , nusser , a. 2005 , astro - ph/0511025 ferrara , a. , scannapieco , e. , bergeron , j. 2005 , apj , 634 , 37 ferrara , a. , tolstoy , e. 2000 , mnras , 313 , 291 fujita , a. , mac low , m .-
m . , ferrara , a. , meiksin , a. 2004 , 2004 , apj , 613 , 159 kollmeier , j. a. , weinberg , d. h. , dav , r. , katz , n. 2003 , apj , 594 , 75 kollmeier , j. a. , miralda - escud , j. , cen , r. , ostriker , j. p. 2006
, apj , 638 , 52 , c. , et al .
1999 , , 123 , 3 , m .-
m . , & mccray , r. 1988 , , 324 , 776 mac low , m .-
m . , ferrara , a. 1999 , apj , 513 , 142 madau , p. , ferrara , a. , rees , m. j. 2001 , apj , 555 , 92 , n. , sasaki , s. , & suto , y. 1998 , , 497 , 555 malhotra , s. , & rhoads , j. e. 2002 , apj , 565 , l71 maselli , a. , ferrara , a. , bruscoli , m. , marri , s. , schneider , r. 2004 , mnras , 350 , 21 navarro , j. f. , frenk , c. s. , white , s. m. n. 1997 , apj , 490 , 493 pettini , max , rix , s. a. , steidel , c. c. , adelberger , k. l. , hunt , m. p. , shapley , a. e. 2002 , apj , 569 , 742 shapley , a. e. , steidel , c. c. , adelberger , k. l. , dickinson , m. , giavalisco , m. & pettini , m. 2001 , apj , 568 , 95 shapley , a. e. , steidel , c. c. , pettini , m. , adelberger , k. l. 2003 , apj , 588 , 65 ( s03 ) , l. , & hrm , r. 1953 , physical review , 89 , 977 steidel , c. c. , adelberger , k. l. , shapley , a. e. , pettini , m. , dickinson , m. , giavalisco , m. 2003 , apj , 592 , 728 , s. , cecil , g. , & bland - hawthorn , j. 2005 , , 43 , 769 , m. , ferrara , a. , & miniati , f. 1997 , , 483 , 262 | a model for gas outflows is proposed which simultaneously explains the correlations between the ( i ) equivalent width of low ionization and ly@xmath0 lines , ( ii ) outflow velocity , and ( iii ) star formation rate observed in lyman break galaxies ( lbgs ) .
our interpretation implies that lbgs host short - lived ( @xmath1 myr ) starburst episodes observed at different evolutionary phases .
initially , the starburst powers a hot wind bound by a denser cold shell , which after @xmath2 myr becomes dynamically unstable and fragments ; afterwards the fragment evolution is approximately ballistic while the hot bubble continues to expand . as the fragments are gravitationally decelerated , their screening ability of the starlight decreases as the uv starburst luminosity progressively dims .
lbg observations sample all these evolutionary phases .
finally , the fragments fall back onto the galaxy after @xmath3 myr .
this phase can not be easily probed as it occurs when the starburst uv luminosity has already largely faded ; however , galaxies dimmer in the uv than lbgs should show infalling gas .
cosmology : theory galaxies : formation |
10 url # 1#1urlprefix[2][]#2 lhuillier c , sindzingre p and fouet j b 2001 _ can .
j. phys . _ * 79 * 15251535 ; arxiv : cond - mat/0009336 | spin-1/2 magnets with kagom geometry , being for years a generic object of theoretical investigations , have few real material realizations . recently , a dft - based microscopic model for two such materials , kapellasite cu@xmath0zn(oh)@xmath1cl@xmath2 and haydeeite cu@xmath0mg(oh)@xmath1cl@xmath2 , was presented @xmath3janson o , richter j and rosner h , arxiv:0806.1592@xmath4 $ ] . here ,
we focus on the intrinsic properties of real spin-1/2 kagom materials having influence on the magnetic ground state and the low - temperature excitations .
we find that the values of exchange integrals are strongly dependent on o
h distance inside the hydroxyl groups , present in most spin-1/2 kagom compounds up to date .
besides the original kagom model , considering only the nearest neighbour exchange , we emphasize the crucial role of the exchange along the diagonals of the kagom lattice .
two - dimensional ( 2d ) magnets with a kagom lattice arrangement of magnetic ions ( figure [ str ] ) are geometrically frustrated due to a triangular - like arrangement of nearest neighbours ( nn ) leading to an unusual highly degenerate classical ground state ( gs ) @xcite .
strong quantum fluctuations arising from spin-1/2 ions can lift the classical degeneracy and drive the system into a magnetically disordered quantum paramagnetic state , which might be applied for future quantum computational applications @xcite . nonetheless , in real materials couplings to further neighbours @xcite and between the kagom layers @xcite are always present and influence the gs and the thermodynamics . in a recent paper @xcite we have performed dft calculations for two new natural isostructural materials with spin-1/2 kagom lattice kapellasite cu@xmath0zn(oh)@xmath1cl@xmath2 @xcite and haydeeite cu@xmath0mg(oh)@xmath1cl@xmath2 @xcite .
the local density approximation ( lda ) yields a metallic solution ( left panel in figure [ dos ] ) in contrast to the experimentally observed insulating behaviour due to underestimation of strong on - site correlations for the cu@xmath5 sites . to account for this deficit
, we carried out an effective one - band tight - binding ( tb ) fit of the lda bands ( figure [ dos ] , right panel ) and subsequently mapped the resulting transfer integrals @xmath6 to an extended hubbard model and to a heisenberg model with @xmath7 .
this mapping provides an estimate for the antiferromagnetic ( afm ) part of the exchange . to obtain the values for the total exchange , consisting of afm and ferromagnetic contributions
, we have performed supercell calculations for various spin arrangements .
the correlations were treated in the lsda+@xmath8 approximation . we have found that besides the standard kagom model based on nn interactions ( @xmath9 ) , an additional coupling along diagonals of kagom hexagons ( @xmath10 ) is relevant ( figure [ str ] ) .
since the absolute values of @xmath9 and @xmath10 change only the overall temperature scale , their ratio @xmath11 is crucial for the gs and low - energy excitations .
kagom layers in the crystal structure of kapellasite ( haydeeite ) .
cuo@xmath12 plaquettes are shown in yellow , zno@xmath1 ( mgo@xmath1 ) octahedra are shown in blue .
the relevant exchange paths are highlighted.,width=283 ] left panel : total and atom - resolved densities of states of kapellasite cu@xmath0zn(oh)@xmath1cl@xmath2 ( left ) and the corresponding band structure ( right ) .
right panel : band structure of antibonding @xmath13 bands crossing the fermi level and the effective one - band tight - binding fit for different o
h distances ( experimental : 0.78 , lda bands dashed lines , tb fit green squares ; optimized : 1 , lda bands solid lines , tb fit red circles).,width=604 ] finite kagom lattice of @xmath14 sites .
the solid lines represent the nn bonds @xmath9 and the dashed lines the diagonal bonds @xmath10 .
the coloured circles indicate the spin orientations of the twelve - sublattice classical gs relevant for @xmath15 . along the chains formed by diagonal bonds
there is an antiparallel ( nel ) spin alignment ( e.g. the spins at the light and dark green circles ) . on each triangle formed by nn bonds
@xmath9 ( e.g. spins on sites 21 , 22 , 23 ) there is a @xmath16 spin arrangement .
in addition , two antiparallel ( nel ) spin - sublattices are perpendicular to one other group of two nel - like sublattices , i.e. ; ; .,width=604 ] gs spin - spin correlation @xmath17 in dependence on @xmath18 for the finite lattice of @xmath14 sites shown in figure [ 36 ] . the site indices @xmath19 correspond to those of figure [ 36 ] .
@xmath20 @xmath20 @xmath20 @xmath20,width=362 ] one focus of this paper is to emphasize the influence of side groups ( which are commonly neglected in model physics ) on different exchange paths .
a key issue of theoretical investigations of real materials is the construction of a relevant model , which describes experimental data and has predictive power . for cuprates , that are insulators with a @xmath21 cu@xmath5 configuration ,
the extension from the initial heisenberg idea to consider only spin degrees of freedom ( heisenberg model ) to a multiband hubbard model with a separate treatment of cu @xmath22 and o @xmath23 orbitals became computationally feasible only recently .
on the other hand , the famous goodenough - kanamori - anderson ( gka ) rules formulated quite early on an empirical basis provide a simple intuitive picture for the exchange interactions based on geometrical quantities . due to the simplicity of the gka rules
, they are violated in many cases . a natural way to search for a possible origin of these phenomena
is to improve the model by including ligand fields effect . in general ,
such models are very difficult to evaluate . even in case of simple systems like cugeo@xmath0 the models are quite complicated and contain many parameters @xcite . here
, we present the results of a dft - based modeling for cu@xmath0zn@xmath3mg@xmath4$](oh)@xmath1cl@xmath2 . in these compounds ,
the leading exchange interactions are strongly dependent on the position of h atoms the simplest possible side groups . for haydeeite , the experimentally defined h position yields an unusually short o
h distance ( @xmath24 ) of 0.78 ( for kapellasite the h position has not been reported ) .
therefore , in lda calculations we relaxed the h position which resulted in an optimized @xmath25 for both systems . in order to elucidate the influence of @xmath24 on the exchange integrals
, we have studied the range @xmath26 which covers the experimental errors ( that are large due to weak x - ray scattering by h atoms ) .
the details of our calculations are described in @xcite .
we find that @xmath24 has a strong influence on the nn exchange @xmath9 , while @xmath10 is almost unaffected .
obviously , this modifies their ratio @xmath27 and consequently the physical properties .
the underlying reason is the shift of oxygen states down in energy caused by the shortening of an o h bond . in consequence ,
cu o hybridization decreases .
therefore , for a fixed cu o geometry , a decreased @xmath24 reduces the afm contribution to the total exchange .
this is supported by a strongly reduced bandwidth ( which is related via @xmath28 to the afm exchange ) for a reduced @xmath24 ( figure [ dos ] , right panel ) and fits well the results of supercell calculations .
there , we find a quasi - linear @xmath29 dependence with a positive slope for both compounds .
@xmath9 is 2.5 times larger for @xmath30 than for @xmath31 in kapellasite ( 2.5 mev versus 1 mev ) , while it changes the sign in haydeeite ( @xmath32 mev versus @xmath33 mev ) .
the next step towards a physically relevant picture of real materials is to evaluate the influence of @xmath27 onto the gs and the low - lying excitations . here
, we use a @xmath9-@xmath10 spin-1/2 heisenberg model on the kagom lattice ( @xmath9-@xmath10 model ) .
the classical gs of the pure kagom system ( @xmath34 ) is known to be highly degenerate .
the additional diagonal bond @xmath10 reduces this degeneracy drastically and selects non - coplanar classical gs s with twelve magnetic sublattices ( see figure [ 36 ] ) among the huge number of classical gs s existing for @xmath34 .
for the quantum spin-1/2 model we have calculated the gs by exact diagonalization for the finite lattice shown in figure [ 36 ] . obviously , the quantum gs spin - spin correlations are drastically affected when including @xmath10 .
while for _ _
j__@xmath35 all spin correlations except the nn correlation function @xmath36 are close to zero we find a well pronounced short - range order for @xmath37 , that corresponds to the classical magnetic structure . for @xmath38
the strongest spin correlations are @xmath39 , @xmath40 , @xmath41 , and they belong to the chains formed by @xmath10 bonds .
the nn correlation function @xmath42 monotonously decreases with @xmath10 .
the correlation function @xmath43 belongs to two sites located on sublattices being perpendicular in the classical gs , and it is almost zero for @xmath38 .
this leads to the conclusion that even in the quantum model the gs might have a non - coplanar magnetic structure giving rise to enhanced chiral correlations .
to summarize , we have shown the relevance of the ligand field for the low - energy physics by example of two spin-1/2 cuprates with a kagom lattice geometry .
we found that there are two relevant exchange integrals in both materials .
the ratio @xmath18 is strongly dependent on the o
h distance and has a drastic impact on the physical properties : the quantum gs of a corresponding spin-1/2 heisenberg @xmath9-@xmath10 antiferromagnet on the kagom lattice exhibits strong magnetic correlations along @xmath10 bonds for @xmath44 , whereas the other correlations remain weak .
therefore , the low - energy excitations might be @xmath45 spinons causing an effectively one - dimensional low - temperature physics as has been discussed previously for other 2d models such as the crossed - chain model @xcite as well as for the anisotropic triangular lattice @xcite . as the physical properties are related to @xmath27 which is strongly dependent on the o
h distance , the h position should be reinvestigated experimentally with better accuracy .
generally , a precise determination of the position of side group atoms is crucial for an accurate model derivation .
we suggest that the dependence of exchange integrals on side group positions can be used to tune a magnetic ground state under changed conditions like external pressure . |
local magnetic moments resulting from partially filled localized orbitals can experience an indirect magnetic exchange coupling mediated via a system of conduction electrons .
the prime example is the ruderman - kittel - kasuya - yoshida ( rkky ) effective interaction @xcite which has an oscillatory dependence on the distance between the magnetic impurities .
@xcite in particular , the rkky mechanism in many cases explains the coupling of magnetic adatoms on non - magnetic metallic surfaces . due to the recent methodical advances in spin - resolved scanning tunneling microscopy techniques @xcite it is nowadays possible to map out the strength and the oscillatory distance dependence of the rkky interaction with atomic resolution .
@xcite moreover , the possibility to manipulate the positions of individual magnetic atoms offers the exciting perspective to build artificial adatom magnetic nanostructures with tailored magnetic properties .
@xcite in view of possible future applications for magnetic data storage , ferromagnetically ordered nanostructures deserve particular attention .
the simplest theoretical model of correlated electrons on a lattice that may sustain ferromagnetism induced by indirect rkky coupling is the periodic anderson model .
@xcite here , the magnetic impurities or the magnetic adatoms are described by sites with a non - zero on - site hubbard - type interaction of strength @xmath0 .
the impurity sites are coupled via a hybridization term @xmath1 to a system of non - interacting conduction electrons which hop between the nearest - neighboring sites of a lattice with a hopping amplitude @xmath2 .
local magnetic moments on the impurity sites are formed in the strong - coupling limit @xmath3 where @xmath4 is the conduction - electron density of states at the fermi energy .
a local polarization of the conduction electrons is induced via the antiferromagnetic local kondo exchange @xcite @xmath5 and results in an effective interaction @xmath6 between the impurities coupling to lattice sites @xmath7 and @xmath8 that is given in terms of the non - local static magnetic susceptibility @xmath9 .
consider the conceptually most simply case of a half - filled and particle - hole symmetric periodic anderson model on a bipartite @xmath10-dimensional lattice with @xmath11 conduction - electron sites . as the conduction - electron spin - spin correlations between nearest - neighbor sites
are antiferromagnetic in this case , i.e. @xmath12 , a _
rkky coupling is only possible if impurities are placed at sites that are separated by two ( or generally by an even number ) of nearest - neighbor hops .
therefore , a corresponding spin - depleted periodic anderson model with at most @xmath13 impurity spins is expected to sustain ferromagnetic order .
apart from the perspectives in modeling artificial ferromagnetic nanostructures , the depleted periodic anderson as well as the related kondo model are interesting to describe the electronic structure of one - dimensional organic ferromagnets .
@xcite the limit of strong kondo coupling @xmath14 is accessible to studies of ultracold alkali earth atoms trapped in optical lattices as has been discussed recently .
@xcite here , however , we would like to emphasize that the considered depleted models are in first place interesting from a more fundamental point of view .
due to the geometrically regular removal of impurities , an insulating state is avoided , and _
metallic _ ferromagnetism can be studied in a simple particle - hole symmetric model .
this is a similar motivation as for the antiferromagnetic case studied in ref . .
another motivation for the present study is that there are several and conceptually very different theoretical approaches that apply to the case of the depleted anderson or kondo lattice and provide independent explanations for the emergence of ferromagnetic order : ( i ) first of all , beyond the regime of weak hybridization @xmath15 , the indirect exchange is expected to compete @xcite with the kondo screening @xcite of the impurity magnetic moments .
( ii ) interestingly , in the limit of strongly coupled impurities ( strong @xmath15 or strong @xmath14 ) , the kondo effect has actually been seen to _ corroborate _ ferromagnetic ordering .
namely , an `` inverse '' indirect magnetic exchange ( iime ) provides a coupling of local moments formed _ between _ almost localized kondo singlets .
@xcite ( iii ) moreover , a variant @xcite of the lieb - mattis theorem @xcite may apply which predicts a ferro- or actually ferrimagnetic ground state in the entire @xmath15 or @xmath14 regime .
this is compatible with a crossover from the rkky to the iime limit .
( iv ) even the non - interacting ( @xmath16 ) limit is interesting : the lattice geometry inevitably leads a completely non - dispersive band right at the fermi energy . as a result
there is a highly degenerate fermi sea where the configuration with a fully polarized flat band is one of the possible many - body ground states .
this raises the question which characteristics of the interacting ferromagnet already derive from those of the fully polarized non - interacting ground state .
moreover , the stoner criterion @xcite applies to the weakly interacting system , i.e. the flat band makes the system extremely susceptible to magnetic ordering .
( v ) another question is whether the ferromagnetic order can be understood within the concept of `` flat - band ferromagnetism '' @xcite as well .
comparison of numerically exact data with standard hartree - fock theory may show whether the ground state is correlated at all and thus different from the fully polarized fermi sea that is predicted in the flat - band ferromagnetism scenario .
finally , another motivation of the present paper is a methodical one : the physics of the models studied here are very well accessible to reliable numerical techniques .
the density - matrix renormalization group ( dmrg ) @xcite provides essentially exact results for the one - dimensional case .
interestingly , dynamical mean - field theory ( dmft ) @xcite turns out as very reliable , too , even for @xmath17 , since the correlated sites are separated by 4 or more nearest - neighbor hops and thus the local approximation for the self - energy becomes rather accurate .
this allows us to address the two - dimensional lattice as well and motivates us to analyze the low - energy part of the single - particle excitation spectrum to study the `` fate '' of the flat band for a strongly correlated system .
results will be discussed mainly for the case of anderson but also for kondo impurities .
the paper is organized as follows : in the next section , we introduce the depleted anderson model .
its properties in the non - interacting limit and in particular the emergence of a flat band are analyzed in sec .
a study of the properties of the non - interacting state with a fully polarized flat band is given in sec .
the related concept of flat - band ferromagnetism is discussed in sec .
v along with calculations based on hartree - fock theory for the strong - coupling ( strong @xmath0 ) regime .
dmrg data for the one - dimensional model in the whole @xmath15 range are presented in sec .
this not only demonstrates the limitations of the hartree - fock approach but also allows us to analyze the crossover from the weak-@xmath15 to the strong-@xmath15 limit .
the role of the lieb - mattis theorem in this context is clarified in sec .
vii , while in sec .
viii the underlying physical pictures and mechanisms are discussed .
ix introduces the dynamical mean - field approach and therewith allows us to shift the perspective and to consider two - dimensional systems and particularly the low - energy one - particle excitation spectrum .
this is worked out in secs .
x and xi for the coherent part of the green s function and the density of states , respectively . in sec .
xii once more shifts and extends the perspective by considering the depleted kondo model in a regime where it is different from the related anderson model .
qualitatively , the low - energy excitation spectra of the two models turn out as very similar as discussed in sec .
xiii along with the question for the fate of the flat band in the strongly interacting limit .
the smaller local hilbert space of the depleted kondo model allows to carry out precise dmrg calculations for the charge susceptibility , discussed in sec .
xiv , which underpin the interpretations of the dynamical mean - field analysis .
a summary of the main results and the conclusions are given in sec .
we consider conduction electrons hopping between the nearest - neighbor sites of a @xmath10-dimensional bipartite lattice ( @xmath17 chain and @xmath18 square lattice ) consisting of @xmath11 sites .
the hopping amplitude fixes the energy scale , i.e. @xmath19 .
`` impurities '' , i.e. sites with a finite local hubbard interaction @xmath0 , are coupled via a hybridization of strength @xmath15 to the b sites of the lattice consisting of the two sublattices a and b , see fig . [
fig : geo ] for the two - dimensional case . in total
, there are @xmath13 impurities .
the hamiltonian reads as : @xmath20 here @xmath21 , @xmath22 and @xmath23 create an electron with spin @xmath24 at a site @xmath7 in sublattice a , at a site @xmath8 in sublattice b and at an impurity site @xmath25 , respectively . with @xmath26
we denote the impurity site attached to the b site @xmath8 .
furthermore , @xmath27 , @xmath28 and @xmath29 denote the corresponding occupation - number operators .
we will consider the model at half - filling where the average total particle number @xmath30 .
this is ensured by choosing @xmath31 for the chemical potential and @xmath32 for the on - site energy of the impurity sites .
impurities ( orange ) with on - site hubbard interaction @xmath0 are coupled via a hybridization of strength @xmath15 to the b sites ( red ) of a bipartite lattice with @xmath33 sites in total ( a sites : blue ) .
periodic boundary conditions are assumed .
the energy scale is fixed by setting @xmath19 for the nearest - neighbor hopping in the conduction - band system .
dashed lines indicate a unit cell .
the system is considered at half - filling , i.e. , the average total particle number is @xmath30 .
, scaledwidth=40.0% ]
assuming periodic boundary conditions and exploiting translational symmetries , the non - interacting part of the hamiltonian is block - diagonalized by switching to a wave - vector representation .
there are 3 sites in a unit cell ( a , b , impurity ) .
consequently , @xmath34 with a @xmath35 hopping matrix for each wave vector @xmath36 : @xmath37 here , @xmath38 is the bare ( @xmath39 ) dispersion of the conduction - electron band .
diagonalization of @xmath40 , @xmath41_{mm } \ ; , \ ] ] is achieved with the unitary transformation matrix @xmath42 this yields three bands ( @xmath43 ) , @xmath44 where @xmath45 .
each band consists of @xmath46 eigenenergies , i.e. there are @xmath46 allowed wave vectors in the first brillouin zone .
the @xmath47 band is dispersionless , i.e. `` flat '' .
the presence of the flat band originates from the ( i ) bipartite lattice structure , ( ii ) the restriction to inter - sublattice hopping , ( iii ) homogeneous on - site energies and ( iv ) from the fact that there are more a sites ( including impurity sites ) than b sites ( for the present discussion we have to regard the impurity sites as a sites as they are coupled to b sites only ) .
this can easily be seen from the real - space representation of the hopping matrix @xmath48 .
we have @xmath49 where @xmath50 is an arbitrary real parameter and @xmath51 an arbitrary matrix of dimensions @xmath52 where @xmath53 is the number of a - sites and @xmath54 the number of b sites ( in our case , @xmath55 , @xmath56 ) .
we choose @xmath57 for simplicity .
an eigenvector @xmath58 of @xmath48 corresponding to the eigenvalue @xmath59 must satisfy @xmath60 @xmath58 has @xmath53 components @xmath61 referring to a sites and @xmath54 components @xmath62 referring to b sites .
the first set of @xmath53 conditional equations is satisfied by setting @xmath63 for all @xmath64 .
the second set of equations represents a homogenous set of @xmath54 linear equations for @xmath53 unknowns .
consequently , there are at least @xmath65 linear independent and normalized solutions , i.e. the eigenvalue @xmath59 is at least @xmath66-fold degenerate ( in our case , the degeneracy is @xmath46 ) .
for @xmath31 and in the non - interacting case for @xmath16 , the @xmath67 band is fully occupied by @xmath11 electrons , and the @xmath68 band is completely empty .
further @xmath46 electrons with spin @xmath24 occupy the flat band @xmath47 .
this leads to a @xmath69-fold degeneracy of the non - interacting ground - state energy which is expected to be lifted due to interactions .
the stoner criterion @xcite shows that the paramagnetic ground state is unstable towards ferromagnetic ordering at arbitrarily small but finite @xmath70 .
let us define the ground - state expectation value of the ordered magnetic moment on a , b and impurity sites as @xmath71 that the ground state , for weak @xmath70 , is in fact fully polarized , i.e. that @xmath72 can be seen easily by first - order perturbation theory in @xmath0 : as is detailed in appendix [ appendix:1oc ] , the fully polarized state has lowest total energy . at @xmath16
we therefore pick the fully polarized state with total magnetization @xmath67 and study the @xmath15-dependence of the different contributions to @xmath73 .
this can be done by computing the spin - dependent occupation numbers for the sites @xmath74 in a unit cell via @xmath75 where @xmath76 for the lowest band , @xmath77 for the highest band , whereas @xmath78 and @xmath79 for the flat band .
using eqs .
( [ eq : qmat ] ) and ( [ eq : magdef ] ) , this yields the following site - dependent magnetic polarization in the @xmath80 limit : @xmath81 in the thermodynamical limit @xmath82 , the @xmath36-sums can be rewritten as an energy integration weighted by the free tight - binding density - of - states and evaluated numerically for @xmath18 .
the result is shown in the upper part of fig .
[ fig : mag ] ( dashed lines ) . for @xmath17
the calculation can be done analytically .
we find @xmath83 and @xmath84 these are plotted in the lower part of fig .
[ fig : mag ] ( dashed lines ) . as can be seen in the figure
, there is a crossover of the fully polarized magnetic state of the non - interacting system from the weak-@xmath15 to the strong-@xmath15 regime .
this crossover is in some respects reminiscent of the corresponding crossover of the strongly interacting system for large @xmath0 , namely from a ferromagnet driven by the effective indirect rkky interaction for weak @xmath15 to a ferromagnet driven by the effective inverse indirect exchange interaction ( iime ) for strong @xmath15 ( see discussion in secs . [
sec : dmrg ] and [ sec : iime ] ) . for @xmath39 , electrons at impurity sites are perfectly localized and form a local spin @xmath85 . in the fully polarized state , the resulting magnetic moment is @xmath86 . for finite but weak @xmath15 ,
electrons delocalize . in the non - interacting system
there is only the fermi - gas contribution to the local moments , and sizeable moments are only formed in the symmetry - broken state considered here .
these ordered local moments are mainly present on the impurity sites . in the one - dimensional case
we have @xmath87 and @xmath88 for @xmath89 , while for @xmath18 there is a non - analytical behavior of the moments @xmath90 and @xmath91 .
on the other hand , in the limit @xmath92 , electrons are perfectly localized at a sites , and thus @xmath93 in the symmetry - broken state . for finite but strong @xmath15 ,
delocalization of a - site electrons then tends to reduce @xmath94 .
we have @xmath95 and @xmath96 , where @xmath97 is the coordination number . in the entire @xmath15 range , and also independent of the dimensionality ,
the magnetic polarization at the b - sites remains zero , and hence @xmath98 .
let @xmath99 be the fermi sea that is constructed by completely filling the @xmath67 band with @xmath11 electrons and the @xmath47 band with @xmath46 spin-@xmath100 electrons , i.e. @xmath99 is a single slater determinant with a fully polarized @xmath47 band : @xmath101 for @xmath16 , this is only one ground state among the @xmath69 ground states . obviously , the macroscopically large degeneracy is due to the fact that the @xmath47 band is flat .
`` flat - band ferromagnetism '' was recognized by mielke and tasaki @xcite as a possible route to itinerant ferromagnetism in the hubbard model . for the hubbard model on lattices with certain topologies ( not necessarily bipartite ) and at certain fillings ( not necessarily half - filling ) , a fully polarized ferromagnetic state can be obtained by filling the flat band with @xmath100 electrons only . the very non - trivial and exact statement made by mielke and tasaki is that this state becomes the _
unique _ ground state for any @xmath70 ( apart from the @xmath102-fold degeneracy ) .
while the fully polarized state is not affected by @xmath0 , all other ground states gain a higher energy .
the situation considered here is somewhat different because the hubbard interaction is not present at all sites and also because there is a dispersive band below the flat band in the non - interacting band structure ( see , however , ref . ) .
in fact , we can easily show that @xmath99 can not be the exact ground state for any @xmath70 and that , therefore , the concept of flat - band ferromagnetism does not apply to the present case .
the argument is based on hartree - fock theory which can be defined by the mean - field decoupling @xmath103 which replaces the original hamiltonian eq .
( [ eq : ham ] ) by a mean - field hamiltonian that is bilinear in creators and annihilators .
the solution of the resulting effective hartree - fock equations must be obtained by computing the expectation values self - consistently from the mean - field hamiltonian . at zero temperature , the approach is fully equivalent with the variational optimization of a slater determinant consisting of @xmath104 _ a priori _ unknown single - particle orbitals using the ritz principle .
note that this _ ansatz _ also comprises the slater determinant @xmath99 in particular .
the numerical calculations clearly show that the optimized hartree - fock ground state @xmath105 has a lower total energy : @xmath106 if @xmath107 and @xmath108 .
qualitatively , this can be understood in the following way : for any @xmath15 with @xmath109 , the flat band is fully polarized and singly occupied with @xmath100-electrons in the state @xmath99 .
nevertheless , there is always a finite double occupancy on the impurity sites , @xmath110 as the impurity orbitals have contributions from all three bands ( already this fact is conflicting with flat - band ferromagnetism , see ref . , for example ) .
hartree - fock theory tends to reduce this double occupancy , @xmath111 , by increasing the polarization at the impurity sites , @xmath112 .
thereby , the interaction energy is lowered .
the magnetic polarizations at the different sites , as obtained from hartree - fock theory , are shown by the solid lines in fig .
[ fig : mag ] .
calculations have been done for @xmath113 ( @xmath17 ) and @xmath114 ( @xmath18 ) .
one clearly notes that ( i ) @xmath115 is different from @xmath116 for all @xmath15 with @xmath109 , which already reflects that @xmath117 can not be the ground state , and that ( ii ) @xmath112 due to the reduced double occupancy on the impurity sites . opposed to @xmath117
, one finds a non - zero but negative polarization at the b sites .
this is an indication of local antiferromagnetic kondo correlations .
the total magnetization is still unity , @xmath118 . , and on a and b sites , @xmath94 and @xmath119 , of the @xmath18 ( top ) and the @xmath17 ( bottom ) lattice as functions of the hybridization strength @xmath120 where @xmath121 ( @xmath18 ) and @xmath122 ( @xmath17 ) is the free band width .
dashed lines : fully polarized ground state of the non - interacting system ( @xmath16 ) .
solid lines : results of a hartree - fock calculation for @xmath123 . filled symbols for @xmath17 : dmrg calculation @xcite for @xmath123 ( @xmath124 lattice sites , @xmath125 , open boundary conditions , results are shown for the central unit cell ) . dotted line for @xmath17 : dmft calculation @xcite for @xmath123 ( @xmath126 lattice sites ) .
open symbols for @xmath18 : dmft calculation for @xmath123 ( @xmath127 ) . for the dmft calculations ,
periodic boundary conditions are assumed .
, scaledwidth=40.0% ]
for the one - dimensional case , the hartree - fock results can be compared with numerically exact data from ref .
obtained by the density - matrix renormaliization group .
@xcite dmrg calculations have been performed with a standard implementation using matrix - product forms for the trial many - body state as well as for the operators ( see ref . for some details ) .
the data shown in fig . [ fig : mag]b ( filled symbols ) have been obtained for a system with @xmath124 sites and open boundaries .
dmrg indeed predicts a ferromagnetic ground state with a total magnetization independent of @xmath15 ( see secs .
[ sec : lm ] and [ sec : iime ] for further discussion ) .
the magnetic polarizations at the different sites have been plotted for the central unit cell of the chain for the sector with maximum magnetic quantum number @xmath128 .
the dmrg ground state is unique , apart from the @xmath129-fold degeneracy . on the scale used in fig .
[ fig : mag ] , the dmrg results are representative for an infinite chain as has been verified by comparing data for different system sizes . for @xmath89 and for @xmath130 ,
hartree - fock theory reproduces the exact result for @xmath94 , @xmath119 @xmath131 and for the total magnetization @xmath73 . for intermediate @xmath15
, however , there are strong deviations : here , hartree - fock theory overestimates the absolute strengths of the polarizations on the impurity and on the b sites , and underestimates @xmath132 .
we conclude that hartree - fock theory is able to roughly describe the trends of the magnetic polarizations with @xmath15 but does not allow for quantitative predictions and even qualitatively fails to describe the magnetic order in terms of spin - correlation functions .
the magnetic ground state of the considered periodic anderson model with depleted magnetic impurities must therefore be seen as being correlated .
it can not be described by a fully polarized fermi sea or by any other ( variationally optimized ) slater determinant .
although the ground state is correlated , there are exact statements available for the particular geometrical and electronic structure considered here .
this goes back to the lieb - mattis theorem @xcite which states that the antiferromagnetic spin-@xmath85 heisenberg model on an arbitrary bipartite lattice has a unique singlet ground state if the number @xmath53 of a sites equals the number @xmath54 of b sites . for @xmath133 ,
the ground state has total spin @xmath134 and is unique apart from the trivial @xmath135-fold degeneracy .
the proof of the theorem is based on the positivity of the hamiltonian after a unitary spin - reflection transformation which permits the application of the perron - frobenius theorem .
the concept of `` spin - reflection positivity '' has been used to generalize the lieb - mattis theorem to the half - filled repulsive hubbard model with constant @xmath136 for any site @xmath7 on a bipartite lattice .
@xcite for the half - filled and symmetric periodic anderson model ( with as many localized as conduction orbitals , i.e. , @xmath137 ) on a bipartite lattice , a modified proof is possible .
@xcite analogous results are also available for the half - filled kondo - lattice model .
@xcite let us now focus on the depleted periodic anderson model at half - filling that is considered here .
as there are additional sites ( the a sites ) with vanishing hubbard - type interaction , the ground state is not necessarily unique , i.e. the lieb - mattis theorem ( and generalizations thereof ) can not be applied in this respect .
in fact , for certain @xmath11 and @xmath138 , one can easily find examples with a degenerate ground state .
the most simple case is probably given by @xmath139 and @xmath140 where the impurity site is attached to the central site of the three - site tight - binding chain with open boundaries .
a simple calculation shows that the eigenstates of lowest energy are given by a triplet , @xmath141 , and a singlet , @xmath142 , for any @xmath70 , i.e. the ground - state energy is four - fold degenerate .
however , one of the ground states ( if there is degeneracy ) is in the sector with @xmath141 .
this is similar to the depleted kondo lattice .
there is a theorem , proven by shen , @xcite stating that for @xmath138 localized spins @xmath85 coupling via a local antiferromagnetic exchange to a half - filled system of otherwise non - interacting conduction electrons on a bipartite @xmath10-dimensional lattice with @xmath143 sites , there is at least one ground state with @xmath134 .
degeneracy can be excluded for the dense case , @xmath137 , or for a finite hubbard - type interaction @xmath70 among the conduction electrons.@xcite the dmrg calculations ( fig .
[ fig : mag ] ) for @xmath124 and @xmath125 yield a ground - state total spin @xmath144 .
this is consistent with the above prediction , i.e. @xmath145 . actually , the system is ferrimagnetic , also consistent with the sign of the spin - spin correlation functions that can be predicted exactly .
@xcite furthermore , the ground state turns out to be unique ( apart from the trivial @xmath102-fold degeneracy ) .
opposed to the exact theorems available and opposed to the essentially exact numerical data , perturbative approaches are able to provide a more physical understanding for the emergence of ferromagnetic order .
moreover , they can explain which sites , depending on @xmath15 , carry the main part of the total magnetic moment and also why the ground state is in fact unique . for @xmath146 and weak @xmath15 ,
the model maps onto a depleted kondo lattice where the correlated sites are replaced by spins @xmath85 coupled via an antiferromagnetic local exchange @xmath147 to the conduction - electron system . @xcite
as @xmath14 is small in this limit , rkky second - order perturbation theory applies .
@xcite at half - filling the effective non - local spin - spin interaction @xmath148 has an oscillatory dependence on the distance @xmath149 between two local spins and is ferromagnetic for the model considered here .
hence , the ferromagnetic ground state results from ferromagnetic rkky coupling of well - formed local moments at the impurity sites . however , the total spin is @xmath150 rather than @xmath151 as could have been expected by simply coupling all impurity spins .
the reason for the `` missing spin 1/2 '' is a variant of the kondo effect that has been studied in ref .
and that applies to systems of finite size where perturbation theory in @xmath14 is well behaved .
namely , in the weak - coupling limit @xmath89 and for the particular system with @xmath124 and @xmath125 , exactly one of the impurity spins is kondo screened , and only the remaining ones are subjected to the ferromagnetic rkky interaction .
the `` missing spin '' can be understood with the help of the _ original _ lieb - mattis theorem : using standard degenerate perturbation theory in first order in @xmath14 for the depleted kondo lattice , we obtain the following effective model : @xmath152 where @xmath153 is a spin @xmath85 at the impurity site @xmath26 , and @xmath154 is the spin of the conduction electron at the fermi energy . the effective exchange couplings @xmath155 are given in terms of its single - particle wave function @xmath156 and are all positive .
( [ eq : csm ] ) represents an antiferromagnetic central - spin heisenberg model .
as this has the topology of a bipartite lattice , the lieb - mattis theorem @xcite immediately predicts a unique ground state with @xmath150 . in the opposite limit of strong hybridization @xmath130 , the electrons at the impurity and at the b sites form strongly localized and magnetically inert singlet bound states
this implies that the total magnetic moment is no longer carried by the impurity sites .
however , these `` anderson singlets '' tend to localize the electrons at the a sites , and this results in the formation of strong local a - site magnetic moments
. there is an indirect coupling between the a - site moments which is ferromagnetic and induced via virtual excitations of the anderson singlets .
this is similar to the strong-@xmath14 limit of the depleted kondo lattice studied in ref . .
a analogous construction of an effective low - energy model by employing strong - coupling perturbation theory is also possible in the anderson case ( and will be published elsewhere ) .
the resulting `` inverse indirect magnetic exchange '' ( iime ) represents the strong - coupling ( strong @xmath15 ) analogue of the rkky mechanism at weak coupling .
the transition from the rkky to the iime regime as a function of @xmath15 is a smooth crossover : as can be seen in fig .
[ fig : mag ] , the polarization at the impurity sites continuously decreases while the a - site polarization increases .
the negative polarization at the b sites attains a maximal absolute value in the crossover regime @xmath157 .
the symmetry of the ground state and the total spin @xmath158 does not change with @xmath15 . in the strong - coupling limit ,
the total spin @xmath150 is due to the ferromagnetic iime coupling of the @xmath159 a - site moments .
it is instructive to compare the results for the magnetic polarization at the different sites that have been obtained by static mean - field theory and ( for @xmath17 ) by dmrg with corresponding results of dynamical mean - field theory ( dmft ) .
@xcite the dmft treats the local correlations exactly , and in particular the kondo screening of the magnetic impurities .
it also accurately predicts the indirect inter - impurity magnetic coupling as has been studied before in ref . .
on the other hand , the feedback of the non - local magnetic correlations on the one - particle excitation spectrum is neglected within dmft as this would imply non - local contributions to the electron self - energy .
this feedback effect , however , can be expected as weak for the present case of a system with regularly depleted impurities : as the self - energy is non - zero on the correlated sites only , a non - local self - energy diagram must include non - local green s functions between sites that are separated by 4 or more nearest - neighbor hopping steps . an almost local self - energy and thus a single - site dmft approach should therefore be a reasonable approximation .
we employ a standard implementation of the dmft using the exact - diagonalization solver , @xcite i.e. the ground state and the single - electron excitation spectrum of the effective impurity problem is obtained by the lanczos method .
calculations have been performed for anderson impurity models with @xmath160 sites .
a fictitious temperature of @xmath161 is used for the low - energy cutoff of the dmft self - consistency equation .
lattices with periodic boundary conditions and a sufficiently large number of sites @xmath11 are considered to ensure that the results are free of finite - size errors . for @xmath123 and in the entire range of hybridization strengths @xmath15 , there is a ferromagnetic solution of the dmft equations which is found by running through the usual dmft self - consistency cycle . a paramagnetic solution , on the other hand , could not be stabilized without enforcing spin - symmetric parameters .
we have checked that the ferromagnetic state has the lower total energy .
the results for the site - dependent polarizations of the depleted anderson lattice are shown fig .
[ fig : mag ] as dotted lines for dimension @xmath17 and as open circles for @xmath18 . in both cases ,
we find strong differences between the static ( hartree - fock ) and the dynamical mean - field results .
hence , there are sizable effects resulting from local correlations .
comparing with the numerically exact dmrg data for the @xmath17 case , we furthermore see that the dmft is not only able to qualitatively reproduce the crossover from the rkky to the iime regime but also predicts the magnetic properties of the system correctly on a quantitative level .
this is a remarkable finding as the @xmath17 case is actually the worst case from the dmft perspective .
one may exploit this to study more complex configurations of magnetic impurities in higher dimensions .
here , we will make use of the dmft by accessing the single - electron excitation spectrum to study the fate of the flat band for a correlated system .
this can be done by looking at the low - frequency `` coherent '' part of the single - electron excitation spectrum . within the dmft framework
, the coherent spectrum can be analyzed easily .
the @xmath36 and @xmath162 dependent single - electron @xmath35 greens - function matrix is obtained from dyson s equation as @xmath163 the @xmath35 hopping matrix @xmath164 is given by eq .
( [ hopping ] ) .
the ( @xmath36-independent ) self - energy matrix @xmath165 and the matrix of on - site energies @xmath166 are diagonal and non - zero on the impurity sites only . assuming that the system is a ferromagnetic fermi - liquid
, we can expand the self - energy for low frequencies : @xmath167 where @xmath168 is a real on - site energy shift and @xmath169 is the quasi - particle weight . inserting this into eq .
( [ green ] ) and neglecting quasi - particle damping effects @xmath170 , we obtain the coherent part of the green s function : @xmath171 comparing with the green s function of the non - interacting system , we find ( i ) a spin - dependent shift of the effective on - site energy at the correlated site , @xmath172 $ ] , ( ii ) a band renormalization @xmath173 with the quasi - particle weight matrix @xmath174 as well as ( iii ) an overall scaling of the local quasi - particle density of states at the impurity sites . and
on - site energy shift @xmath175 for a half - filled two - dimensional depleted anderson lattice with @xmath176 sites for @xmath114 as functions of @xmath15 as obtained by dmft ( red circles , bottom and left axis ) and @xmath177 for the case of kondo impurities as a function of @xmath14 ( green circles , top and right axis ) .
spin dependence : @xmath178 , anderson impurities : @xmath179 , kondo impurities : @xmath180 .
dashed line : static mean - field result for @xmath181 in the anderson case .
note that the dmft self - energy is non - zero on the correlated sites in the anderson model while for the kondo model the self - energy is non - zero at the conduction - electron b sites
. , scaledwidth=40.0% ] the poles of the coherent green s function close to @xmath182 determine the dispersion of the coherent quasi - particle band . at half - filling , given for @xmath31 , and in the spin - symmetric paramagnetic state with @xmath183 , the effective hopping matrix , eq .
( [ eq : effhopp ] ) , is identical with the hopping matrix in the non - interacting case except for a renormalization @xmath184 of the hybridization .
hence , we find a flat quasi - particle band at @xmath185 .
this is consistent with the expectation that a correlation - induced `` band narrowing '' of an already non - dispersive band does not have any effect .
however , this must be seen as an artifact of the dmft as generally the self - energy and thus the parameters @xmath186 and @xmath187 acquire a @xmath36-dependence which directly leads to a dispersion of the quasi - particle band .
anyway , already on the dmft level , the ferromagnetic long - range order implies that the coherent part of the excitation spectrum is dispersive since the spin - dependence of @xmath175 also implies different effective on - site energies of the a and b sites of the bipartite lattice ( see sec .
[ sec : non ] ) .
[ fig : qpw ] displays the parameters @xmath188 and @xmath175 as obtained from a dmft calculation for a two - dimensional strongly correlated system ( @xmath189 ) as functions of @xmath15 ( red symbols ) .
a lattice with @xmath176 sites is sufficient to ensure that the results are not affected by finite - size effects . in the ferromagnetic state at
half - filling we have a spin - independent quasi - particle weight @xmath178 but @xmath190 while @xmath191 .
the modulus of the deviation from @xmath192 ( plotted in the figure ) is spin - independent , similar but smaller than the corresponding static mean - field result @xmath193 . in the weak-@xmath15 limit ,
the system is effectively equivalent to the corresponding depleted kondo lattice .
localization of electrons at the impurity sites and local - moment formation drive the system to a strongly correlated state with @xmath194 . for large @xmath15 , we find @xmath195 and @xmath196 : due to strong charge fluctuations on the impurity sites , the system s low - frequency one - electron excitation spectrum is well described by non - interacting values of the parameters .
calculations for a two - dimensional depleted anderson lattice with @xmath197 sites using the low - frequency parameters @xmath198 from a dmft calculation for @xmath114 and @xmath199 ( see fig . [
fig : qpw ] ) .
blue : spin - up , red : spin - down coherent density of states . dashed lines : the same but for the case of kondo impurities ( dmft calculation for @xmath200 ) .
, scaledwidth=40.0% ] using the parameters from the dmft calculation , we obtain the `` coherent '' part of the spin - dependent local density of states ( dos ) projected onto a , b , and impurity sites , @xmath201 see fig .
[ fig : qps ] .
a somewhat larger lattice ( @xmath202 ) is necessary to completely suppress finite - size effects on the scale of the figure .
we discuss results for hubbard interaction @xmath114 . in this strong - coupling regime ,
the one - electron excitation spectrum shows incoherent hubbard bands separated by an energy of the order of @xmath0 .
the coherent part of the one - electron excitation spectrum consists of three quasi - particle bands which are separated by an energy of the order of @xmath15 . only the central band around the fermi edge ( @xmath185 )
is plotted in fig .
[ fig : qps ] , and here we focus on the low - frequency limit only where the fermi - liquid form of the self - energy , eq .
( [ eq : expand ] ) is expected to hold .
first of all , the central quasi - particle band is dispersive , and the coherent part of the density of states has a finite width as discussed above .
however , the coherent dos at the b sites is roughly two orders of magnitude smaller than at the a and at the impurity sites .
this is still reminiscent of the flat - band picture that is found at @xmath16 or , at least within the dmft , in the ( metastable ) paramagnetic phase where the central quasi - particle band is of a- and of the impurity - sites character only while the weight of the b sites is zero .
interestingly , the coherent density of states exhibits a van hove singularity at @xmath203 which may be seen as another reminiscence to the flat - band case at @xmath16 where the dos includes a @xmath204-peak at @xmath203 . in the correlated state , however , the singularity is weaker and only the dos at the a sites is divergent .
hence , the metallic character of the system is dominated by electron transport via the a sites .
furthermore , the coherent part of the dos is fully polarized , i.e. @xmath100 and @xmath205 dos do not overlap . note that not only for the impurity and the a sites but also for b sites only the @xmath100 coherent dos is occupied .
however , @xmath206 as is seen from fig .
[ fig : mag ] for @xmath207 and @xmath114 where the magnetic state of the system is found as intermediate between the rkky and the iime limits .
this implies that larger ( negative ) high - frequency contributions to the magnetic moment at the b sites must outweigh this effect .
concluding , strong correlations drive the flat - band system to a ferromagnetic fermi - liquid state with unconventional low - energy particle - hole excitations . while the spin-@xmath100 as well as the spin-@xmath205 dos are gapped , the total dos is gapless .
hence , low - energy particle - hole excitations contributing to the static conductivity must be accompanied by a spin flip .
this scenario can be tested in various ways : first , to simplify the system and to exclude the effect of charge fluctuations , we compare the results obtained for anderson impurities with corresponding ones for kondo impurities , i.e. we turn to the following hamiltonian for a depleted kondo lattice : @xmath208 here , @xmath153 is a spin-@xmath85 at site @xmath26 that couples via an antiferromagnetic local exchange @xmath209 to the spins of conduction electrons at site @xmath8 , the chemical potential is @xmath31 at half - filling , and @xmath210 is the vector of pauli matrices . for the dmft calculations this depleted kondo - lattice model is self - consistently mapped onto an effective impurity model where the correlated impurity consists of a b - site @xmath8 with the local spin at @xmath26 attached and is embedded in an uncorrelated bath .
@xcite an effective impurity model with @xmath211 sites ( plus the local spin @xmath85 ) is considered in practice and treated with the lanczos technique .
dyson s equation is again given by eq .
( [ eq : dyson ] ) but now the hopping and the self - energy in @xmath36 space are @xmath212 matrices : @xmath213 expanding the self - energy for @xmath214 , eq . ( [ eq : expand ] ) , leads to eq .
( [ eq : gcoh ] ) with an effective on - site energy at the correlated b site , @xmath215 $ ] , and to a band renormalization , @xmath216 , with the quasi - particle weight matrix @xmath217 in the metastable paramagnet at half - filling , the effective hopping matrix is given by the non - interacting one , except for a renormalization @xmath218 of the bare dispersion .
this leads to a completely flat quasi - particle band . as in the anderson case
, however , the ferromagnetic long - range order makes the band dispersive since @xmath219 implies different effective on - site energies of the a and b sites .
the parameters @xmath188 and @xmath175 characterizing the low - frequency part of the self - energy are shown in fig .
[ fig : qpw ] as a function of @xmath14 .
the quasi - particle weight @xmath178 at the b sites and the modulus of the on - site energy shift @xmath220 are spin - independent but @xmath221 while @xmath222 . when comparing the low - energy electronic structure of the anderson and the kondo model with each other , it is important to recall that the dmft construction is different : while for the anderson case a correlated site with @xmath70 in the lattice model must be identified with the impurity site of the effective anderson impurity model , the spin @xmath85 _ and _ the attached b site have to be considered as the impurity in the kondo case .
@xcite consequently , the self - energies are non - zero on correlated ( @xmath70 ) sites in the anderson and non - zero on b - sites in the kondo case . in the weak-@xmath14 limit
, the anderson and the kondo model can be mapped onto each other . here
, electrons are only weakly scattered from the impurities and thus @xmath195 . for strong @xmath14
, on the other hand , an impurity spin and an electron at a b site form an almost local kondo singlet .
scattering of a - site electrons from these local kondo singlets tends to localize electrons at a sites and results in a strongly correlated fermi - liquid state with small @xmath188 .
furthermore , the iime mechanism drives the system to the ferromagnetic state as discussed in sec .
[ sec : iime ] .
the resulting `` coherent '' part of the spin - dependent local density of states ( dos ) projected onto a and b sites is shown in fig . [
fig : qps ] for @xmath200 ( dashed lines ) .
the dos resulting from the coherent low - energy band around the fermi edge has a finite width .
its projection onto the b sites is an order of magnitude smaller than the projection on the a sites .
the coherent dos is again fully polarized and also exhibits the same ( singular ) structure at @xmath203 that was found for the anderson case .
we conclude that even in the strong-@xmath14 limit where the kondo lattice does not map onto the anderson lattice , the low - frequency physics is qualitatively the same .
green s function for the one - dimensional depleted anderson model as obtained from dmft for @xmath223 and @xmath113 . imaginary ( blue ) part , @xmath224 , and real part ( red ) , @xmath225 , for @xmath74 and different system sizes @xmath11 as indicated .
note the double logarithmic scale . ,
scaledwidth=40.0% ] but for the depleted kondo lattice at @xmath226 . ] as the dmft is able to quantitatively describe the correlated magnetic ground state even for the one - dimensional system , it is worthwhile to discuss the @xmath17 single - particle excitation spectra . the local spin-@xmath100 green s functions for a , b and impurity sites in the range of small imaginary frequencies are shown in figs . [
fig : gand ] and [ fig : gkon ] for the case of anderson ( @xmath227 ) and for the case of kondo impurities ( @xmath226 ) , respectively .
the spin-@xmath205 green s functions are fixed by the relations @xmath228 which are enforced by particle - hole symmetry .
results are displayed for different finite systems with @xmath229 and for the infinite system . as can be seen from the figures ,
finite - size effects start to play a dominant role in the low - frequency regime for @xmath230 .
the low - frequency behavior of the green s function for @xmath231 can be obtained from the dyson equation ( [ eq : dyson ] ) by computing the matrix inverse analytically and by replacing the @xmath36-sum with an integration @xmath232 weighted with the non - interacting density of states @xmath233 .
for @xmath17 we have @xmath234 , and the @xmath50-integration can be done analytically . for low frequencies , i.e. using eq .
( [ eq : expand ] ) , we then find for the anderson case : @xmath235 while for the case of kondo impurities @xmath236 this perfectly agrees with the data shown in figs . [ fig : gand ] and [ fig : gkon ] . as in the @xmath18 case
( see fig .
[ fig : qps ] ) , we find a diverging a - site quasi - particle density of states at the fermi edge @xmath237 for @xmath238 while the b - site and the impurity dos vanish .
in fact , for @xmath18 the low - frequency quasi - particle dos is similar , @xmath239 and @xmath240 , but with additional logarithmic corrections that are traced back to the van hove singularity of the @xmath18 non - interacting density of states .
concluding , the interaction - induced renormalization of the low - energy one - particle excitation spectrum generates dispersive quasi - particle bands with characteristic van hove singularities .
let us emphasize that this is the mean - field and fermi - liquid picture for the excitation spectrum which can be provided by the dmft but which is likely to be invalidated for @xmath17 below some low - frequency scale by coupling to bosonic long - wave - length modes . for the finite - size systems ,
the dmft picture is expected to be more adequate as those modes are cut by the finite - size gap .
[ fig : gand ] and [ fig : gkon ] show that , below a certain frequency scale of the order of @xmath241 , the a - site green s function behaves as @xmath242 while for the b - site green s function ( and likewise for the impurity green s function ) we have @xmath243 with positive constants . according to the lehmann representation of the green s function for a system of finite size
, @xmath244 this implies that there is a pole at zero frequency , @xmath245 , with a finite weight @xmath246 on the a sites while @xmath247 .
generally , @xmath248 , where @xmath249 label the mutually orthogonal ground states of the system , and @xmath250 is the ground - state degeneracy . here , a two - fold degeneracy of the ground state arises from the fact that , for any finite @xmath11 , two eigenvalues of the effective hopping matrix @xmath251 in eq .
( [ eq : gcoh ] ) are vanishing , namely at @xmath252 and at @xmath253 .
the corresponding eigenvectors have 100% a - character .
the same behavior of @xmath254 is also found for higher dimensions but finite @xmath11 , where there is at least a two - fold ground - state degeneracy .
a diverging total density of states at the fermi edge implies a diverging charge response to a change of the chemical potential , i.e. a diverging charge susceptibility @xmath255 where @xmath256 is the average total particle number per site .
hence , we can check the dmft results by comparing with dmrg data for @xmath257 available for @xmath17 , finite @xmath11 and half - filling .
@xmath257 is related to the charge gap , @xmath258 which is obtained from the ground - state energy @xmath259 in the sector with total particle number @xmath104 and @xmath260-component of the total spin @xmath261 .
we have @xmath262 calculations have been performed for sectors with different @xmath261 . a finite @xmath257 is only obtained if @xmath263 , and the results for different @xmath261 agree . for @xmath128 ,
corresponding to the symmetry - broken state that is realized in the related dmft calculation , however , @xmath257 vanishes in the limit @xmath82 .
the reason is that particle - hole excitations with arbitrarily low excitation energy would have to be accompanied by a spin flip , as it has already been discussed in sec .
[ sec : coh ] on the dmft level , while @xmath257 is sensitive to the spin - independent charge response only . as a function of inverse system size @xmath11 for the one - dimensional depleted kondo lattice .
result obtained by dmrg for different @xmath14 as indicated and for @xmath264 .
the dashed line is a fit to the data for @xmath226 .
, scaledwidth=40.0% ] fig .
[ fig : kappa ] displays dmrg data for @xmath257 obtained for the @xmath264 sector of the kondo lattice at different @xmath14 in the strong - coupling regime and different system sizes @xmath11 .
as is demonstrated by finite - size scaling for @xmath226 , the charge susceptibility diverges with increasing system size .
this is fully consistent with the interpretation of the dmft results .
the fact that @xmath257 is increasing with increasing @xmath14 can be attributed to the increasing tendency to form local kondo singlets at the b sites and thereby to localize electrons at a sites .
a diverging @xmath257 in many cases indicates an instability of the ground state towards phase separation ( see ref . for an example of phase separation in a different fermion - lattice model at half - filling ) :
an s - shaped trend of @xmath265 with @xmath266 for @xmath267 implies that there is a finite range of chemical potentials , @xmath268 , where @xmath269 , i.e. a range where a homogeneous phase is thermodynamically unstable and where an inhomogeneous state composed of macroscopically separated regions with different fillings has a lower grand potential . in the present case , however , the divergence of @xmath257 originates from the van hove singularity of the density of states at the fermi edge and will thus be absent for fillings off half - filling .
we therefore expect that there is no _ finite _ @xmath270 range , i.e. @xmath271 , or , stated differently , that @xmath272 has a saddle point at half - filling rather than a maximum and a minimum at fillings below and above half - filling , respectively .
this is corroborated by the dmrg calculations for @xmath273 where we did not find any tendencies for the formation of an inhomogeneous ground state .
the periodic anderson model with regularly depleted impurities , or the depleted kondo - lattice model , provide the most simple setup to study ferromagnetic order of localized moments induced by the indirect rkky magnetic exchange . on a bipartite lattice with conduction - electron hopping between nearest - neighbor sites and for a half - filled conduction band ,
the period of the effective rkky exchange couplings is commensurate with the lattice and is ferromagnetic if the distance between the impurities is @xmath274 .
we have studied the magnetic properties of this model on a one - dimensional chain and on the two - dimensional square lattice in the rkky regime , i.e. at weak hybridization strength @xmath15 , or weak kondo coupling @xmath14 , but also the crossover to the strong-@xmath15 or strong-@xmath14 regime . besides the rkky approach , there are several and rather different theoretical concepts that apply to this case .
a variant of the lieb - mattis theorem for the kondo model @xcite predicts that there is a ferromagnetic ground state with a macroscopically large total spin quantum number @xmath158 among the different ground states ( if there is ground - state degeneracy ) . in certain geometries ,
@xmath158 is less by one as compared to the naive application of the rkky theory , and the ( finite - size ) kondo effect must be considered in addition for an explanation .
@xcite perturbative approaches in the weak - coupling @xcite and the strong - coupling ( strong @xmath15 or @xmath14 ) cases @xcite can be employed to clarify whether the ground state is degenerate or not ( apart from the trivial @xmath129 spin degeneracy ) . for strong @xmath15 or @xmath14 in particular
, the concept of the inverse indirect magnetic exchange ( iime ) explains why there is ferromagnetic order despite the fact that the impurity magnetic moments are kondo screened , namely local magnetic moments at the intermediate a sites are formed and couple magnetically via virtual excitations of the kondo singlets .
for the one - dimensional case the crossover from the rkky to the iime limit is nicely seen in the local ordered magnetic moment and in spin correlation functions with the help of density - matrix renormalization .
another route to ferromagnetic order that may apply to the class of systems considered here is the concept of flat - band ferromagnetism .
@xcite in fact , the bipartite geometry , the restriction to nearest - neighbor hopping and the depletion of impurities with @xmath274 straightforwardly implies the emergence of a flat band at the fermi energy of the non - interacting ( @xmath16 ) system .
this implies that the @xmath16 ground - state energy is highly degenerate and that the `` fully polarized '' state where the flat band is exactly half - filled with spin-@xmath100 electrons only is among the ground states .
it is remarkable that the computed @xmath15- and site - dependent local magnetic polarizations in this state already roughly capture the main trend , namely ferromagnetic order sustained by the impurities with @xmath90 but @xmath91 in the rkky limit @xmath275 while for the iime limit @xmath130 we have @xmath276 but @xmath277 and ferromagnetism is sustained by conduction electron localized at a sites .
the slater determinant with a fully polarized flat band would be the only ground state for any finite @xmath0 if the concept of flat - band ferromagnetism applies .
however , already hartree - fock theory shows that this is not the case .
furthermore , comparing the predictions of hartree - fock theory with the essentially exact dmrg data , we conclude that the ground state is highly correlated rather than a simple fermi sea for intermediate or strong @xmath0 .
hartree - fock theory in fact provides a fairly good but still rough description of the symmetry broken state .
more surprising is the fact that the dynamical mean - field theory , even for the one - dimensional case , yields _ quantitatively _ almost exact results as is verified by comparing with the dmrg data .
we conclude that the depletion of the impurities drives the system to a state which is strongly correlated ( and thus different from the static mean - field state ) but where the correlations are mainly temporal rather then spatial ( and thus accessible to the dynamical mean - field approach ) .
technically , the finite distance @xmath274 between the impurities implies that the non - local contributions to the electron self - energy become negligibly small as already the lowest - order non - local corrections scale with the third power of the fourth - nearest neighbor element of the non - interacting green s function .
this offers the exciting perspective that a comparatively simple dmft approach can be employed to quantitatively describe the indirect magnetic coupling and the resulting magnetic order of nanostructures in higher spatial dimensions , e.g. on a two - dimensional metallic surface .
future work will have to address the magnetic and correlated electronic structure of depleted anderson- and kondo - lattice models away from the particle - hole symmetric point off half - filling .
this brings in different new aspects , such as , for example , the absence of a lieb - mattis theorem or the incommensurability of the rkky couplings with the lattice constant .
a rather complex magnetic phase diagram can be expected .
furthermore , it will be interesting to study the filling dependence of the single - particle excitation spectrum . here , for the case of half - filling , we have found a rather unconventional low - frequency electronic structure using the dmft , namely a gapless metallic spectrum but with a completely filled @xmath100 and empty @xmath205 `` coherent '' part of interacting density of states .
ferromagnetic order at @xmath70 has been seen to necessarily result in a finite quasi - particle dispersion of the originally ( for @xmath16 ) entirely flat band , i.e. a `` correlated flat band '' is no longer flat .
due to the bipartite lattice structure and the manifest particle - hole symmetry at half - filling the interacting density of states develops ( weaker ) singularities at the fermi edge which can be understood as van hove singularities of the low - frequency coherent quasi - particle band structure .
these result in a diverging homogeneous charge susceptibility as could also be verified for the one - dimensional case by dmrg but are not expected to drive the system to a phase - separated state . financial support of this work by the deutsche forschungsgemeinschaft within the sfb 668 ( project a14 ) and within the excellence cluster `` the hamburg centre for ultrafast imaging - structure , dynamics and control of matter at the atomic scale '' is gratefully acknowledged .
according to stoner s criterion and due to the occurrence of a flat band at the fermi energy , the weakly interacting depleted periodic anderson model at half - filling should be unstable towards ferromagnetic ordering . here , we summarize the results of first - order perturbation theory in @xmath0 for the hamiltonian eq .
( [ eq : ham ] ) . the mean - field decoupling eq .
( [ mf_decoupling ] ) is exact up to first order in @xmath0 and provides us with a simplified hamiltonian @xmath278 which is given by @xmath279 , eq .
( [ h0 ] ) , with @xmath280 replaced by @xmath281 , and where @xmath282 is a @xmath35 matrix with @xmath283 for all @xmath284 except for @xmath285 where we have @xmath286 with @xmath287 . obviously , there is no difference to the non - interacting case for the paramagnetic state . for the ferromagnetic state
, @xmath281 is diagonalized by the unitary transformation @xmath288 up to first order in @xmath0 where @xmath289 is given by eq .
( [ eq : qmat ] ) and @xmath290 therewith , up to first order in @xmath0 , we find the eigenvalues @xmath291 [ compare with eq .
( [ eq : spectrum ] ) ] where @xmath45 .
analogously to eqs .
( [ eq : magdef ] ) and ( [ eq : ns ] ) , we get the site - dependent magnetic polarizations @xmath292 the non - interacting values are given by eq .
( [ eq : ms ] ) .
note that @xmath293 .
the first - order effect of the hubbard - type interaction is thus to increase @xmath131 and @xmath294 and to decrease @xmath94 .
the total energy of magnetic ground state , @xmath295 ( @xmath296 ) is lower than the energy of the paramagnetic state which is obtained by setting @xmath297 and which is equal to the total energy of the non - interacting system . | the magnetic ground - state properties of the periodic anderson model with a regular depletion of the correlated sites are analyzed within different theoretical approaches .
we consider the model on the one - dimensional chain and on the two - dimensional square lattice with hopping between nearest neighbors . at half - filling and with correlated impurities present at every second site ,
the depleted anderson lattice is the most simple system where the indirect magnetic coupling mediated by the conduction electrons is ferromagnetic .
we discuss the underlying electronic structure and the possible mechanisms that result in ferromagnetic long - range order . to this end
, different numerical and analytical concepts are applied to the depleted anderson and also to the related depleted kondo lattice and are contrasted with each other .
this includes numerical approaches , i.e. hartree - fock theory , density - matrix renormalization and dynamical mean - field theory , as well as analytical concepts , namely a variant of the lieb - mattis theorem and the concept of flat - band ferromagnetism , and finally perturbative approaches , i.e. the effective rkky exchange in the limit of weak and the `` inverse indirect magnetic exchange '' in the limit of strong coupling between the conduction band and the impurities . |
accretion and clustering are the most fundamental aspects of ob star formation ( see garay & lizano 1999 , mckee & ostriker 2007 and zinnecker & yorke 2007 for reviews ) .
ob stars are observed to form in very special regions of molecular clouds massive molecular clumps , typically of parsec scale in size .
large mass and high density are needed for the accretion flow to feed the forming massive stars as well as the associated stellar cluster at a high enough rate .
however , how the massive clumps form , whether the internal conditions of the massive clumps facilitate the accretion process , and how the massive clumps fragment into clusters , are still open questions . in the present work ,
we improve our understanding of these issues observationally , using the submillimeter array ( sma ; ho , moran , & lo 2004 ) and the iram 30 m telescope .
we focus on the well studied uc hii region g10.6 - 0.4 , for which a wealth of complementary data are available to help address these questions .
g10.6 - 0.4 is a well studied ob cluster forming region at a 6 kpc distance ( caswell et al .
1975 ; downes et al . 1980 ) . in the central @xmath21 pc region ,
high bolometric luminosity ( 9.2@xmath310@xmath4 l@xmath1 ) and bright free - free continuum emission ( @xmath52.6 jy within a 0.05 pc radius , at 1.3 cm band ) were detected ( ho & haschick 1981 ; sollins et al . 2005 ; sollins & ho 2005 ) , suggesting that a cluster of o - type stars has formed ( o6.5b0 ; ho and haschick 1981 ) .
observations of the 20 cm continuum emission using the nrao very large array ( vla ) unveil a @xmath65 pc scale ionized bubble to the north of the brightest ob cluster , which suggests that the evolution of the cloud can be affected by the impact of the hii region projected close to g10.60.4 ( ho , klein & haschick 1986 ) .
interferometric observations with high resolutions detected groups of water and oh masers ( genzel & downes 1977 ; ho & haschick 1981 ; ho et al . 1983 ; hofner & churchwell 1996 ; fish et al .
2005 ) , and multiple high velocity @xmath7co outflows ( liu , ho & zhang 2010 ) , which indicates on going and in situ star formation . observations of various molecular transitions ( ho & haschick 1986 ; keto , ho , & haschick 1987 ; keto , ho , & haschick 1988 ; guilloteau et al .
1988 ; omodaka et al . 1992 ; ho , terebey , & turner 1994 ; klaassen et al .
( 2009 ) ; liu et al . 2010 ; liu , zhang & ho 2011 ; beltrn et al .
2011 ) suggest that the general motion of the molecular gas at the 1 pc scale is dominated by gravity and shows rotation along a flattened geometry .
the overall geometry and the excitations of the molecular gas in the central @xmath61 pc region resemble a scaled up low
mass star forming core ( liu , zhang , & ho 2011 ) , implying that the global contraction is efficient .
a similar geometry in the central 1 pc scale is also resolved in the luminous ( [email protected]@xmath810@xmath4 l@xmath1 ) massive cluster forming region g20.08 - 0.14 n ( galvn - madrid et al .
2009 ) . the high resolution spectral line observations further show the clumpiness and the spatial asymmetry of the mass and the velocity field at a scale of 0.5 pc ( liu et al . 2010 ; liu , zhang & ho 2011 ) .
this asymmetric flow appears to have a spin up rotational motion , and might continue to the highly clumpy 0.1 pc scale , flattened , hot toroid , which was first resolved by the nh@xmath9 optical depth studies .
the hot toroid has a high temperature and should immediately surround the central 200 m@xmath1 ob cluster ( sollins & ho 2005 ; liu et al .
2010 ; beltrn et al .
2011 ) , suggesting that massive stars may accrete from the very clumpy and geometrically thick molecular materials .
observations of the radio recombination lines suggest that part of the material that is ionized by the embedded ob stars can continue across the hii boundary , to feed the embedded ob stars ( keto 2002 ; keto & wood 2006 ) .
despite the rich literatures in the studies of the kinematics in the inner 1 pc region , how the parsec scale massive molecular clump connects to the more extended structure , how the massive clumps form , and how the detailed morphology and dynamics lead to the fragmentation and the formation of a cluster , remain unanswered questions . to improve the understanding in these aspects , we observed the 1.2 mm continuum emission in a 10 pc scale area using the iram 30 m telescope .
at such a large scale , the 1.2 mm continuum emission is dominated by the thermal dust emission .
the large dish of the iram 30 m telescope allows us to resolve the dust emission , and simultaneously provides the short spacing information , to complement the sma data . this yields high resolution and high dynamic range images for our morphological studies .
in addition , we observed the dust polarization properties at the 0.87 mm wavelength using the sma , to gauge the role of the magnetic field in the formation of the massive molecular clump .
for a few 1.2 mm clumps resolved in the iram 30 m observations , we followed up with the sma at 0.87 mm , in order to resolve their internal structures and subsequent fragmentation .
we also compare the large scale 1.2 mm continuum emission with the spitzer mipsgal 24 @xmath0 m map and the glimpse 8 @xmath0 m map .
the diffuse 24 @xmath0 m and 8 @xmath0 m emissions mainly trace the ionized gas and the photon dominant regions ( pdr ) at the boundary of the neutral material .
these comparisons help to define the locations and the effects of both the embedded and the external ionizing sources .
we retrieved the archived vla 20 cm continuum data and the 3.6 cm and 1.3 cm continuum data to trace the diffuse and the confined ionized gas in the observed area .
the observations and data reductions are introduced in section [ chap_obs ] .
the observational results are presented in section [ chap_result ] .
the physical implications of the results are discussed in section [ chap_discussion ] . a brief summary and our plan of the follow up researches are given in section [ chap_summary ] and section [ chap_future ] .
we will follow up the studies of the kinematics in a future publication .
we use the coordinate system of epoch j2000 throughout the present paper .
[ [ iram-30m - telescope ] ] iram 30 m telescope + + + + + + + + + + + + + + + + + + we perform a 7@xmath10@xmath37@xmath10 otf scanning observation using the iram 30 m telescope mambo-2 array receiver , on 2011 january 20 .
the observation is carried out at a frequency of 250 ghz ( 1.2 mm ) , with a primary beam size of 10.5@xmath11 .
we made three maps ( map1 , 2 , 3 hereafter ) , with one hour on source time for each , centered at r.a.=18@xmath1210@[email protected] , decl.=-19@xmath1555@[email protected] , using the wobbler switching mode .
the wobbler throw is 120@xmath11 ; the scanning directions of these three maps have different position angles relative to the r.a .
axis , due to the rotation of the sky .
basic data reduction was carried out using the ` mopsic ` software .
the map1 is observed in a low sky noise condition , while map2 and map3 showed higher noise and stripes in the image before the sky noise subtraction was performed . without performing the sky noise subtractions50 % .
we use the sky noise subtracted maps as a reference for the distribution and the geometry of the faint structures .
however , we do not present the sky noise subtracted map to avoid the confusions in the brightness distributions . ] , we consistently obtain the peak flux of 9.5 jybeam@xmath16 for all three maps .
this value is comparable with the previous sma measurement of @xmath610 jy at 1.3 mm , in the central 10@xmath11 region ( liu et al .
2010 ) . a few percents of errors in estimating the flux distribution in the brightest central region will couple to the uncertainties in the modeling of the local zero flux level
. this can propagate significant ( @xmath6100 mjybeam@xmath16 ) repetitive artifacts in the scanning direction.we formed the final image based on map1 .
we used the measurements in map2 and map3 to amend the pixels in map1 which are corrupted by the repetitive artifacts .
however , because of the sky subtraction effects noted above , the replaced pixels may still bias the local brightness distribution in certain areas .
these effects , however , are limited to a level that does not affect the discussions in the present paper .
the rms noise level of map1 ( without sky noise subtraction ) is about 7 mjybeam@xmath16 .
[ [ sma ] ] sma + + + we performed the 1.3 mm band observations using the 6 meter dishes of the sma in the subcompact configuration , the compact configuration and the very extended configuration on 2009 february 09 , 2009 june 10 and 2009 july 12 , respectively . in the compact configuration and the very extended configuration observations ,
the observing frequencies were centered on 231 ghz ( 1.30 mm ) in the lower sideband , and centered on 241 ghz ( 1.24 mm ) in the upper sideband , respectively ; in the subcompact configuration observation , the observing frequencies of the two sidebands were centered on 221 ghz ( 1.36 mm ) and 231 ghz ( 1.30 mm ) .
these observations were carried out with 8 antennas .
the pointing center of these observations is r.a.=18@xmath1210@[email protected] , decl.=-19@xmath1555@[email protected] .
the primary beam size of these observations is 55@xmath11 .
the basic calibrations were carried out using the ` mir idl ` and the ` miriad ` software package ; the self - calibrations of these data were carried out using the ` aips ` package .
we constructed the continuum band visibility data at 1.3 mm from the line free channels .
the combined sma data sets cover a uv sampling range of 5410 k@xmath17 .
we detect @xmath614 jy of stokes
i emission in the field of view of the sma observations . the continuum image of this combined sma data set ( @xmath18@xmath3@xmath19 = [email protected]@xmath30@xmath1158 ) has been published in liu , ho & zhang ( 2010 ) .
we performed the follow
up 0.87 mm band observations using the sma on 2011 march 15 in the subcompact array configuration with 7 antennas , and on 2011 march 25 in the compact array configuration with 8 antennas .
the observing frequencies of the two sidebands were centered on 336 ghz ( 0.89 mm ) and 348 ghz ( 0.86 mm ) .
the pointing center of these observations are r.a.=18@xmath1210@[email protected] , decl.=-19@xmath1557@[email protected] ( p1 region hereafter ) , and r.a.=18@xmath1210@[email protected] , decl.=-19@xmath1557@[email protected] ( p2 region hereafter ) .
the observations were carried out in the snapshot mode , with 20 minutes on source integration per pointing in the subcompact configuration , and 16 minutes on source integration per pointing in the compact configuration ( after flagging the bad data ) . the theoretical rms noise level after combining the subcompact array and the compact array data is about 3.4 mjybeam@xmath16 .
we provide 2 versions of the continuum map .
in one version , we limit the uv sampling range to 845 k@xmath17 , to optimize the sensitivity and the shape of the synthesized beam ( [email protected]@[email protected] ) . without limiting the uv sampling range ( 8 - 80 k@xmath17 ) , we obtain a higher resolution 0.87 mm continuum image with [email protected]@[email protected] synthesized beam , to resolve the detailed structures in the brighter p1 region .
[ [ iram-30m - telescope - sma ] ] iram 30 m telescope + sma + + + + + + + + + + + + + + + + + + + + + + + + we approximate the sma observations at 1.2 mm using the sma data taken at 1.24 mm and at 1.30 mm .
the iram 30 m observations detect @xmath630 jy of stokes
i emission within the sma primary beam .
we convert the iram 30 m image into a uv visibility data set by using the ` miriad ` tasks ` demos ` , ` uvrandom ` and ` uvmodel ` .
we limit the uv sampling range of the iram 30 m visibility to be 04 k@xmath17 , to fill in the missing short spacing information in the sma data .
we note that in such a small uv sampling range , the visibility model of the iram data is minimally affected by the attenuation of the single dish primary beam .
we have inspected the data and confirm the consistent absolute flux levels of the sma data and the visibility model of the iram 30 m data .
we limit the uv sampling range of the sma data to be within 0220 k@xmath17 to optimize the sensitivity to the extended structures and the angular resolution . by applying ` invert ` and ( non box ) ` clean ` to the sma and the iram 30 m visibility data sets with 5000 iterations in ` miriad ` , we obtain a synthesized beam of [email protected]@[email protected] ( p.a.=15@xmath15 ) , and an rms noise level of 5 mjybeam@xmath16 .
the recovered flux by ` clean ` is 29.93 jy .
the final combined 1.2 mm image is formed by the ` miriad ` task ` restor ` .
we performed observations in the 0.87 mm band in polarization mode using the sma , in the compact north array on 2010 july 31 , and in the extended array on 2010
september 12 .
the observing frequencies of the two sidebands were centered on 336 ghz ( 0.89 mm ) and 348 ghz ( 0.86 mm ) .
the pointing center of both observations is r.a.=18@xmath1210@[email protected] , decl.=-19@xmath1555@[email protected] .
the primary beam size of these observations is 40@xmath11 .
these observations cover the uv sampling ranges of 10135 k@xmath17 and 20220 k@xmath17 , respectively .
the recovered stokes
i fluxes are 15 jy and 10 jy , respectively .
calibrations and imaging were carried out using the ` miriad ` software package .
we averaged the line free channels in the individual observations to generate the continuum channels , and jointly imaged the upper 4 ghz and the lower 4 ghz sideband .
for the compact north array observations , we obtain an rms noise level of 2.7 mjybeam@xmath16 , for the stokes q and the stokes u images . for the extended array observations
, we obtain an rms noise level of 3.0 mjybeam@xmath16 , for the stokes q and the stokes u images . jointly imaging the data from these observations
yield a synthesized beam of [email protected]@[email protected] , and the rms noise level of 2.3 mjybeam@xmath16 for the stokes q and the stokes u images . we retrieved the archived 3.6 cm continuum emission toward g10.6 - 0.4 in the vla a configuration including the vlba pie town antenna on 2005 january 2 ; and we observed the 3.6 cm continuum emission in the nrao ( expanded ) very large array ( vla / evla ) c configuration on 2009 july 27 .
the pointing center of these observations is r.a.=18@xmath1210@[email protected] , decl.=-19@xmath1555@[email protected] .
the primary beam size of these observations is 330@xmath11 .
the basic calibrations , self - calibration , and imaging of these data were carried out using the ` aips ` package .
we combined the a array+pie
town visibility data with the c - array visibility data , and applied a gaussian taper in the uv domain with a fwhm of 170 k@xmath17 , yielding a [email protected]@[email protected] synthesized beam with a position angle of -10.9@xmath15 .
the observed rms noise of the 3.6 cm continuum image is about 1 mjybeam@xmath16 ( @xmath65 k in terms of brightness temperature ) .
we note that the 3.6 cm continuum emission traces free free emission based on spectral index measurements .
we retrieved the archived vla 20 cm continuum data , taken on 1984 june 13 .
the primary beam size of this observation is 1833@xmath11 ( @xmath630.5@xmath10 ) .
the basic calibrations , self - calibration , and imaging of these data were carried out using the ` aips ` package , yielding an rms noise level of 1 mjybeam@xmath16 , with a 21@xmath11@xmath314@xmath11 synthesized beam ( see also ho , klein & haschick 1986 ) .
we retrieved the archived vla 1.3 cm continuum data , taken on 2002 february 1 .
the primary beam size of this observation is 2@xmath10 .
the basic calibrations , self - calibration , and imaging of these data were carried out using the ` aips ` package , yielding an rms noise level of 0.13 mjybeam@xmath16 , with a [email protected]@[email protected] synthesized beam ( see also sollins et al .
2005 ; sollins & ho 2005 ) .
we observed the cs ( 1 - 0 ) transition using the nrao ( expanded ) very large array ( vla / evla ) in the dnc configuration , on 2009 september 27 .
these observations had a continuous lst duration of 9 hours , with 20 available antennas after flagging .
the pointing center is r.a.=18@xmath1210@[email protected] , decl.=-19@xmath1555@[email protected] .
the primary beam size of these observations is 55@xmath11 .
the observations cover the uv sampling range of 4245 k@xmath17 .
continuum emission are averaged from the line - free channels and then subtracted from the line data .
calibrations , self
calibrations , and imaging were carried out using the ` aips ` package , yielding a synthesized beam of [email protected]@[email protected] , and an rms noise level of 22 mjybeam@xmath16 .
this cs dataset was previously published in liu , zhang & ho ( 2011 ) .
the observations discussed in this paper cover a tremendous amount of molecular structures , which have a broad range of linear size scales , and are embedded in different ambient environments ( e.g. different molecular volume density , geometry , morphology , gravitational potential , etc ) . to clarify the terminology ( e.g. filaments , clumps , cores , protrusive features / structures , envelope , etc ) , we present a schematic model in figure 1 , which also forecasts parts of the observational results .
this model presents a filamentary 10 pc scale , filamentary molecular cloud , which can be embedded in a giant molecular cloud .
the local concentrations of mass in the filaments fragment into smaller and denser substructures on a shorter dynamical timescale than the timescale of the global contraction . depending on the physical environments in the parent structures , the processes of fragmentation may also be hierarchical .
we note that the scale bars in the schematic model reflect the physical size scales of the structures that we will be referring to .
part of the schematic model , the massive molecular envelope region , has been introduced in a previous publication ( liu , zhang & ho 2011 ) .
[ chap_result ] figure [ fig_rgb ] shows the 1.2 mm continuum map from the iram 30 m observations . in this figure , we label the significant 24 @xmath0 m emission regions ( a o ) as the suspected ob star forming regions .
most of these ob star forming regions already show 20 cm or 6 cm free free continuum emissions ( ho , klein & haschick 1986 ) .
the resolved uc hii regions in our 3.6 cm observations are also labeled ( a e ; see also figure [ fig_2panel ] ) .
the uc hii region a located at the geometric center , which contains the ob cluster with 200 m@xmath1 , is resolved in the highest resolution vla observations at 1.3 cm ( sollins et al . 2005 ; sollins & ho 2005 ; see also section [ chap_context ] ) . from figure [ fig_rgb ]
we see at least 5 extended ( @xmath5 3 pc in the projected scale ) massive filaments in the 1.2 mm continuum map .
some of the extended structures may still be blended , and remain to be resolved with higher angular resolution .
the most significant feature in figure [ fig_rgb ] is the clear separation between the hii region in the northeast , and the neutral material in the southwest .
a @xmath2010 pc scale ionized arc traced by the diffuse 24 @xmath0 m emission is immediately followed by a slightly more extended 8 @xmath0 m bright pdr shell southwest of it .
most of the dense neutral material are distributed further southwest of the pdr shell , and is traced by the 1.2 mm continuum emission .
the neutral material is filamentary , and is extremely clumpy , with massive stars forming in localized dense structures .
multiple 24 @xmath0 m bright hii regions with significant pdr shells are seen over the @xmath2010 pc scale region ( e.g. massive cluster a , c , d , e , f , j , m , n ) .
uv photons from the massive cluster o appear to photoionize a filament in the south , and generate a bow shock shaped hii region ( see also section [ chap_bow ] ) .
the extended filament in the northwest seems to continue to the massive clusters e , f , and g. how this filament is dynamically associated with the densest structure in the center of this map is uncertain .
the rest of the extended filaments appear to connect to the geometrical center , an extremely bright 1.2 mm source with a projected scale of @xmath623 pc .
a high concentration of 8 @xmath0 m point sources is resolved in this central region , indicating the active fragmentation and the formation of a stellar cluster .
this region has high extinction , and there might be more embedded massive stars which are not visible in the optical and near infrared observations .
clusters of water masers ( hofner & churchwell 1996 ) and the high velocity @xmath7co ( 21 ) outflows ( liu , ho & zhang 2010 ) were also reported in this region .
we observe the 1.3 mm continuum emission using the sma , and observe the 3.6 cm free
free continuum emission and the cs ( 10 ) transitions using the vla / evla in the center of this dense structure ( section [ chap_context ] ) .
figure [ fig_contour ] shows the iram 30 m 1.2 mm continuum map with much more detailed contours , overlaid with the combined iram 30m+sma 1.2 mm continuum map in the central region . assuming @xmath21=2 , the gas
to dust ratio of 100 ( lis et al .
1998 ) , an average temperature of 3050k , and subtracting the free free continuum flux of 4 jy ( liu et al . 2010 ) , the detected 1.2 mm flux within the central 55@xmath11 sma primary beam corresponds to 1.84@xmath810@xmath22 m@xmath1 of molecular mass .
we note that from the cs observations , omodaka et al .
( 1992 ) estimated the mass to be 4@xmath810@xmath22 m@xmath1 in the central 30@xmath11 region .
we show the combined iram 30m+sma 1.2 mm continuum map , and the velocity integrated cs ( 10 ) maps in figure [ fig_2panel ] , to demonstrate the structures in the inner region with higher angular resolutions .
while figure [ fig_contour ] shows multiple filaments on the scale of a few parsecs , in figure [ fig_2panel ] , we see multiple protrusions in the inner parsec scale region .
these protrusions seem to have no preferred alignment , exhibiting instead a radial configuration connecting to a common center .
we visually identify the protrusions in the combined 1.2 mm map ( figure [ fig_2panel ] ) .
the identified structures are labeled by fl
ne , e , se , s , sw2 , sw1 , nw2 , nw1 , respectively . the elongated structure fl nw1 , fl e and fl sw1 are consistent with the more extended molecular filaments , which were previously detected in the nh@xmath9 ( ho & haschick 1986 ) and the cs ( 10 ) observations ( omodaka et al .
we note that the velocity gradient of fl
e was marginally resolved by ho & haschick ( 1986 ) , and was interpreted as the envelope rotational motion .
the fl s might have been detected in the previous c@xmath23o observation ( ho , terebey , & turner 1994 ) .
however , we emphasize that the 1.2 mm continuum emission presented in this paper traces the thermal dust emission , without being biased by the excitation conditions and the molecular abundances .
some of the protrusions continue to a large scale , which can only be followed in the iram 30 m 1.2 mm continuum map , for example , fl ne , fl
. there might be some foreshortened structures blended around the middle of the map ( figure [ fig_contour ] ) , which are only marginally resolved in the iram 30 m continuum map .
there is also an apparent filament or protrusion , which continues from the central parsec region to a large scale , and is associated with the massive cluster c. we suspect another filamentary structure that continues to the large scale , which is related to the formation of the massive cluster m and n. these suggestions can be tested with interferometric mosaic observations . comparing
the combined 1.2 mm continuum map with the velocity integrated cs ( 10 ) map ( figure [ fig_2panel ] ) suggests that some protrusions might continue directly into the central 0.05 pc , where the fast spinning hot toroid ( sollins & ho 2005 ; liu et al .
2010 ; beltrn 2011 ) is immediately around the most massive ( @xmath6200 m@xmath1 ) ob cluster .
we note the filamentary and clumpy nature of the massive molecular envelope traced by the cs ( 10 ) map has been highlighted in liu , zhang & ho ( 2011 ) . as an example of the continuation of the filaments , we zoom in on the fl
se region and show the combined 1.2 mm continuum map and cs ( 10 ) map in figure [ fig_filament ] . in the figure , we see that both the 1.2 mm emission and cs ( 10 ) emission show a number of elongated structures .
the most prominent one follow the suggested plane of rotation ( keto , ho & haschick 1988 ; sollins & ho 2005 ; liu et al .
figure [ fig_freefree ] demonstrates the relation between these elongated cs emission and the 1.3 cm free
free continuum emission from the centrally embedded uc hii region .
we note that the arc shaped features in the east of the 1.3 cm continuum map are explained as signatures of the externally ionized molecular clumps , which are approaching the embedded ob cluster ( sollins & ho 2005 ) .
this is the first time we resolve the geometry of the massive accretion flows in the distant ( few kpc ) ob cluster forming regions , on such a large scale and with such a high spatial dynamic range .
however , the resolved structures mentioned here are all projected onto the two dimensional images .
high resolution spectral line observations ( liu et al . 2010 ; liu , zhang & ho 2011 ) can be used to study the three dimensional geometrical context and the overall dynamics .
this will be the subject in our next paper .
the parsec scale filaments have complicated internal structures . as an example , figure [ fig_flrg ] shows the iram 30 m 1.2 mm continuum map in fl
e , overlaid with the saturated 24 @xmath0 m and 8 @xmath0 m ` spitzer ` images . from this figure , we see that both the 1.2 mm continuum emission and the near infrared opacity suggest a structure continuing from the southwest , which may still contain substructures .
more detailed structures are suggested by the near infrared opacity , although it is hard to distinguish if those structures are physically associated or a mere projection along the line of sight .
in such an evolved region , the distribution of the foreground and background extended infrared emissions severely confuse the structure identification .
the local structures can best be recognized from the millimeter and submillimeter thermal dust emission . however , the iram 1.2 mm continuum map has much poorer angular resolution as compared to the 24 @xmath0 m and 8 @xmath0 m ` spitzer ` images . here
, we define the millimeter clumps as compact structures identified in the iram 1.2 mm map , which have sizescales of @xmath60.5 pc .
we define the submillimeter cores as compact structures identified in the sma 0.87 mm stokes
i image , which have sizescales @xmath240.2 pc .
we identify the millimeter clumps as well as the submillimeter cores in regions where the fluxes are predominantly contributed by fl e and fl s , since structures in other regions , particularly in the east appear to be blended due to the still insufficient angular resolution .
in addition , the millimeter and the submillimeter continuum emissions in fl
e and fl s are minimally contaminated by the free free continuum emission , and are therefore good tracers of the molecular mass ( figure [ fig_contour ] , [ fig_ionized ] ; see also ho , klein & haschick 1986 for the free free continuum emission ) . note that these observationally defined features are still limited by the insufficient angular resolution and sensitivity of our observations , with likely unresolved substructures .
we perform 2dimensional gaussian fit using the ` aips ` task ` imfit ` to obtain the size .
we note that the size from the fitting is limited by the ( synthesized ) beam of the observations .
the gaussian fits of the fainter cores are biased by the confusion with the emission from the filaments .
the profiles of the brightness distribution of the identified structures may be non
gaussian due to the internal structures . hence ,
to obtain the total flux , we simply sum the flux within the ellipses as defined by the gaussian fits .
we provide the preliminary mass estimates based on the optically thin dust emission formulation in lis et al .
( 1998 ) , assuming an average temperature of 25k , the gas to dust ratio of 100 , and @xmath21=2 .
we approximate the volume of each of the clumps and cores by assuming a spherical geometry , with the radius being the average of their projected major and minor axes
. the estimated flux and mass of the clumps and cores can be biased to be factors of 23 higher , due to those uncertainties in the gaussian fits .
however , while estimating the density , these biases will be compensated by the similar bias in the volume . the coordinates ,
size , millimeter or submillimeter flux , mass , and the average density are summarized in tables [ table_mm ] and [ table_submm ] . with the similar assumptions , the overall molecular mass in fl
e and fl s , without including the structures in the central 3 pc dense region , are 6@xmath810@xmath25 m@xmath1 and 4.5 @xmath810@xmath25 m@xmath1 , respectively .
the identified structures are also labeled ( mm 19 , smm 16 , and smm 15 a , b , c ) in figures [ fig_contour ] , [ fig_ionized ] , and [ fig_condensations ] . the millimeter clumps in fl
e have an average projected separation of 0.82 pc ; and the millimeter clumps in fl
s have an average projected separation of 0.54 pc .
our follow up submillimeter stokes
i observations focus on the millimeter clumps mm1 , mm2 and mm4 .
we resolved 6 submillimeter substructures in the lower resolution versions of the 0.87 mm stokes
i image ( smm 16 , figure [ fig_ionized ] ) .
smm 1 and smm 2 have a projected separation of 0.27 pc ; and smm 3 , 4 , 5 have an average projected separation of 0.15 pc .
we note that the jeans length is 0.14 pc , assuming an average temperature of 25 k , and an average density of n@xmath26 = 10@xmath4 @xmath27 . without limiting the uv sampling ( section [ chap_0p87 ] ) , we made a higher angular resolution ( [email protected]@[email protected] ) 0.87 mm stokes
we find that the core smm1 is resolved into three independent submillimeter cores , while smm 2 , 3 , and 5 are also each resolved into two further submillimeter cores ( figure [ fig_condensations ] ) .
this implies smaller actual sizes and average separations of submillimeter cores in mm1 and mm2 .
we label the resolved substructures by smm 15 with suffix a , b , c. the submillimeter cores smm 1a , 1b , 1c , 2a and 2b seem to follow one elongated distribution ; the submillimeter cores smm 3a , 3b , 4 , 5a , 5b are aligned in another elongated distribution . with the still insufficient angular resolution of the sma 0.87 mm stokes i image ( section [ chap_0p87 ] ) , we may underestimate the volume density ( table [ table_submm ] ) of the submillimeter cores .
the submillimeter core smm 6 has an elongated shape ( figure [ fig_ionized ] ) , which may indicate further internal structures , but is too faint to be resolved .
we do not detect the stokes q and stokes u emissions at the 3@xmath28 level , in any combination of the compact north array and the extended array data .
we note that our polarized intensity maps have achieved comparable resolutions and lower rms noise levels as compared with the previous sma studies of distant ob star formation regions ( g31.41 + 0.31 : girart 2009 ; g5.89 - 0.39 : tang et al .
2009a ; w51 e2/e8 cores : tang et al .
our non detection suggests that g10.60.4 is less than 2.8% polarized .
possible explanations for the lower level of the polarized emission and the physical implications are discussed in section [ chap_mag ] .
the early vla 20 cm observations show an ionized bubble to the north of the central ob cluster .
this ionized bubble can be seen in the 24 @xmath0 m emission ( figure [ fig_rgb ] ) , and also with the 8 @xmath0 m bright pdr shell external to it .
the 24 @xmath0 m emission additionally shows some patchy structures on the eastern edge of the ionized bubble .
those patchy structures have high opacities , and are dark in the 24 @xmath0 m and 8 @xmath0 m maps .
the observed area has strong diffuse infrared emission .
the structures in the background infrared brightness distributions and the foreground infrared emission can confuse the identifications of structures . we report the patchy dust emission from the same area in the 1.2 mm band .
a blow up view of the iram 30 m 1.2 mm continuum map overlaid with the vla 20 cm continuum map is provided in figure [ fig_bubble ] .
the distribution of the 20 cm continuum emission suggests that the 1.2 mm emission from the patchy molecular gas are not contaminated by the free free emission .
this patchy molecular gas shows rich subparsec scale structures . with our current data , how these patchy structures form , and
if they are interacting with the ionized bubble or the hii region , are unclear .
we suspect that they are naked dense cores which lost their ambient gas through some interactions ; or they can be some foreground , isolated , subparsec scale molecular cloudlets .
these have to be distinguished in future observations .
from figure [ fig_rgb ] , we see the formation of the ob stars over the entire @xmath510 pc projected area .
the formation of these ob stars seems to be synchronized on a time interval , which is short in the sense that it takes @xmath2910@xmath30 years to propagate the sound wave in this @xmath610 pc cloud .
this propagation timescale is much longer than the collapsing timescale plus the life time of the massive stars .
the alfvn waves can propagate supersonically .
however , they will be restricted by the detailed magnetic field line configurations and the magnetic field strength . are there large scale physical phenomena , which are globally regulating or triggering the local structure formations ( e.g. the formation of the massive core , or the collapsing of the massive cores , etc ) ?
the impulse from the hii region in the northeast might be a candidate of this global mechanism , which was suggested by ho , klein & haschick ( 1986 ) with their vla 20 cm continuum observations .
this scenario has to be tested by the numerical hydrodynamical simulations . with the 1.2 mm dust continuum maps
, one may argue that the filamentary geometry of the neutral material could be a projection effect . in this section ,
we present the case of the ionizing pattern around the parsec scale filament in the south , and the massive cluster o ( figure [ fig_rgb ] ) , to argue in favor of the elongated geometry of the parsec scale structures .
while fragmentation in filamentary infrared dark clouds are ubiquitously seen , we argue that our result provides hints for their evolutionary context .
we can not rule out the possibility that there are edge on sheets which in projection appear like filaments .
however , it is questionable whether a @xmath205 pc scale sheet is likely at such a late evolutionary stage where clusters of smaller structures and massive stars are already present .
the idea is that if there is an ob cluster located close to the parsec scale massive filaments ( with a gradient of volume density along the radius ) , its ionizing photons and stellar wind will be blocked in the direction of the dense neutral material .
the uv photons emanating from the ob stars will only efficiently ionize the outer edges of the filament .
if the filament presents a small solid angle to the ob cluster , most of the uv photons , the stellar wind and the ionized gas will flow around the filament , and form a * u * shaped free
free continuum emission region .
such a * u * shape illumination pattern is seen around the massive filament in the south ( figure [ fig_ionized ] ) . if that filament were just filamentary in projection , but is an extended smoothed sheet in reality , we do not expect to see the ionized gas penetrating to the northeast of it ; instead , the free free emission should be limited to the southwest .
alternatively , the * u * shape illumination pattern can also be explained by the addition of two photoionized molecular filaments .
this scenario can be examined by deeper observations of thermal dust emission with higher angular resolution .
we note that the cloud with a filamentary morphology may be self shielded from an ob cluster deeply embedded in the geometrical center of the cloud .
while the parsec scale filament in the south is illuminated by the massive cluster o and shows significant 8 @xmath0 m emission ( figure [ fig_rgb ] ) , the filament in the east remains dark at 8 @xmath0 m . if there are smaller and denser filaments continuing to the centrally embedded ob cluster , the small cross sections of the filamentary accretion flow is minimally affected by the radiation from the luminous ob cluster , and the pressure force of the ionized gas . therefore , it is conducive for the accretion of the massive stars .
previous observations have suggested that the major , massive filaments can be formed from the merging of the smaller ones ( jimnez - serra et al .
the filaments may also collide with each other , which is conducive for the formation of cores and the massive stars ( galvn - madrid et al .
2010 ) .
at 10 pc scale , the filamentary morphology has a low volume filling factor , or a high porosity even after being projected onto a 2dimensional observing area .
the intensity ratios of molecular gas tracers ( e.g. hcn / co ) can then be easily converted to volume filling factors ( e.g. dense gas volume filling factor ) as in extragalactic studies . from figures
[ fig_contour ] , [ fig_ionized ] , and figure [ fig_condensations ] , we see concentration of dense gas in the form of massive molecular clumps and molecular cores .
the volume densities of these clumps and cores ( tables [ table_mm ] , [ table_submm ] ) are high enough to excite the dense gas tracers cs , hco@xmath31 , hcn , and h@xmath32co .
the parsec scale filaments , especially the two in the south and in the east , show rich marginally resolved 1.2 mm substructures , the millimeter clumps ( figure [ fig_contour ] ) .
the clumps in the eastern filament show the generally stronger 1.2 mm emission than those in the southern filament , which can be due to their difference in mass , temperature , or the dust properties .
this can be further examined by observing the nh@xmath9 emission with multiple transitions , and by observing the dust continuum emission in multiple frequency bands .
at least three massive clumps in the eastern filament may be forming massive stars , and show bright 8 @xmath0 m point sources around the region we label cluster l ( figure [ fig_rgb ] ) .
clumps in the southern and the eastern filaments seem to be regularly separated by 0.51 pc , implying a scale length for the local contractions which may be related to the length and width of the filaments . in the clumps mm1 and mm2 , our follow up sma observations have resolved much more submillimeter cores , which have an averaged projected separation comparable to the thermal jeans length ( section [ chap_mm ] , figure [ fig_condensations ] ) .
compared with other high resolution interferometric observations of filamentary infrared dark clouds , it is likely that hierarchical fragmentation is common . undoubtedly , there will be more blended substructures in our iram 30 m map due to the still insufficient angular resolution .
for example , in the case study of the infrared dark cloud g28.34 + 0.06 ( @xmath334.8 kpc ; wang et al .
2008 ; zhang et al . 2009
; wang et al . 2011 ) , the 1.3 mm continuum observation of sma resolved five cores within a single local maximum in the iram 30 m 1.2 mm continuum map .
we note that the submillimeter cores in the molecular clumps mm 1 and mm 2 ( figure [ fig_condensations ] ) are not randomly distributed , but are aligned linearly , forming the main axis of the filament . that the successive fragmentation at even smaller scales , continue to be aligned along the filaments ,
suggests a common process .
the process of fragmentation into clumps can be regulated by the density , mass concentration , the geometry or the morphology of the clump , the local magnetic field structure , the local feedback and the ambient pressure , the velocity field , interactions , and so on .
successive fragmentation takes place as the local conditions change .
furthermore , collision or merging of filaments may also be an important process .
we postpone the detailed statistical studies of these cores until we obtain follow up observations that achieve the physical resolution of @xmath340.1 pc , and achieve a much better sensitivity to recover the solar mass or smaller substructures .
some of the parsec scale filaments appear to converge into the geometric center of a 23 pc scale densest structure .
the abundance of protrusions in both the iram 30 m 1.2 mm continuum map , and the combined iram 30 m + sma 1.2 mm continuum map ( figure [ fig_contour ] ) support such a scenario .
we note however that the massive stars are distributed over the entire 10 pc area .
hence , it appears that the star formation process may be driven by local processes , and may not be related to inflow from large scales via the filaments .
we provide simple arguments based on the global and the local free fall timescale , to justify that the filamentary accretion flows with a radial alignment will allow efficient fragmentation during the quick global contraction .
theories on how contractions depend on morphology , geometry , and initial conditions , are described by ledoux ( 1951 ) , ostriker ( 1964 ) , larson ( 1985 ) , vishnaic ( 1994 ) , whitworth et al .
( 1994 ) , curry ( 2000 ) , myers ( 2009 ) , and myers ( 2011 ) .
our estimates specifically focus on the materials in the central 2 pc region ( i.e. inside the sma primary beam shown in figure [ fig_2panel ] ) , in which we see a concentration of bright infrared point sources ( figure [ fig_rgb ] ) .
we do not consider the magnetic field in our estimates , which is consistent with our observations that strong and organized magnetic fields may not be present at the sampled sizescales ( see also section [ chap_mag ] ) . with some rescaling , the same calculations can be applied to the more extended area while a lack of understanding of the large scale initial turbulence brings large uncertainties .
the free fall timescale is inversely proportional to the square root of the averaged density @xmath35 .
the value of @xmath35 can be estimated by @xmath36 while @xmath37 is the radius around a local center of mass , and @xmath38 is the total mass enclosed in the radius @xmath37 . for the case of g10.60.4 ,
in the central 2 pc region , we have resolved @xmath68 protrusions ( figure [ fig_2panel ] ) . assuming a filamentary geometry for those protrusions , of which the width is a fraction of a parsec , say 0.10.3 pc
, the volume occupied by each filament can be approximately estimated as the volume of a 2 pc rod , which is @xmath39@xmath40@xmath41@xmath42 @xmath43 0.0160.14 pc@xmath25 .
the volume filling factor in the inner 2 pc region can be approximately estimated by @xmath44 if we project all structures along the line of sight , we can also estimate the surface filling factor , which is @xmath550% .
we note that the smaller beam filling factor of 30% reported in the previous low resolution nh@xmath9 observations ( keto , ho , & haschick ) , suggests that most of the mass are concentrated to filaments or in the central massive core .
the estimated small volume filling factor implies potentially a much higher local volume density than the averaged volume density in that region . within the region enclosed by the 1 pc radius ,
the free fall contraction timescale of the local perturbations can be a factor of 1.95.8 times shorter than the global free fall contraction timescale .
depending on the support of the rotational velocities , the timescale of the large scale contraction could be a few times longer than the global free fall contraction timescale ( ho & haschick 1986 ; keto , ho & haschick 1987 , 1988 ; keto 1990 ; liu et al .
2010 ; liu , zhang & ho 2011 ) .
thus we suggest that although the system may have a quick global contraction ( see also the _ clump fed _ scenario in wang et al . 2010 ) , the existence of small scale structures implies that efficient fragmentation and subsequent star formation may have already occurred .
the accretion flows are clearly highly clumpy , which have been resolved in the high resolution nh@xmath9 absorption experiment ( sollins & ho 2005 ) , and the pv diagram of @xmath45cs ( 54 ) ( liu et al .
the fragmentation of the accretion flows is consistent with the observed clusters of uc hii regions , multiple water masers and high velocity @xmath7co outflow sources ( figure [ fig_2panel ] ) , many dusty cores with strong 1.3 mm emission , all external to the centrally located , compact , fast spinning , hot toroid ( liu , ho & zhang 2010 ; liu , zhang & ho 2011 ) .
the earliest epoch of star formation will generate turbulence , which will help regulate the subsequent star formation efficiency and the accretion of the centrally embedded massive star ( g10.60.4 : liu , ho & zhang 2010 ; simulations : li & nakamura 2006 ; nakamura & li 2007 ; carroll et al .
2009 ; wang et al .
2010 ) . from the three color image ( figure [ fig_rgb ] )
, we see a higher concentration of the 8 @xmath0 m sources in the central parsec region than in the extended region , which may be explained by the faster fragmentation and contraction of the filamentary structures in the higher density environment .
we note that the interferometric observations of the nearby massive cluster forming region , the orion
kl central core , have also demonstrated the highly filamentary nature of the molecular gas , as well as the fragmentation in those filaments , in the inner @xmath61 pc region ( wiseman & ho 1996 ; wiseman & ho 1998 ) .
note that on such a small scale , the filaments do not need to be primordial nor a continuation of the outer structures .
the interactions between the ambient gas with the ( proto)stellar feedback can also lead to the formation of filaments in the later evolutionary stages . of course
, the filaments can also be destroyed by the same feedback mechanisms ( wang et al .
the process of infall and fragmentation is highly dynamical and chaotic .
the _ red excess _ of the velocity field in g10.6 - 0.4 was previously reported in the nh@xmath9 absorption line experiments , which is consistent with a high velocity infall ( ho & haschick 1986 ; keto , ho & haschick 1987 , 1988 ; keto 1990 ; sollins & ho 2005 ) . from 1@xmath11 resolution emission line observations ,
liu et al .
( 2010 ) confirms the red excess in the central @xmath60.06 pc ( 2@xmath11 ) region , and further sees the high spatial asymmetry and the clumpiness in the mass distribution , from the position
velocity ( pv ) diagram .
motivated by the resolved protrusions in the present work ( figure [ fig_2panel ] ) , we propose that the spin up rotational motions of the accretion flow with a filamentary morphology may coherently explain the red excess , the spatial asymmetry and the clumpiness in those pv diagrams .
the detailed modeling will be provided in our follow up kinematics studies .
sma observations of luminous massive star forming regions ( g31.41 + 0.31 : girart 2009 ; g5.89 - 0.39 : tang et al . 2009a ;
w51 e2/e8 cores : tang et al .
2009b ) have shown that the polarization percentage of the 0.87 mm thermal dust emission , range from a few percent ( e.g. 4% in g31.41 + 0.31 ) , to up to @xmath68% in the w51 e2/e8 regions , and up to @xmath622% in the g5.89 - 0.39 region .
the polarization observations of g10.6 - 0.4 achieved an improved sensitivity over the previous sma studies of similar targets .
while those previous cases all show significant detections of polarized flux , in the present work , we do not detect the polarized emission above the 3@xmath28 rms noise level . in g10.6 - 0.4 ,
the polarization percentage of the dust emission at 0.87 mm is constrained to be less than 2.8% .
such a polarization percentage is lower than the reported polarization percentages in those previous cases .
if g10.6 - 0.4 is a strongly magnetized case , the low polarized flux can be interpreted by some differences in the dust properties from the other targets , which produce less polarized flux , the relatively inefficient dust alignment in g10.60.4 , and/or the canceling of the polarized emission from structures overlapped along the line of sight .
the feedback of the high velocity molecular outflows and the expansions of the hii regions may , but not necessarily disturb the magnetic field and lead to the non detection of the polarized flux . in the case of g5.89 - 0.39 , the embedded uc hii region has already undergone an expansion phase , and multiple molecular outflows have been detected , and the magnetic field structure is strongly affected ( tang et al .
. the magnetic field might also be more important on a much smaller scale , which is unresolved in our sma observations .
the radial morphology of the filamentary structures in the central core , suggests contraction by gravity to be dominant over all possible sources of support .
this needs to be confirmed via modeling of the kinematics .
the absence of detectable magnetic field structures is consistent with this scenario .
comparing our high resolution 1.2 mm continuum maps and the cs ( 10 ) observations , suggests that the massive molecular clump containing a luminous ( @xmath2910@xmath4 l@xmath1 ) ob cluster is part of a @xmath510 pc scale organized structure .
the extended molecular gas has a filamentary geometry , starting from a @xmath65 pc extent in radius , to a geometrical center of a 23 pc scale dense structure .
the parsec scale filaments show regularly spaced molecular clumps , each containing massive molecular cores that are associated with the local massive star formations .
part of these parsec scale filaments might directly continue into the central 0.05 pc radius hot toroid , which encircles the highest mass stellar cluster . in this central region
, the radial filamentary configuration suggests accretion flows which would allow simultaneously a quick global contraction and the efficient local fragmentation .
such a model would be consistent with the previously seen centrally located 0.1 pc scale fast spinning hot toroid , and the rich cluster of water masers , millimeter cores , and high velocity @xmath7co ( 21 ) outflows .
we propose that the spin up rotational motions of a filamentary accretion flow can explain the red / blue excess of the velocity field , and the spatial asymmetry of the mass distributions , although the interpretation is not unique . in this scenario , the magnetic field may not , and does not necessarily play an important role on the scale that we are observing ( 0.055 pc radius ) .
the large numerical hydrodynamics simulations to compare the strongly and the weakly magnetized cases will help to improve the physical insights , and also improve the interpretation of the data . the proposed scenario in this paper
might not be the unique mode of ob cluster formation .
survey observations of a larger sample will be useful to identify or to classify other modes of ob cluster formation .
we emphasize that the subject under the consideration in the present paper is specifically the luminous ( @xmath2910@xmath4 l@xmath1 ) ob cluster .
in the present work , we suggest the convergence of the massive filamentary structures is one very important aspect in ob cluster formation
. however , how such filaments of a few parsec scale form , is still uncertain , and should be a very fundamental aspect in the study of the interstellar medium .
the very sensitive high resolution and high dynamic range spectral line and dust polarization observations might help to address this question .
another aspect which can not be addressed by our present research is how the 0.1 pc scale hot toroid converge to form the embedded ob clusters .
high resolution molecular and recombination line observations with alma will certainly address this issue .
we thank the sma staff for making these observations possible .
we also thank the iram staff for help with the observations and for developing the data reduction routines .
liu thanks asiaa for supporting this research .
this work is based in part on observations made with the _
spitzer space telescope _ , which is operated by the jet propulsion laboratory , california institute of technology .
we thank the glimpse team ( pi : e. churchwell ) and mipsgal team ( pi : s. carey ) for making the irac and mips images available to the community . k.
w. acknowledges the support from the sma predoctoral fellowship and the china scholarship council . and r@xmath46 .
the purple arrows indicate the parsec scale protrusions ( fl ne , e , se ,
s , sw1 , sw2 , nw1 , nw2 ) .
the blue contours are 16 mjybeam@xmath16 @xmath3 [ -2 , -1 , 1 , 2 , 3 , 4 , 5 , 6 , 7 , 8 , 9 , 10 , 12 , 14 , 16 , 20 , 60 ] ( 1@xmath28= 5 mjybeam@xmath16 ) ; the orange contours are 3 mjybeam@xmath16 @xmath3 [ 1 , 2 , 4 , 8 , 16 , 32 , 64 ] ( 1@xmath28= 0.3 mjybeam@xmath16 ) ; the magenta contours are 40 mjybeam@xmath16 @xmath3 [ 1 , 2 , 3 , 4 , 5 , 6 , 7 , 8 ] ( 1@xmath28= 7 mjybeam@xmath16 ) .
the negative contours are shown in dotted lines .
the cs distributed in projection against the bright uc hii regions show negative flux due to strong absorption .
we avoid integrating the negative flux while generating the cs map to emphasize the structures in emission . ]
@xmath3 [ 1 , 2 , 3 , 4 , 5 , 6 , 7 , 8 , 9 , 10 , 12 , 14 , 16 , 20 , 60 ] ( 1@xmath28= 5 mjybeam@xmath16 ) .
the green dashed line show the pv cut as well as a proposed plane of rotation in liu et al .
2010 , liu , zhang & ho ( 2011 ) .
the dashed line is 25@xmath11 ( 0.75 pc ) in length , and the top right end of this line corresponds to the zero angular offset .
we note a continuation of an elongated cs structure to the 1 mm protrusion along the dashed line .
we also note two more elongated cs structures in the north of the previously mentioned one . ] | we observe the 1.2 mm continuum emission around the ob cluster forming region g10.6 - 0.4 , using the iram 30 m telescope mambo-2 bolometer array and the submillimeter array .
comparison of the spitzer 24 @xmath0 m and 8 @xmath0 m images with our 1.2 mm continuum maps reveals the ionization front of an hii region , the photon dominated layer , and several 5 pc scale filaments following the outer edge of the photon dominated layer .
the filaments , which are resolved in the mambo-2 observations , show regularly spaced parsec
scale molecular clumps , embedded with a cluster of submillimeter molecular cores as shown in the sma 0.87 mm observations . toward the center of the g10.6 - 0.4 region , the combined sma+iram 30 m continuum image reveals several , parsec scale protrusions .
they may continue down to within 0.1 pc of the geometric center of a dense 3 pc size structure , where a 200 m@xmath1 ob cluster resides .
the observed filaments may facilitate mass accretion onto the central cluster forming region in the presence of strong radiative and mechanical stellar feedbacks .
their filamentary geometry may also facilitate fragmentation .
we did not detect any significant polarized emission at 0.87 mm in the inner 1 pc region with the sma . |
the recent discovery of a cloud moving towards the galactic center @xcite creates a potential opportunity for testing models of sgr a*. the @xmath7 cloud will interact strongly with gas near nominal pericenter at @xmath8 ( @xmath9 black hole mass ) , and may change the black hole accretion rate @xmath0 . since the structure of the cloud and the surrounding medium are uncertain , possible outcomes range from very small changes in the accretion rate over timescales of decades to rapid , large changes in the accretion rate .
the dynamical timescale at @xmath10 is @xmath11yr , and the viscous timescale @xmath12yr @xmath13 , assuming @xmath14 and @xmath15 , i.e. a hot , radiatively inefficient accretion flow .
after an initial transient phase while the flow circularizes accompanied by transient emission
it is natural to think the flow will settle into a steady state .
the settling timescale could be as little as a few @xmath16 , and so the steady state may arrive as soon as mid-2014 . if the resulting flow can be modeled as a steady disk , the excess mass will drain away on the viscous timescale , i.e. the source will remain bright well into the 22nd century .
it is therefore interesting to ask how changes in @xmath0 will manifest themselves observationally .
current observations of sgr a * show @xmath17 jy at 1 - 50 ghz with a nearly flat spectral slope @xcite ; @xmath18 .
the spectral slope becomes flatter and variable at 230 - 690 ghz , with @xmath19 , @xcite ) , which is commonly interpreted as signaling a transition from optically thick to optically thin synchrotron emission .
the discovery of polarized emission ( polarization fraction at level of a few per cent ) at @xmath20 mm and subsequent measurement of faraday rotation imply a model dependent limit @xmath21 in the inner accretion flow ( @xcite , @xcite ) .
sgr a * is resolved by vlbi at @xmath22 mm where interstellar scattering is comparable to the intrinsic source fwhm @xmath23 @xcite .
this is smaller than the apparent diameter of the event horizon @xmath24 .
sgr a * fluctuates rapidly in the near - infrared ( nir ) , with average @xmath25 ( @xcite ) .
it is not yet detected at mid - infrared ( mir ) wavelengths ( @xmath26 , e.g. @xcite ) . in the x - ray sgr
a * exhibits flares with characteristic duration of order an hour and a duty cycle of order @xmath27 .
the upper limit for quiescent state x - ray emission is @xmath28 @xcite .
we are not aware of any secular trends in these observed properties of sgr a*. one model of sgr a * that fits most observational constraints is our relativistic accretion model @xcite , where submillimeter , ir , and x - ray emission arise in an optically thin , geometrically thick disk close to the event horizon ; radio emission is assumed to arise nonthermally in a synchrotron photosphere at larger radius @xcite , but is not predicted by our model , which focuses on the inner parts of the accretion flow .
the underlying flow model is a general relativistic magnetohydrodynamic ( grmhd ) simulation ( @xcite , @xcite , @xcite ) .
the emerging radiation is calculated using monte carlo @xcite and ray - tracing schemes .
the @xmath22 mm flux originates as doppler boosted synchrotron emission from the approaching side of the disk between @xmath29 and the innermost stable circular orbit ( isco ) .
our relativistic disk model assumes a thermal electron distribution function , neglects thermal conduction , and assumes a constant ratio of ion to electron temperature @xmath30 .
the model does not produce the observed ir flaring at the correct amplitude , but can with the addition of a small nonthermal component in the distribution function @xcite .
if we fix the black hole mass ( @xmath31 , @xcite ) and distance ( @xmath32 , @xcite ) , the remaining model parameters are the source inclination @xmath33 , black hole spin @xmath34 ( @xmath35 ) , @xmath30 , and @xmath0 .
we fix @xmath0 so that the @xmath22 mm flux matches the observed @xmath36 jy .
the relativistic accretion models are not tightly constrained by the data , but they reveal the following : ( 1 ) face - on models that reproduce the millimeter flux would look like rings and therefore , in vlbi data , have dips in visibilities on fixed intermediate baselines , while existing observations suggest that the ring radius would need to vary to fit the data ( @xcite ; see also the discussion of @xcite ) .
more nearly edge - on models are therefore favored ; ( 2 ) models with @xmath37 that reproduce the millimeter flux have a hot , dense inner disk that would overproduce x - rays via inverse compton scattering .
lower spin , @xmath38 , models are therefore favored ; ( 3 ) the observed source size and flux fix the temperature of the emitting electrons @xmath39 ( @xmath40 the rms size of the source on the sky ) and this turns out to favor @xmath41 models .
the `` best - bet '' model from @xcite has @xmath42 , @xmath43deg , @xmath44 , and @xmath45 .
the relativistic disk model uses self - consistent dynamics and radiative transfer but is not unique .
the electron distribution function is particularly poorly constrained .
it is likely anisotropic , may contain multiple temperature components ( @xcite ) and power - law components , and vary in basic functional form with time and position .
alternative accretion models ( e.g. @xcite , @xcite ) make different assumptions about the flow and/or distribution function and favor slightly different @xmath46 and @xmath33 .
these models may respond differently to an increase in @xmath0 .
indeed , radically different models may also fit the data .
the model dynamics depends on the initial magnetic field distribution , particularly the distribution of vertical flux through the accretion disk ; models with large vertical magnetic flux ( e.g. @xcite ) are likely to respond quite differently to variations in the mass flux . also , jet models for sgr a * ( @xcite , @xcite , @xcite ) posit a luminous jet and a comparatively dim accretion disk . again
, these may respond differently to an influx of mass .
how , then , does our relativistic disk model respond to an increase in @xmath0 ? in this _ letter _
we calculate the @xmath22 and @xmath47 mm flux and source size as well as the spectrum that would result for the best - bet model over a range of @xmath0 .
one key question we seek to answer is whether a small increase in @xmath0 would hide the event horizon ( and the signature ring - like appearance of the photon orbit , also known as the shadow or silhouette of the event horizon ) underneath a synchrotron photosphere .
this might prevent detection of the photon orbit by the planned event horizon telescope @xcite .
another key question is whether the increased @xmath0 would make sgr a * detectable in its quiescent state in the ir and x - ray .
below we describe variation of the flux and source morphology at @xmath3ghz ( @xmath22 mm ) and @xmath5ghz ( @xmath47 mm ) ( [ results ] ) , describe variation of the spectrum ( [ spectra ] ) , and finally discuss which features of the results are likely to be most robust ( [ discussion ] ) .
how do we naively expect the mm disk image size and flux to respond to changes in @xmath0 ? in our model ,
@xmath22 mm emission in sgr a * is thermal synchrotron emission from plasma with optical depth @xmath48 , near the isco .
the true electron distribution undoubtedly contains nonthermal components ( see , riquelme et al .
models with thermal + power - law distribution functions ( e.g. @xcite ) contain an o(1/3 ) nonthermal contribution to the flux at @xmath22 mm , which hints at how uncertainty in the distribution function translates into uncertainties in the spectrum .
the thermal synchrotron absorptivity is @xmath49 , where @xmath50 , @xmath51 , @xmath52 , @xmath53 is an angle between the magnetic field vector and emitted photon , @xmath54 is a modified bessel function of the second kind @xcite and @xmath55 . near @xmath22 mm , @xmath56 , and
the emissivity is nearly independent of frequency , so @xmath57 and @xmath58 .
we will assume that @xmath59 , @xmath60 , and @xmath61 const .
, so that @xmath62 for @xmath63 , and ignore relativistic corrections
. then for @xmath64 ( or @xmath65 ) the source has size @xmath66 and the flux @xmath67 . for @xmath68 the source size is set by the photosphere radius @xmath69 where @xmath70 ( i.e. for @xmath71 , but not so large that @xmath72 so that our emissivity approximation fails ) .
then @xmath73 , and the flux @xmath74 . these simple scaling laws , unfortunately , are not a good description of the variation of source size and flux with @xmath0 . there are at least three reasons for this .
first , relativistic effects are very important ; for @xmath75 emission comes from close to the photon orbit and the source size is determined by doppler beaming and gravitational lensing .
second , for the best - bet model at @xmath22 mm @xmath76 , so in a turbulent flow there is a complicated variation of the size of the effective photosphere with @xmath0 .
third , the emissivity is not precisely frequency independent near peak .
we therefore need to turn to numerical models .
the best - best model is taken from a survey of 2d models . here
we adopt the best - bet model parameters ( @xmath77 , @xmath78 and @xmath79 ) and use them to set parameters for a fully 3d model @xcite .
we use the same data set as @xcite , and choose three representative snapshots taken at times when the flow is quiescent ( @xmath80 and @xmath81 , where @xmath82s ) .
we then recalculate disk images and spectra for @xmath83 .
the 3d model with @xmath84 is broadly consistent with observational data but is slightly more luminous at higher energies than 2d models ( @xmath85 is lower in the 3d models , and this changes the x - ray to millimeter color ) .
the model is self - consistent only for @xmath86 . at higher @xmath0
the efficiency of the flow is @xmath87 and therefore our neglect of cooling in the underlying grmhd model is not justified . at higher @xmath0
the @xmath22 mm photosphere also lies outside the limited range in radius where @xmath88 , so the flow is not in a steady state .
the images and total fluxes emitted by the disk at @xmath3 and @xmath5ghz are calculated using a ray tracing scheme @xcite .
to estimate the size of the emitting region we calculate the eigenvalues of the matrix formed by taking the second angular moments of the image on the sky ( the principal axis lengths ) .
the major and minor axis eigenvalues , @xmath89 and @xmath90 respectively , are related by @xmath91 to the fwhm of the axisymmetric gaussian model used to interpret the vlbi observations .
we use @xmath92 to measure the average radius of the emitting spot .
figures [ fig:1 ] and [ fig:2 ] show the variation of @xmath3 and @xmath5 ghz sgr a * model images with @xmath0 ( based on a single snapshot from the 3d grmhd simulation ) .
evidently modest increases @xmath0 will make the ring - like feature that is the observational signature of the photon orbit easier to detect . for @xmath2 , however , the ring ( or black hole silhouette ) is hidden beneath the synchrotron photosphere at @xmath3 ghz .
the silhouette survives for higher @xmath0 at @xmath5 ghz , disappearing only at @xmath93 . for low @xmath0
the silhouette is also difficult to detect because the emitting region is too small .
+ + + figure [ fig:3 ] , shows the accretion flow image size ( @xmath94 , circle radii in figure [ fig:1 ] and [ fig:2 ] ) and flux at @xmath3 and @xmath5ghz as a function of @xmath0 .
different point types correspond to three different snapshots from the 3d grmhd simulation .
the size of the emission region and flux increase with @xmath0 .
the 345/230 ghz flux ratio increases with the increasing @xmath0 ; this is caused by the shift of the synchrotron peak towards higher energies at higher @xmath0 .
the following simple fitting formulas describe how @xmath95 and @xmath96 depend on @xmath0 : @xmath97 } \label{fitsig230}\ ] ] @xmath98 } \label{fitsig345}\ ] ] @xmath99 } \label{fitf230}\ ] ] @xmath100 } \label{fitf345}\ ] ] the above constants are nontrivial to interpret because they encapsulate the complexities of the accretion flow structure and relativistic effects in the radiation transport .
the advantage of the above formulas is their simplicity .
the fitting functions are shown in figure [ fig:3 ] as dashed and dotted lines .
figure [ fig:4 ] shows the relation between two observables , @xmath95 and @xmath96 . the size is a linear function of the flux and increases more steeply at 230 ghz than at 345 ghz .
we also provide two phenomenological scaling laws fitted to the data : @xmath101 } \label{form1}\ ] ] @xmath102 } \label{form2}\ ] ] notice that these fits apply to the best - bet model only . for other @xmath34 , @xmath30 , or
@xmath33 the scalings will be slightly different , although if the source is optically thick the linear scaling follows directly from @xmath103 .
notice that equation [ form1 ] predicts the change of sgr a * size by 9 per cent when 230 ghz flux changes from 2 to 2.7 jy .
taking into account observational and theoretical uncertainties this is consistent with the observed variations of the size of the source which increases by a few per cent as the flux increases from 2 to 2.7 jy @xcite .
our model spectra are generated by thermal synchrotron emission in the submillimeter / far - ir bump , and by compton scattering in the x - rays .
the spectral slope of flaring nir emission , and its high degree of linear polarization @xcite , imply that it is synchrotron from a small , nonthermal tail of high energy electrons that is not ( but can be ; see @xcite ) included in our best - bet model .
what are the expected scaling laws ?
again , @xmath104 .
the luminosity around the synchrotron peak is @xmath105 , where @xmath106 is such that @xmath107 .
the emission rightward of the mir / nir is produced by compton up - scattered synchrotron radiation . the thomson depth @xmath108 , the x - ray luminosity is expected to scale as : @xmath109 , assuming that x - rays are produced primarily by singly scattered synchrotron photons . what do the numerical models show ?
figure [ fig:5 ] shows spectra emitted from the 3d disk model as observed at @xmath110 .
the seds are calculated using a general relativistic monte carlo scheme @xcite .
the nir luminosity @xmath111 , which is only slightly steeper than the expected dependence for the synchrotron peak .
@xmath112 agrees well with the expected scaling .
we conclude that sgr a * would become a persistent mir and x - ray source ( above the present upper limits of @xmath113 in mir and @xmath114 in x - rays ) if @xmath6 .
this is conservative in the sense that our models are strictly thermal .
the addition of a high energy nonthermal tail would only increase the mir / nir and x - ray flux .
we do not consider higher accretion rate models because for @xmath115 ( @xmath116 ) the model becomes radiatively efficient , @xmath117 , and our neglect of radiative cooling in the underlying 3d grmhd model is unjustified
. finally , notice that the mev flux increases sharply with @xmath0 .
this suggests that electron - positron pair production by photon - photon collisions in the funnel over the poles of the hole would increase sharply ( the pair production @xmath118 , @xcite ) .
if this pair production is connected to jet production , it is reasonable to think that a high @xmath0 sgr a * might also produce a jet .
in summary , we have used general relativistic disk models and relativistic radiative transfer to recalculate millimeter images and spectra for sgr a * at a range of @xmath119 .
our models predict the following : ( 1 ) if the 230 ghz flux increases by more than a factor of 2 , corresponding to an increase of @xmath0 by more than a factor of 2 , the central accretion flow will become a persistent , detectable , mir and x - ray source ; ( 2 ) the photon orbit , which produces the narrow ring of emission visible in figures 1 and 2 , becomes easier to detect for modest increases in the 230 ghz flux ; ( 3 ) the photon orbit is cloaked beneath the synchrotron photosphere at @xmath3 ghz for @xmath120 , or 230 ghz flux @xmath121 jy ; ( 4 ) the photon orbit is cloaked at @xmath5 ghz only at higher @xmath93 , or 230 ghz flux @xmath122 jy ; ( 5 ) the size of the source increases in proportion to the flux at both @xmath3 and @xmath5 ghz .
we suspect that almost any accretion model for sgr a * with a spatially uniform model for the plasma distribution function will reach qualitatively similar conclusions , but that jet models may differ significantly .
there are order - unity uncertainties in our model predictions due to uncertainties in the plasma model .
what range of changes in @xmath0 are reasonable ? in our best - bet model the mass at radii within a factor of two of the pericenter radius is @xmath123 @xmath124 , assuming steady mass inflow from @xmath10 to the event horizon .
the addition of even a fraction of the inferred cloud mass to the accretion flow in a ring near @xmath10 would ( eventually ) increase @xmath0 by a factor of @xmath125 . on the other hand , stellar winds supply mass in the neighborhood of the central black hole at @xmath126 .
models by @xcite and @xcite suggest that most of this mass is ejected in the form of a wind , and that @xmath127 to @xmath128 flows inward .
a reasonable extrapolation of these models suggest the accretion flow at @xmath129 has a mass of @xmath130 ; this is comparable to estimates of the mass of the inflowing cloud , so in this case we might expect a factor of @xmath131 increase in @xmath0 . | in accretion - based models for sgr a * the x - ray , infrared , and millimeter emission arise in a hot , geometrically thick accretion flow close to the black hole .
the spectrum and size of the source depend on the black hole mass accretion rate @xmath0 . since gillessen et al .
have recently discovered a cloud moving toward sgr a * that will arrive in summer 2013 , @xmath0 may increase from its present value @xmath1 .
we therefore reconsider the `` best - bet '' accretion model of moscibrodzka et al . , which is based on a general relativistic mhd flow model and fully relativistic radiative transfer , for a range of @xmath0 .
we find that for modest increases in @xmath0 the characteristic ring of emission due to the photon orbit becomes brighter , more extended , and easier to detect by the planned event horizon telescope submm vlbi experiment . if @xmath2 this `` silhouette '' of the black hole will be hidden beneath the synchrotron photosphere at @xmath3ghz , and for @xmath4 the silhouette is hidden at @xmath5ghz .
we also find that for @xmath6 the near - horizon accretion flow becomes a persistent x - ray and mid - infrared source , and in the near - infrared sgr a * will acquire a persistent component that is brighter than currently observed flares . |
this work is supported in part by fundao de amparo pesquisa do estado de so paulo ( fapesp ) and the conselho nacional de desenvolvimento cientfico e tecnolgico ( cnpq ) .
m.s.h is the martin gutzwiller fellow , 2007/2008 . | we derive closed expressions for the nuclear breakup cross sections in the adiabatic limit using the austern - blair theory .
these expressions are appropriate for the breakup of weakly bound nuclei .
the concept of an exit doorway that mediates the coupling between the entrance channel and the breakup continuum is used .
we prove the validity of the scaling law that dictates that the nuclear breakup cross section scales linearly with the radius of the target .
we also compare our results for the nuclear breakup cross section of @xmath0be , @xmath1b on several targets with recent cdcc calculation .
the breakup of nuclei is a common occurance when the bombarding energy is high enough and/or the binding energies are sufficiently low . in the case of weakly bound nuclei the threshold for breakup is small and more so for bound unstable nuclei .
the mechanism of breakup is assumed to consist of elongating the projectile , through the action of the interaction , which eventually leads to the production of two or more fragments .
this interaction is composed of a short range , nuclear piece and a longer ranged electromagnetic one .
a debate has been going on in the literature concerning the way the nuclear part of the breakup cross section depends on the mass of the target nucleus which supplies the interaction . in most references @xcite , it is assumed that the dependence goes as the cubic root of the mass number . in reference
@xcite , however , it is claimed that this dependence is more like linear ! in a recent paper @xcite , through a careful continuum discretized coupled channels ( cdcc ) calculation , the former dependence ( a@xmath2 ) has been established , which corroborates the contention that the nuclear breakup cross section should follow the prediction of the serber model @xcite .
it is interesting to compare the numerical cdcc calculation alluded to above with those of simpler analytical models .
specifically , the ausern - blair adiabatic theory for inelastic scattering comes to mind .
if one assumes that the breakup proceeds through a so - called exit doorway @xcite , then the process can be treated as an inelastic excitation .
the idea of exit doorway has been used in the case of the influence of breakup on fusion @xcite and in the excitation of giant resonances @xcite with success . in a very recent paper
@xcite a comparison of a preliminary cdcc calculation for the breakup cross section of the system @xmath3he@xmath4al at low bombarding energies with a simple formula derived by us]using the austern - blair model , showed that such an idea is quite reasonable and encouraged us to pursue the matter further . we do this in the present paper , where we fully develop the austern - blair model for the nuclear elastic breakup reaction cross section assumed to proceed through the excitation of an exit doorway @xcite .
the exit doorway concept has been used in the devolopment of reaction theories invloving the the excitation of a `` doorway '' in the final state , in contrast to the conventional cases where such resonances are populated in the entrance channel @xcite . in the breakup reactions of halo nuclei
one may envisage that the process proceeds through the breakup doorway dipole , quadrupole etc . ) into the continuum .
as such , the detailed description of the exclusive reaction , where the final channels are especified , will necessarily contain the full information about the exit doorway ( its energy , width etc . ) .
this is the case that was encountered in the theory of the excitation of multiple giant resonances @xcite and of the influence of the pygmy resonance on the fusion of halo nuclei @xcite . in the current paper
we will be content with the inclusive quantity of the integrated breakup cross section and the only reference to the exit doorway is made implicitly as a final state that has to be populated for breakup to occur .
the full hamiltonian which describes the colliding ions can be written as @xmath5 where @xmath6 is diagonal in open channel space , @xmath7 is the intrinsic part that describes the structure of the projectile and the target nuclei , @xmath8 is the kinetic energy operator and @xmath9 is the optical potential which contains the complex nuclear plus the coulomb parts .
the operator @xmath10 describes the coupling among the open channels .
the intrinsic hamiltonian @xmath7 , which for simplicity is taken here to represent the excitable projectile nucleus with the target considered structureless , is now written as : @xmath11 + \nonumber \\
\sum_{ij}[|i>\omega_{ij}<j| + cc ] \end{aligned}\ ] ] the first three terms on the rhs above refer to the ground , exit doorway and discretized continuum states , respectively . the fourth term couples the doorway to the discretized continuum states , and the last term represents the continuum - continuum coupling .
if we remove the doorway from the above we get the continuum disctretized coupled channels ( cdcc ) intrinsic hamiltonian @xmath12 + \nonumber \\
\sum_{ij}[|i>\omega_{ij}<j| + cc ] \end{aligned}\ ] ] the exit doorway modulated cdcc hamiltonian , eq . ( 2 ) , is our subject of study here . a full development of this new cdcc will be left for a future work .
here we concentrate our effort on understanding the consequence of reaching the breakup continuum from the entrance channel only through the exit doorway @xmath13 . for this purpose
we ignore the last term in eq .
( 2 ) and remind ourselves that , whereas @xmath14 and @xmath15 are eigenstates of @xmath7 , @xmath13 is not . the full doorway - modulated cdcc equations can be obtained as follows .
the full schrodinger equation of the colliding system is , @xmath16|\psi>=0\ ] ] which when projecting onto the different channels gives : @xmath17 @xmath18 we now invoke the `` exit - doorway '' hypothesis , @xmath19 the overlaps @xmath20 and @xmath21 and can be easily obtained from eq.(2)(without the last term)@xcite , @xmath22 \label{alfa}\ ] ] where @xmath23 the exit - doorway spreading width describing its average coupling to the continuum states of the projectile , is related to the @xmath24 factors through , @xmath25 where @xmath26 is an average value and @xmath27 is the average density of discretized continuum sates in the vicinity of d. clearly the need to the continuum - continuum coupling terms would be very important if exclusive cross sections are to be calculated , since through them ( and through the doorway ) the elastic channels coupling to the breakup channel continuum can be fully acounted for . including the c - c coupling term ,
would result in a more complicated expression for @xmath28 than that of eq.[alfa ] .
equations ( 5 ) and ( 6 ) can be recast into the following , after setting @xmath29 and @xmath30 , @xmath31 @xmath32 the breakup cross section , within the exit doorway model then becomes , @xmath33 where the sum over @xmath34 has been performed by appropriate contour integration over @xmath35 .
note that the q - value in @xmath36 is complex owing to the non - zero width of the exit dooway whose energy is @xmath37 .
a simple way to see how the complex q - value arises is to eliminate @xmath38 in eq.(5 ) in favor of @xmath39 by employing eq.(6 ) , which gives @xmath40 . with this eq.(5 ) becomes @xmath41 . with the exit doorway hypothesis , the polarization potential contribution , @xmath42 becomes @xmath43 this suggests defining the exit - dorway scattering wave function by setting @xmath44 such that eqs.(10 ) and ( 11 ) become : @xmath45 @xmath46 the `` inelastic '' cross - section is thus given by eq.(13 ) above with the aforementioned proviso that the q - value of the excited state is complex .
the width of this q - value is a measure of the continuum contribution to the coupling . at this point
we comment on the inclusion of the continuum - continuum coupling , namely the last term in eq.(3 ) .
in this situation the amplitudes @xmath47 are obtained by matrix diagonalization and , among other things , the resulting overlap probability @xmath28 , deviates from the breit - wigner form of eq.(8 ) . a possible form which may incorporate
some of the c - c effects is a lorentzian : @xmath48 the above form results in an equation for @xmath49 with a modified form factor which depends on the position and width of the exit doorway @xmath50 where @xmath51 is generally complex .
accordingly the cross - section would be : @xmath52 . in the limiting case of @xmath53 , the factor @xmath51
is approximately given by @xmath54 resulting a cross - section given by : @xmath55 .
in the case of coupling to the breakup continuum considered here , the other limit , @xmath56 is more appropriate , as @xmath57 is roughly given by the q - value of the breakup ( @xmath58mev ) while @xmath59 measures the extent in continuum excitation the discretization is performed ( @xmath60 mev ) .
the function @xmath51 can be calculated in such a situation , but we leave this for a future investigation . the important point we are making here , is that a dwba calculation with complex excitation energy in the final state , and with a form factor of the type @xmath61 , should be an adequate candidate to treat the elastic breakup process . in the following
we take the exit doorway to be excited sates of different multipolarities and use the austern - blair sudden / adiabatic theory @xcite .
we employ the distorted wave born approximation for @xmath39 and @xmath36 . the elastic breakup cross section and its dependence on the target masss can be analysed within the distorted wave born approximation ( dwba ) .
if we treat the breakup as an inelastic multipole process , the amplitude @xmath62 would look like : @xmath63 the unpolarized cross section of the dipole transition is then obtained from the expression @xmath64 the radial integrals @xmath65 for pure nuclear excitation are given by , @xmath66 where the form factors @xmath67 are given by the following expressions for the monopole,@xmath68 , dipole , @xmath69 and quadrupole excitations @xmath70 @xcite , @xmath71 & & \\ f_1(r)=-\delta_1^{(n)}(\frac{3}{2 } ) ( \frac{\delta r_p}{r_p})[\frac{dv(r)}{dr}+(\frac{r}{3})\frac{d^2v(r)}{dr^2 } ] & & \\ f_2(r)=-\delta_2^{(n)}\frac{dv(r)}{dr } & & \end{aligned}\ ] ] with @xmath72 being the difference between the rms radii of the neutron and proton distributions of the projectile , and @xmath73 is the elastic scattering channel optical potential .
the quantities @xmath74 and @xmath75 can be extracted from the analysis of refs.@xcite . in ref.@xcite
a power expansion in @xmath76 was employed in the analysis of @xmath77-inelastic scattering from neutron skin nuclei . in the adiabatic limit , @xmath78 , and for large orbital angular momenta , @xmath79 , the radial integral
can then be evaluated in closed form following the procedure of austern and blair @xcite . for the dipole and quadrupole cases we have @xmath80 where @xmath81 is
the nuclear deformation length given by @xmath82 with @xmath83 being the nuclear deformation parameter and @xmath84 is the radius of the excited projectile .
the above expression for the radial integrals can be associated with the nuclear elastic breakup radial integral .
thus we can obtain analytical expression for the integrated nuclear breakup cross section by simply integrating the cross section formula , eq.(2 ) . in performing this calculation the angular momentum
coupling coefficients are evaluated exactly and the sum over @xmath85 can be performed by putting the coulomb phase shifts both as functions of @xmath86 .
the amplitude of eq .
( 13 ) is given now by @xmath87 with the condition that @xmath88 if @xmath89 is odd .
the integrated pure nuclear breakup cross section containing dipole and quadrupole contributions then becomes the following @xmath90 \sum_{l=0}^\infty(2l+1)|\frac{ds^{(n)}(k)}{dl}|^2\ ] ] where terms proportional to the second derivative of @xmath91 have been dropped .
a simple estimate of the above formula can be made by approximating the sum in @xmath92 by an integral in @xmath93 : @xmath94 assuming a real nuclear s - matrix which depends on @xmath95 through @xmath96^{-1}$ ] then the derivative of @xmath97 would peak around the grazing angular momentum @xmath98 with a width given by @xmath99 .
the integral ( [ i ] ) is then obtained as : @xmath100 for @xmath101 . using @xmath102 @xmath103 , with @xmath104 and @xmath105 being the diffuseness of the optical potential we find the simple formula for @xmath106 .
@xmath107 \frac{r}{3a } \label{ab}\ ] ] where @xmath108 is a constant normalization factor which depends among other things on the exit doorway nature of the excited state exemplified by the factor @xmath109 .
it is clear that @xmath106 depends linearly on the radius of the target and , more importantly on the square of the nuclear dipole and quadrupole deformation lengths .
thus , the @xmath110dependence is established . in the calculation to follow we use the cluster model to calculate de deformation lengths for the different multipolarities @xcite this model assumes that the projectile is composed of two clusters , a core of mass and charge @xmath111 and @xmath112 and a `` valence '' particle with @xmath113 and @xmath114 . the separation energy is denoted by @xmath115 , the q - value of the breakup .
calling the spectroscopic factor of finding the cluster configuration in the ground state of the projectile , @xmath97 one obtains the following expression for the distribution of @xmath116 in the excitation energy @xmath117 @xcite .
@xmath118 where @xmath119 is normalization factor which takes into account the finite range , @xmath120 of the @xmath121 potential .
the latter is assumed to be such as to give a yukawa type wave function at large distances , @xmath122 with @xmath123 and @xmath124 .
it is easy to obtain @xmath125 by simply integrating of eq.([b ] ) and employing the expression : @xmath126 } \prod_{k=1}^{2\lambda+1}(\lambda+\frac{3}{2}-k)\ ] ] we get for the cluster - model deformation lenghts @xmath127 and @xmath128 the following : @xmath129 @xmath130 where @xmath131 refers to the projectile . for our three nuclei discussed here
, we have @xmath0be@xmath132be@xmath133 , @xmath134b@xmath135be@xmath136 and @xmath137be@xmath138he@xmath139he , which define their cluster character , with the corresponding breakup q - values , @xmath140 mev , @xmath141 mev and @xmath142 mev .
the factor @xmath143 could be related to the asymptotic normalization coefficient ( anc ) of the bound state wave function and is taken as a parameter to be adjusted so as to account for the experimentally known @xmath125 .
cdcc calculations for the nuclear breakup ( dots ) compared to the results of eq .
( [ ab ] ) .
see text for details.,scaledwidth=50.0% ] simple estimate of @xmath144 can be obtained from @xcite who gave @xmath145 . in table 1
we present the results of the deformation lengths obtained using the cluster model of formulas [ d1 ] and [ d2 ] .
for @xmath0be , we used @xmath146 e@xmath147fm@xmath147 @xcite and we get @xmath149 fm@xmath147 .
further,@xmath150 fm@xmath147 from the same reference . in figure 1
we compare our results , eq.([ab ] ) using @xmath151 fm with the cdcc calculation of ref.@xcite at @xmath152 mev.a .
clearly we underestimate the cdcc calculation .
the reason resides in the neglect , in our model , of the higher - order channel coupling terms alluded to above .
we also show in figure 1 the comparison with the cdcc calculation for @xmath1b and @xmath153be using the formula([ab ] ) .
clearly the scaling law is better obeyed in the `` normal '' nucleus @xmath153be as has already been discussed in @xcite .
the value of the normalization @xmath108 is close to unity for the `` normal '' nuclei @xmath153be for the @xmath1b the normalization is very close the average value of the asymptotic normalization coefficient @xmath154 measured in ref.@xcite . for the @xmath0be a higher normalization
is obtained probably due to higher order effects which are not accounted by our dwba description .
.deformation lengths for the @xmath153be , @xmath1b and @xmath0be projectiles .
the deformation lengths for @xmath153be and @xmath134b have been calculated using formulas ( [ d1 ] ) and ( [ d2 ] ) using @xmath155 for the @xmath0be the @xmath156 and @xmath157 are the values from ref.@xcite [ cols="^,^,^,^,^",options="header " , ] in conclusion , we derived an expression for the nuclear breakup cross - section using the austern - blair theory .
the obtained cross - section exhibits the scaling law and should serve to supply a simple mean for an estimate of the nuclear breakup contribution .
the expression found should be contrasted with the purely geometric serber - like expression @xmath158@xcite . |
over the last three decades the orbital physics of @xmath6 systems and its interplay with spin has focused primarily on transition metal oxides , especially 3@xmath7 transition metals ( tms ) .
while the magnetic @xmath6 configuration has been studied mostly in @xmath8 systems , it can also occur in mid- to late-5@xmath7 tm ionic systems .
for these heavy ions spin - orbit coupling ( soc ) becomes a competing factor , mixing the various spin , orbital , charge , and lattice degrees of freedom .
the interplay of strong electron correlation and large soc is relatively less explored , and certainly not well understood at all , because the behavior involves so many comparable energy scales .
this situation arises in a broad family of magnetic mott insulating systems in which three - fold degenerate @xmath9 orbitals are partially filled.@xcite in such systems orbital degeneracy is protected only by cubic lattice symmetry , and typically the crystal field splitting is large enough that @xmath10 orbitals are out of the picture . in @xmath11 @xmath9
subshells where soc remains unquenched ( cubic symmetry ) , the six one - electron levels split into an upper @xmath12=1/2 doublet and a lower @xmath12=3/2 quartet . in this category
ir - based based magnets have been studied actively.@xcite a prominent class of such systems is the ordered double perovskites , with chemical formula a@xmath0bb@xmath13o@xmath1 .
we are interested in the case where b is a closed shell cation and b@xmath13 is a magnetic ion ; in such cases unusually high formal valence states can arise .
a few examples attracting recent interest are b@xmath13=ru@xmath14 and os@xmath14 in a@xmath0nab@xmath13o@xmath1 ( a = la and nd),@xcite mo@xmath14 in ba@xmath0ymoo@xmath1 , os@xmath15 in ba@xmath0caoso@xmath16@xcite and heptavalent os in ba@xmath0boso@xmath1 ( b = li , na ) .
if we narrow our focus to @xmath6 b@xmath13 ions only , the possibilities are practically confined to mo@xmath14 , re@xmath15 and os@xmath17 ) .
et al . _ have reported a theoretical study@xcite of koso@xmath18 with heptavalent os , where large soc , strong correlations , and structural symmetry breaking conspire to produce an unexpectedly large orbital moment in the @xmath19 shell that nominally supports no orbital moment .
the recently studied compounds@xcite ba@xmath0boso@xmath1 ( b = li , na ) show many features to make them of current interest .
besides the double - perovskite structure , and being a rare example of a heptavalent osmium compound , ba@xmath4naoso@xmath5 ( bnoo ) is exotic in being a _ ferromagnetic mott insulator_,@xcite with order appearing at t@xmath20 = 6.8k with curie - weiss temperature @xmath21 -10k . although its single @xmath9 electron orders magnetically , it shows no evidence of the anticipated orbital order that causes jahn - teller distortion and should destroys its cubic symmetry .
the sister compound la@xmath0naoso@xmath1 , on the other hand , with high - spin @xmath22 os configuration and a nominally cubic symmetry , is observed to be highly distorted.@xcite this distortion is ascribed to geometrical misfit arising from incompatible ionic radii .
there is a recent example of an os - based based @xmath23 perovskite compound baoso@xmath24 that remains cubic;@xcite on the other hand a related perovskite @xmath25 naoso@xmath24 does distort.@xcite the question of origin of the magnetic ordering in bnoo is surely a delicate one , since isostructural , isovalent , and also mott insulating ba@xmath0lioso@xmath1 ( bloo ) orders _ antiferromagnetically _ in spite of a very similar curie - weiss susceptibility@xcite and similar volume .
lee and pickett demonstrated@xcite that , before considering magnetism and on - site interaction effects , soc splits the @xmath9 bands into a lower @xmath12=@xmath26 quartet and an upper @xmath12=@xmath27 doublet , as expected .
since bnoo is observed to be insulating and effects of spin - orbit coupling drive the behavior , it provides the first `` j@xmath28=@xmath26 '' mott insulator at quarter - filling , analogous to the `` j@xmath28=@xmath27 '' mott insulators at half - filling that are being studied in @xmath25 systems.@xcite including spin polarization and on - site hubbard u repulsion beyond the semilocal density approximation ( dft+u+soc ) with both the wien2k and fplo codes gave essentially full spin polarization but was not able to open a gap@xcite with a reasonable value of u. the complication is that the occupied orbital is a oso@xmath5 cluster orbital with half of the charge on os and the other half spread over the neighboring o ions .
u should be a value appropriate to this cluster orbital and should be applied to that orbital , however the codes applied u only to the os @xmath11 orbitals .
xiang and whangbo@xcite neglected the hund s rule @xmath29 in the dft+u method , and did obtain a gap .
however , neglecting @xmath29 omits both the hund s rule exchange energy and the anisotropy ( orbital dependence ) of the hubbard interaction @xmath30 , whereas one of our intentions is to include all orbital dependencies to understand the anisotropy on the os site . in this paper
we first establish how to model this system faithfully including all anisotropy , then address the interplay of soc coupling with correlation effects and crystal field splitting . the density functional extension to include some fraction of hartree - fock exchange the hybrid functional open the gap , but only when soc is included .
the inclusion of nonlocal exchange and soc provides an understanding of the mott insulating ground state and a [ 110 ] easy axis , both in agreement with experimental data.@xcite we conclude that bnoo provides an example of a @xmath31 , quarter - filled mott insulator .
some comparison is made to isovalent bloo , which had 6% smaller volume and aligns antiferromagnetically rather than ferromagnetically .
two density functional theory based studies , mentioned briefly above , have been reported for bnoo .
one was performed within a fully anisotropic implementation of the dft+u method@xcite while the other was dft+u@xcite ( gga+u ) , but neglecting anisotropy of the interaction .
an overriding feature of this system is a strong hybridization of os 5@xmath32 orbitals with o 2@xmath33 states , with the result that the `` @xmath9 '' bands have half of their density on the four neighboring oxygen ions in the plane of the orbital . in keeping with this strong hybridization ,
the `` o @xmath34 '' bands have considerable os @xmath11 charge , such that the @xmath11 occupation of the nominally @xmath2 ion is actually 4 - 5 electrons , still leaving a highly charged ion but less than half of the formal 7 + designation .
lee and pickett reported@xcite that fully anisotropic dft+u could not reproduce a mott insulating state because u is applied on the os ion whereas half of the occupied local orbital ( cluster orbital ) density lies on neighboring oxygen ions . a model treatment
in which u is applied to the cluster orbital did produce the mott insulating state.@xcite whangbo s results indicate that part of the complication in this system involves the anisotropy of the repulsion within the os ion . the study reported by whangbo _ et .
al._@xcite addressed three spin directions ( [ 001 ] , [ 110 ] , and [ 111 ] ) . within their treatment ,
the [ 111 ] spin direction led to the minimum energy , indicated a calculated band gap of 0.3 ev for u=0.21 ry = 2.85 ev ( j@xmath35=0 ) . using dft+u+soc and the same code but including anisotropy of the interaction @xmath30
, we have not reproduced this gap , indicating their gap is due to the neglect of hund s rule coupling and anisotropy of the interaction . because of the need to include all interactions and all anisotropy , we have adopted a different approach based on the hybrid exchange - correlation functional , described in the next section .
this approach seems to be more robust , allowing us to probe the interplay of soc and strong correlation of bnoo , and also obtain results of the effect of pressure on the ground state of bnoo .
our first challenge was to obtain a mott insulating state in bnoo when all interactions ( correlation and soc ) are accounted for . with large soc
the result depends on the assigned direction of the moment .
the hybrid functional approach plus soc leads directly to a ferromagnetic ( fm ) mott insulating ground state , as observed.@xcite in our studies we observed a strong preference for fm alignment , versus the commonplace antiferromagnetic ( afm ) alignment that often arises on the simple cubic lattice of perovskite oxides .
ba@xmath4lioso@xmath5 however orders antiferromagnetically , which for nearest neighbor antialignment exchange coupling leads to frustration of ordering .
all calculations reported here for either compound are for fm orientation .
we study specifically the effects of spin orientation on the electronic structure , and initiate a study of the pressure dependence of bnoo considering the zero pressure lattice constant and at 1% , 2% , and 5% reduced lattice constants .
the present first - principles dft - based electronic structure calculations were performed using the full - potential augmented plane wave plus local orbital method as implemented in the wien2k code.@xcite the structural parameters of bnoo with full cubic symmetry of the double perovskite structure were taken from experimental x - ray crystallographic data:@xcite @xmath36=8.28 , @xmath37=0.2256 . non - overlapping atomic sphere radii of 2.50 , 2.00 , 1.80 , and 1.58 a.u . are used for the ba , na , os , and o atoms , respectively .
the brillouin zone was sampled with a minimum of 400 k points during self - consistency , coarser meshes were sometimes found to be insufficient . for the exchange - correlation energy functional for treating strongly correlated insulators , a variety of approaches in addition to dft+u exist and have been tested and compared for a few selected systems.@xcite as mentioned above , for technical reasons
the relevant orbital is an octahedron cluster orbital rather than the standard localized , atomic - like orbital encountered in @xmath8 oxides
the lda+u method is problematic .
we have chosen to apply the onsite exact exchange for correlated electrons ( eece ) functional introduced and evaluated by novak and collaborators.@xcite , implemented similarly to common use in hybrid ( mixture of hartree - fock and local density exchange ) .
this oeehyb functional is an extension of the dft+u method to parameter - free form : exact exchange with full anisotropy is evaluated for correlated orbitals ( os @xmath11 orbitals here ) without explicit reference to any ( screened or unscreened ) hubbard u repulsion or hund s exchange interaction j@xmath35 .
the double counting term is evaluated directly from the density of the occupied correlated orbitals , again without any input parameters .
the exact exchange is calculated within the atomic sphere in atomic - like fashion ( hence `` onsite '' ) .
the onsite exact exchange replaces 25% of the local density exchange .
this functional is implemented in the wien2k code,@xcite and we refer to it here as oeehyb . for the semilocal exchange - correlation functional , the parametrization of perdew , burke , and ernzerhof@xcite ( generalized gradient approximation ) is used .
soc was included fully in core states and for valence states was included in a second - variational method using scalar relativistic wave functions,@xcite a procedure that is non - perturbative and quite accurate for @xmath32 orbitals even with large soc .
this oeehyb method has some kinship with hybrid exchange - correlation functionals ( see tran _
et al._@xcite for a comparison of several hybrids ) .
hybrids replace some fraction @xmath38 , typically 25% , of local density exchange with hartree - fock exchange , which then is approximated in various ways to reduce the expense to a reasonable level .
the oeehyb approach deals with exact exchange only for correlated orbitals , however , making it appropriate for correlated materials but it will not increase bandgaps of ionic or covalent semiconductors .
we note that as an alternative to the commonly used dft+u approach , the more conventional hybrid exchange - correlation functional as implemented in the vasp code@xcite has been applied to the iridate na@xmath4iro@xmath39 by kim _
et al._@xcite to obtain the magnetic insulating ground state .
projected densities of states ( pdoss ) are presented in fig . [ dos_fig1 ] for bnoo at the experimental volume using oeehyb , initially without soc and with full cubic symmetry .
the fermi level ( taken as the zero of energy ) lies in a deep pseudogap , due to small band overlap .
all @xmath9 orbitals participate equally at this level , evident from the observation that the pdos is distributed almost equally over the os @xmath9 orbitals and the @xmath34 orbitals of the six neighboring o ions , a very strong hybridization effect that has been emphasized before.@xcite the conventional picture of a mott insulating state in a @xmath40 shell is that a single orbital , say @xmath41 , is occupied , and the crystal symmetry is broken ( and must be broken in the calculation ) to sustain , indeed to allow , occupation of a single orbital . however , due to the large soc which is expected to produce a substantial orbital moment which requires occupation of a complex ( _ viz . _
@xmath42 ) orbital , we have foregone this intermediate step of obtaining orbital - ordering broken symmetry , which would typically be the final result for a @xmath8 ion with negligible soc .
adding soc to oeehyb , the result mentioned above , which necessarily lowers the symmetry , leads to a soc - driven mott insulating state with a 0.2 - 0.3 ev gap , depending on the direction assumed for the magnetization .
it is for this reason that we use the oeehyb functional to model bnoo , as we have previously reported that the lda+u approach was unable to open a gap.@xcite the gga+u method is more promising , but it requires an unphysically large value of @xmath43 = 6 ev to open a gap for [ 001 ] orientation .
one can obtain a band gap along for [ 110 ] and [ 111 ] orientations in gga+u with u values of 3.2 ev and 2.0 ev , respectively.@xcite our calculated band gaps , provided in table [ table1 ] , depend on direction of the magnetization , which experimentally can be manipulated with an applied field .
it should be kept in mind that symmetry is lowered by including soc , and the resulting symmetry depends on the direction of magnetization . for spin along [ 001 ]
the band gap is lower by 70 mev than for [ 110 ] and [ 111 ] .
.calculated spin , orbital , and total ( @xmath44=@xmath45+@xmath46 ) moments of os , and band gap of ba@xmath4naoso@xmath5 , for four values of lattice parameter and for the three high symmetry directions of the magnetization .
note : the total moment per f.u .
will include an @xmath470.5@xmath3 spin moment not included within the os sphere , thus primarily on the oxygen ions . [ cols="^,^,^,^,^,^,^,^,^",options="header " , ]
the hybrid density functional has been used for the modeling of ba@xmath4naoso@xmath5 with considerable success . unlike our attempts with gga+u+soc , when soc is included , and only then , it provides a robust fm mott insulating ground state for the moment aligned along any of the three symmetry directions .
thus bnoo ( and bloo ) is a @xmath48=@xmath26 mott insulator at quarter filling , driven by the combination of strong exchange - correlation and large soc .
though the spin magnetization remains completely spin - up when soc is included , the large changes in the composition of the occupied orbital are driven by soc .
on the scale of 0.1 - 0.2 ev which is important considering the small gap , the band structure is strongly dependent of the orientation of the magnetization .
in addition , our approach predicts the [ 110 ] direction as the easy axis , as observed , though [ 111 ] is close in energy .
the character of the net moment is unexpected . on the one hand , the spin and orbital moments on os
are 0.5 and -0.4@xmath3 , respectively , leaving a net moment on os of 0.1@xmath3 , similar to earlier indications.@xcite however this value leaves most of the total moment of 0.6@xmath3 arising from the oxygen ions in the oso@xmath5 cluster .
the observed ordered moment is @xmath470.2@xmath3 thus indicates somewhat larger compensation , or small spin contribution , than our results provide .
another unusual feature found here is that for spin along [ 001 ] , there is a 0.1@xmath3 orbital moment on the apical oxygen ion .
this surprisingly large value can not arise from the very small soc on oxygen ; rather it is the hybridization of the os @xmath49 combination with the @xmath50 orbital that transfers orbital angular momentum to o. this effect does not operate for other spin orientations .
we acknowledge many useful conversations with k .- w .
lee , discussions with i. r. fisher , r. t. scalettar , and n. j. curro , and comments on the manuscript from r. t. scalettar .
our research used resources of the national energy research scientific computing center ( nersc ) , a doe office of science user facility supported by the office of science of the u.s .
department of energy under contract no .
de - ac02 - 05ch11231 .
this research was supported by doe stockpile stewardship academic alliance program under grant de - fg03 - 03na00071 .
27 g. chen , r. pereira , and l. balents , phys .
b. * 82 * , 174440 ( 2010 ) .
g. jackeli and g. khaliullin , phys .
lett . * 102 * , 017205 ( 2009 ) .
b. j. kim , h. jin , s. j. moon , j .- y .
park , c. s.leem , j. yu , t.w .
noh , c. kim , s .-
oh , j. h. park , v. durairaj , g. cao , and e. rotenberg , phys .
* 101 * , 076402 ( 2008 ) .
s. j. moon , h. jin , k.w .
kim , w. s. choi , y. s. lee , j. yu , g. cao , a. sumi , h. funakubo , c. bernhard , and t.w .
noh , phys .
rev . lett .
* 101 * , 226402 ( 2008 ) .
b. j. kim , h. ohsumi , t. komesu , s. sakai , t. morita , h. takagi , and t. arima , science * 323 * , 1329 ( 2009 ) .
g. cao , t. f. qi , l. li , j. terzic , s. j. yuan , l. e. delong , g. murthy , and r. k. kaul , phys .
lett . * 112 * , 056402 ( 2014 ) .
a. a. aczel , d. e. bugaris , j. yeon , c. de la cruz , h .- c .
zur loye , and s. e. nagler , phys .
b. * 88 * , 014413 ( 2013 ) .
a. a. aczel , p.j.baker , d. e. bugaris , j. yeon , h .- c .
zurloye , t. guidi , and d. t. adroja , phys .
. lett . * 112 * , 117603 ( 2014 ) .
a. a. aczel , d. e. bugaris , l. li , j .- q .
yan , c. de la cruz , h .- c .
zur loye , and s. e. nagler , phys .
b. * 87 * , 014435 ( 2013 ) .
k. yamamura , m. wakeshima , and y. j. hinatsu , solid state chem .
* 179 * , 605 ( 2006 ) .
song , k .- h .
ahn , k .- w .
lee , and w. e. pickett , arxiv:14087.4078 .
a. j. steele , p. j. baker , t. lancaster , f. l. pratt , i. franke , s. ghannadzadeh , p. a. goddard , w. hayes , d. prabhakaran , and s. j. blundell , phys .
b. * 84 * , 144416 ( 2011 ) .
a. s. erickson , s. misra , g. j. miller , r. r. gupta , z. schlesinger , w. a. harrison , j. m. kim , and i. r. fisher , phys .
. lett . * 99 * , 016404 ( 2007 ) .
k. e. stitzer , m. d. smith , and h .- c .
z. loye , solid state science * 4 * , 311 ( 2002 ) .
w. r. gemmill , m. d. smith , r. prozorov , and h .- c .
zur loye , inorg . chem .
* 44 * , 2639 ( 2005 ) .
y. shi , y. guo , y. shirako , w. yi , x. wang , a. a. belik , y. matsushita , h. l. feng , y. tsujimoto , m. arai , n.wang , m. akaogi , and k.yamaura , j. am .
soc . * 135 * , 44 ( 2013 ) .
jung , y - j .
song , k - w .
lee , and w. pickett , phys .
b. * 87 * , 115119 ( 2013 ) .
k. w. lee and w. e. pickett , epl * 80 * , 37008 ( 2007 ) .
h. j. xiang and m .- h .
whangbo , phys .
b. * 75 * , 052407 ( 2007 ) .
p. blaha , k. schwarz , g. madsen , d. kvasicka , and j. luitz , wien2k , an augmented plane wave + local orbitals program for calculating crystal properties , technical university of vienna , vienna , ( 2001 ) .
we have used version 12 of wien2k , because we have found that soc with the hybrid functional does not always work correctly in version 13 .
f. tran , p. blaha , k. schwarz , and p. novak , phys .
b * 74 * , 155108 ( 2006 ) .
p. novk , j. kune , l. chaput , and w. e. pickett , phys .
b * 243 * , 563 ( 2006 ) .
g. kresse and j. furthmller , phys .
b. * 54 * , 11169 ( 1996 ) .
kim , j .- h . lee , and j .- h .
cho , sci .
rep . * 4 * , 5253 ( 2014 ) doi:10.1038/srep05253 .
a. h. macdonald , w. e. pickett , and d. d. koelling , j. phys .
c * 13 * , 2675 ( 1980 ) . | the double perovskite ba@xmath0naoso@xmath1 ( bnoo ) , an exotic example of a very high oxidation state ( heptavalent ) osmium @xmath2 compound and also uncommon by being a ferromagnetic mott insulator without jahn - teller ( jt ) distortion , is modeled using the density functional theory ( dft ) hybrid functional based exact exchange for correlated electrons ( oeehyb ) method and including spin - orbit coupling ( soc ) .
the experimentally observed narrow gap ferromagnetic insulating ground state is obtained , with easy axis along [ 110 ] in accord with experiment , providing support that this approach provides a realistic method for studying this system . the predicted spin density for [ 110 ] spin orientation is nearly cubic ( unlike for other directions ) , providing an explanation for the absence of jt distortion . an orbital moment of -0.4@xmath3 strongly compensates the + 0.5@xmath3 spin moment on os , leaving a strongly compensated moment more in line with experiment .
remarkably , the net moment lies primarily on the oxygen ions .
an insulator - metal transition by rotating the magnetization direction with an external field under moderate pressure is predicted as one consequence of strong soc , and metallization under moderate pressure is predicted .
comparison is made with the isostructural , isovalent insulator ba@xmath4lioso@xmath5 which however orders antiferromagnetically . |
diffuse interstellar bands ( dibs ) are mysterious absorption structures of interstellar origin observed in spectra of reddened stars of early spectral types .
the detection of the first two dibs was described by heger in 1922 .
these structures , as we know them today , are scattered within the whole region of visible light and in the near infrared ( see e.g. herbig 1995 ) .
individual dibs differ between themselves in intensity , line width and profile shapes .
there are many arguments that 5797 and 5850 dibs have the same carrier . almost all procedures isolating dibs families indicate that 5797 and 5850 bands tend to belong to the same family ( e.g. chlewicki et al . 1986 ,
josafatson and snow 1987 , kreowski and walker 1987 , wszoek and godowski 2003 ) .
wszoek and godowski ( 2003 ) additionally confirmed the status of 5850 band as a member of 5797 spectroscopic family , that means that both these dibs have probably the same carrier .
they composed sequences of spectrograms , of different target stars , in such a way that equivalent width ( ew ) of 5797 was substantially changing from one case to the other .
they realized that the depth of 5850 band regularly follows the depth of 5797 band . to check how good candidate to 5797 spectroscopic family
is the dib at 5850 angstroms , we carried out measurements of ews for both dibs in spectra of 74 target stars .
we also carried out the analysis of mutual correlation between them .
all the spectra analyzed for the purpose of this contribution were taken from the archives of professor jacek kreowski ( astronomical centre , nicolaus copernicus university , toru , poland ) .
we used spectra taken with the canada - france - hawaii telescope and spectra acquired at the mcdonald observatory with an echelle spectrograph fed with the 2.1-m telescope .
considered dibs were present always in the same order of the spectra .
spectra are of quite good quality , as far as resolution and signal / noise are concerned .
using spectra of 74 target stars we have measured ews for dibs at 5797 and 5850 angstroms .
that was done with the help of software package rewia v. 1.4 , written by jerzy borkowski ( nicolaus copernicus astronomical center , polish academy of sciences , toru , poland ) .
we calculated correlation coefficient for measured values .
we drew correlation diagram , which is exposed in figure 1 . on the figure
one may notice small dispersion of points around the straight line .
it is most probably caused by errors in observation and in data reduction .
correlation coefficient r=0.9 was counted and it is very high . in practice , there exist no other pair of dibs with such good correlation as 5797 and 5850 ones .
5850 band lies in the spectra very close to 5797 dib .
although 5850 is substantially weaker than 5797 , it is still a relatively strong dib .
that means that for very reddened stars
ew(5850 ) measurement errors are not too big . furthermore ,
5850 morphologically resembles very much the 5797 band .
all groupings of dibs into families , known from literature , show that 5850 and 5797 belong to the same morphological family .
correlation diagram for large number of target stars , exposed in the figure 1 , and counted coefficient of correlation strengthen very much the hypothesis that 5797 and 5850 dibs have the same carrier .
we would like to thank professor jacek kreowski for giving us an access to his data archives . | the carriers of diffuse interstellar bands are still mysterious species . there exist many arguments that diffuse bands at 5797 and 5850 angstroms have the same carrier . using high - resolution spectra of few dozens of reddened stars
we have searched mutual correlation between intensities of considered bands .
results of our analysis indicate that 5797 and 5850 really tend to have the same carrier .
# 1_#1 _ # 1_#1 _ = # 1 1.25 in .125 in .25 in |
the evolution of pre - main - sequence ( pms ) stars young contracting stars on their way to the main sequence is poorly understood although significant progress have been made in recent years ( see * ? ? ?
* ; * ? ? ?
* for reviews ) .
observationally , these stars are also among the hardest to study because they are usually heavily extincted by their parental molecular cloud and are often seen in projection against bright hii regions
. the ideal laboratory of pre - main - sequence stellar evolution would then be the youngest possible galactic cluster free from dust extinction and nebula contamination , at a distance that permits detection of low - mass members to at least the hydrogen burning limit .
an investigation of such an ideal cluster would provide robust observables to test pre - main - sequence models ( luminosity functions , color - magnitude , and color - color diagrams ) , free from completeness corrections and scatter due to variable extinction and nebula emission . in turn
, the age of this very young cluster would be well constrained , as would be its star formation history , and its underlying initial mass function ( imf ) .
the age of this cluster would also set an upper limit on the timescale for the disruption of the molecular material that originated the cluster .
finally , observations from x - rays down to the mid - infrared wavelengths would also be able to provide a transparent characterization of an entire pre - main - sequence population down to brown dwarf masses , while shedding new light on the x - ray / uv / h@xmath9/infrared - excess phenomena observed in young stars ( see * ? ? ?
* ; * ? ? ?
* for reviews ) .
in this letter we present a ubvri survey of the stellar cluster ngc 2362 , probably the best proxy to the ideal cluster described above .
the cluster is virtually free from dust extinction and shows no signs of nebular emission .
walter baade suggested more than fifty years ago that ngc 2362 appeared to be almost exclusively made up of b stars with little evidence for low mass stars @xcite . this same population of b stars has been used as the standard observational template to define the upper end of the zero age main sequence . despite the claims for an abnormal mass function , a deep @xmath10band study of the central @xmath11 region of the cluster @xcite uncovered a substantial population of lower mass stars whose mass function appeared to be similar to that of the field @xcite .
recently , @xcite presented sensitive jhk observations of ngc 2362 that confirm this finding .
more interesting perhaps , they found virtually no stars with detectable near infrared excess emission , suggesting that this young cluster largely consists of diskless , post - t tauri stars . in a survey of young clusters at l band ( 3.4 @xmath12 m )
, @xcite confirmed this result and further found ngc 2362 to be the youngest known cluster without a significant population of disk bearing stars .
consequently , ngc 2362 plays a pivotal role in the determination of the overall lifetime of circumstellar disks in clusters and for setting the timescale allowed for planet building within such disks .
unfortunately , the age of this cluster is not well constrained .
the best existing age estimate essentially rests on the position in the hr diagram of a single object , @xmath13cma , the cluster s only post - main sequence star @xcite . an improved age determination for the cluster is urgently needed .
our new deep ubvri survey of ngc 2362 reveals a long , narrow , and well defined pre - main - sequence , spanning about 9 magnitudes , in the @xmath8 band , from a - stars down to about @xmath14m@xmath4 , close to the hydrogen burning limit .
this enables us to provide an improved and independent determination of the cluster s age .
comparison with modern calculations of pms evolution yields an age of @xmath0 5@xmath1 myr which closely agrees with the age we obtain from the cluster s early type stars as derived from post - mais - sequence models .
our observations of the early type stars also yield new estimates for the cluster s distance and extinction which we find to be in good agreement with previous determinations .
the remaining of this letter is organized as follows : we present the observations in section 2 , the results in section 3 , and conclude with a discussion on section 4 .
ccd ubvr and gunn _ i _ optical data were collected with dfosc mounted at the 1.5 m danish telescope in la silla during the nights of 2001 february 2 and 4 .
the field - of - view of dfosc was @xmath15 with a plate - scale of 0.39@xmath16 per pixel . due to severe distortions on the edges ,
the images were trimmed which resulted in an effective field - of - view of @xmath17 .
a total of 36 images for the cluster , 10 images for the control field , and 36 images of standard fields , were obtained . to overcome the observational difficulties imposed by @xmath13cma ( severe bleeding , scattered light , ghost images ) resulting in significant data corruption at the very center of the cluster (
@xmath18 , @xmath19 ) , a series of exposures of ngc2362 were taken at four positions ( nsew ) as close as possible to @xmath13cma but avoiding it or its scattered light ( see fig .
[ fig : mosaic ] ) . during the same night we acquired short exposures ( 1 sec ) centered on @xmath13cma to get photometry for the bright stars at the center of the cluster and , in particular , those that were in the region ( an area of @xmath20 ) not covered by the nsew mosaic . a nearby control field ( @xmath21 north of the cluster )
was also observed , with the same exposure times as the cluster field ( @xmath22 and @xmath23s in the @xmath24 and gunn _ i _ bands , respectively ) , during the night of february 4th .
the control field observations covered one field of view of the ccd .
although the integration times for the cluster and control field were the same , the control field data is not as deep due to the proximity of the moon during the control field observations .
calibration into the standard system was achieved by the observation of several @xcite fields . the typical seeing ( fwhm ) throughout the run was @xmath25 .
all ccd frames were processed within iraf .
images were subjected to the the usual bias and flatfield corrections , and cosmic rays were removed .
photometry was performed with the iraf / daophot package .
a @xmath26 ( @xmath27 pixel ) aperture radius was used in standard star photometry .
the photometry , aperture corrections , extinction corrections , transformation to the standard system and construction of the photometric catalog were performed following the procedures thoroughly described in @xcite .
the final catalog contains photometry for approximately 8300 stars ( @xmath7 4500 stars in all ubvri bands ) in the cluster field , plus approximately 2000 stars in the control field .
in fig . [ fig : hr]a we introduce the @xmath8 vs. @xmath28 diagram of the stars contained in a @xmath29 diameter circle centered on ngc2362 ( hereafter the _ on _ field ) .
the center of the cluster was estimated to be the peak of the stellar density distribution .
the distribution was computed using the stars falling along the cluster sequence delineated in fig .
[ fig : hr]a ( both zams and pms stars ) .
the center is practically coincident with @xmath13cma .
although previous estimates of the cluster s diameter give a diameter of @xmath30 @xcite , we find a significant concentration of stars , above the background , in the inner @xmath29 diameter .
[ fig : hr]b shows the @xmath8 vs. @xmath28 diagram for the control field ( hereafter the _ off _ field ) .
the area covered on the sky is the same as in the left panel .
the distribution of stars in the _ on _ field color - magnitude diagram follows a sequence which splits into two clearly separated branches below about @xmath8 = @xmath31 magnitudes .
the red branch of this sequence extends at least from @xmath32 to @xmath33 and has no counterpart in the _ off _ field color - magnitude diagram .
this red branch is part of a very long pms band of stars which starts at about @xmath34 mag and is clearly visible from @xmath35 to @xmath36 mag .
the reddening towards ngc2362 was determined by fitting the @xcite zams to the b - star sequence in the @xmath37 ) vs. @xmath38 diagram .
we have determined @xmath39 mag , which is in agreement with the previous estimates by @xcite and @xcite .
no evidence of variable extinction was found at a @xmath40 mag level in @xmath41 .
the distance to ngc2362 was determined by fitting the @xcite zams to the b - star sequence in different color - magnitude diagrams . unlike older clusters , where the a type stars are already on the main sequence , the zams stars in ngc2362 are almost exclusively of spectral type b. this makes the distance determination more difficult since the b - star portion of the zams is nearly vertical for most color - magnitude combinations .
the color that provides a b - star sequence with the lowest possible slope is @xmath42 .
therefore , distance was mainly constrained by the fit in the @xmath8 vs. @xmath42 plane .
the apparent distance modulus was found to be @xmath43 mag . using the standard value for the ratio of total - to - selective absorption , @xmath44
, we find the distance to ngc 2362 to be @xmath45 pc ( @xmath46 ) which is virtually the same as the value adopted by @xcite ( @xmath47 .
the age of ngc2362 was determined following two different approaches . in the first case ,
we have derived a turn - off age by fitting solar composition post - main - sequence isochrones of @xcite to color - magnitude diagrams ( using various color - magnitude combinations ) . in the second case ,
the pre - main - sequence isochrones of @xcite ( solar recipe ) were used in deriving an age based on the location of the low - mass pms stars in the @xmath8 vs. @xmath48 plane .
these particular pms models were selected because they are the most consistent with constraints imposed by recently derived dynamical masses of pms binary stars @xcite .
the solar recipe version of the pms models has only been computed for the @xmath49m@xmath4 mass range , which was the region used in estimating the age .
the @xcite models have also been computed for lower masses ( @xmath50m@xmath4 ) using non solar mixing lengths ( which is unimportant at such low masses ) .
however , the predicted optical colors at these cool temperatures are not robust ( indeed we found departures from the observed sequence ) , and therefore have not been used in the age fits .
the youth of ngc2362 poses two problems to the turn - off age determination .
first , the observed b - star sequence does not show any evident evolutionary deviation from the zams ( in fact , ngc2362 has been used in the past to define the upper zams ) .
second , the high - mass end is dominated by poor statistics , which adds extra difficulty to the analysis : the only star that exhibits evolutionary effects is @xmath13cma which , to make things even harder , is found to be a multiple system ( eg . * ? ? ?
[ fig : age ] shows a 5 myr isochrone fit in the @xmath8 vs. @xmath51 and @xmath8 vs. @xmath48 planes .
the filled circle in the bright end is @xmath13cma , which has been plotted using the photometry of @xcite .
a bar reaching 0.75 mag fainter indicates the magnitude it would have if it was a simple ( non - multiple ) system .
the fit has been essentially constrained by the position of @xmath13cma considering that it is a system of two similar - o9ib - stars .
this age determination agrees with the one by @xcite , which was also based on the position of @xmath13cma .
as previously mentioned , we have also determined the cluster s age by fitting pre - main sequence isochrones of @xcite to the low - mass sequence on the @xmath8 vs. @xmath48 diagram ( fig .
[ fig : age ] ) . in this case ,
none of the problems of the turn - off fit arise : the sequence is well populated , and the evolutionary status is quite obvious ( the pms stars are well above the main sequence ) .
furthermore , the luminosity of the pms branch is a sensitive function of age .
also , the absence of variable reddening within the cluster results in an extremely well defined pms which also helps to constrain the age .
the best fit was obtained for a 5 myr isochrone ( lines on the right of fig .
[ fig : age ] ) .
a 0.75 mag brighter pms isochrone was also plotted to account for the presence of binaries .
that the pms is so tight ( well below the binary limit ) is indicative of a very simple star formation history , most probably a single and quick episode of star formation .
the age spread of the pms stars in the cluster is likely less than 3 myr ( see below ) . to estimate the age uncertainty we have not considered the uncertainties in the models themselves .
taking into account the photometric and fitting errors we estimate an uncertainty of @xmath52 myr .
the low dispersion of the pms , together with the sensitivity of the pms luminosity to age , have allowed us to place tight constraints on the age .
a somewhat higher uncertainty towards a younger age arises from the fact that we fit the pms isochrones to the _ lower / bluer envelope _ of the observed sequence .
although the _ lower / bluer envelope _ fitting is usually performed in order to avoid the effects of binary contamination and rotation , it could lead the pms fit towards slightly older ages if those effects are not important .
we note here that the ages derived from the pre - main sequence and post- main sequence isochrones are in excellent agreement with each other as well as with the previous age determination for the cluster which was based primarily on the position of @xmath13cma in the hr diagram ( balona and laney 1996 ) . moreover these age estimates are also consistent with the dynamical age estimated for the hii region surrounding the cluster and excited by @xmath13cma @xcite .
we report ubvri observations of ngc 2362 , a very young open cluster in cma , and a nearby control field .
the main findings of this letter are : 1 . we derive the following fundamental parameters for ngc 2362 : age @xmath0 5@xmath1 myr , distance @xmath0 1480 pc , and reddening @xmath53 0.10 mag .
cluster ages were obtained separately from both the high - mass population ( ob - stars on the main sequence ) and the low - mass ( pms ) population of the cluster and found to be in excellent agreement .
2 . analysis of this cluster s color - magnitude diagram reveals a long , narrow , and well defined cluster pre - main - sequence , spanning about 9 magnitudes ( in @xmath8 ) in the optical color - magnitude diagrams , with a negligible scatter .
this pms extends from early a stars to late type stars near the hydrogen burning limit .
star formation in the cluster was likely characterized by a single quick episode or burst spanning an interval of no more than about 3 million years .
the observed properties of ngc2362 make it an ideal laboratory for testing pre - main sequence models and investigating early stellar evolution .
we thank the referee for his / her useful comments which improved the paper .
a. moitinho acknowledges financial support by conacyt ( mexico ; grant i33940-e ) and dgapa ( mexico ; grant in111500 ) .
this research made use of the nasa astrophysics data system , of the simbad database operated at the centre de donns stellaires strasbourg , france , and of the webda open cluster database of j .- c . | we present ubvri photometry for the young open cluster ngc 2362 . from analysis of the appropriate color - color and color - magnitude diagrams we derive the fundamental parameters of the ngc 2362 cluster to be : age @xmath0 5@xmath1 myr , distance @xmath0 1480 pc , @xmath2 0.10 mag .
the cluster age was independently determined for both high mass ( @xmath3m@xmath4 ) and low mass ( @xmath5m@xmath4 ) stars with excellent agreement between the ages derived using post - main sequence @xcite and pre - main sequence @xcite evolutionary tracks for the high and low mass stars respectively .
analysis of this cluster s color - magnitude diagram reveals a well defined pre - main sequence ( covering @xmath6 v @xmath7 9 magnitudes in @xmath8 and extending from early a stars to near the hydrogen burning limit ) which makes this cluster an ideal laboratory for pre - main sequence evolution studies . |
in the context of coded communication , the channel coding theorem relates the error probability and the code rate , showing that there exist codes whose error probability tends to zero provided that the code rate is smaller than the channel capacity . for uncoded systems , the error probability and the channel capacity
are also related . in particular , references @xcite show that given one of the two values , tight bounds on the other can be given for the family of binary - input memoryless and symmetric ( bims ) channels .
such channels are described by the channel transition probability @xmath2 , where @xmath3 and @xmath4 .
we assume that the channel output alphabet @xmath5 has finite size , though our approach also holds for well - behaved channels with infinite alphabet size , like the binary - input additive white gaussian noise ( biawgn ) channel .
we adopt gallager s definition of symmetric channel @xcite , that is , a channel is said symmetric if the channel transition probability matrix ( rows corresponding to input values ) is such that it can be partitioned in submatrices for which each row is a permutation of any other row and each column is a permutation of any other column .
both the binary erasure channel ( bec ) and the binary symmetric channel ( bsc ) are symmetric .
more precisely , references @xcite show that the uncoded error probability of any bims channel with capacity @xmath0 is upper - bounded by that of the bec and lower - bounded by that of the bsc of the same capacity .
similar results have been found in @xcite for the bhattacharyya parameter , a simple upper bound to the uncoded error probability ; here only the extremal property of the bec was proved . in the context of iterative decoding , analogous
extremal properties of the bec and bsc have been found @xcite for the building blocks of iterative decoders for low - density parity - check codes , namely variable - node and check - node decoders .
upper and lower bounds to the error probability of good codes can be given in terms of error exponents , e.g. gallager s random coding bound ( * ? ? ?
5.6.3 ) , the sphere - packing bound by shannon _
_ @xcite and arimoto s strong converse bound @xcite .
these exponents are expressed as optimization problems involving gallager s @xmath1 function ( * ? ? ?
* eq . 5.6.14 ) , @xmath6 where @xmath7}{p_{y|x}(y|x)^{\frac{1}{1+\rho}}}\right)^\rho\right]\ ] ] and the pair @xmath8 is distributed according to @xmath9 .
here and throughout the paper @xmath10 $ ] denotes the expectation of a random variable and all logarithms are in base @xmath11 .
equiprobable inputs maximize the @xmath1 function for bims channels @xcite , and we henceforth assume such distribution , i.e. @xmath12 . in this paper , we characterize the feasible values of @xmath13 for an arbitrary bims channel of fixed channel capacity @xmath0 and show that the @xmath1 function is upper - bounded ( resp .
lower - bounded ) by that of the bec ( resp .
bsc ) of the same capacity . since the aforementioned exponents are expressed using the @xmath1 function , we are able to find their extremal values .
in fact , our analysis leads to similar results for the cutoff rate , the bhattacharyya parameter , the channel dispersion and to a number of other extensions .
the @xmath14 functions for the bec and bsc of erasure / crossover probability @xmath15 , respectively denoted by @xmath16 and @xmath17 , are given by @xmath18 using the capacity expressions for the bec , @xmath19 , and bsc , @xmath20 , we can find the erasure / crossover probability corresponding to a given capacity @xmath0 and parametrize the @xmath16 and function @xmath17 as functions of @xmath0 , namely @xmath21 where @xmath22 is the binary entropy function , and @xmath23 denotes the inverse of @xmath24 for @xmath25 $ ] .
@xmath26 are respectively defined as the inverses of eqs .
, with respect to @xmath0 . for bims channels , one has the bounds @xmath27 and @xmath28 for @xmath29 and @xmath30 for @xmath31 .
this is a consequence of the facts that @xmath14 is non - negative and non - increasing for @xmath32 ( * ? ? ?
5b ) , that @xmath33 , and that @xmath34 . it is however not apparent whether further limitations exist on the feasible pairs of capacity @xmath0 and @xmath14 . against this first impression
, the next theorem tightly characterizes the set of possible pairs of capacity @xmath0 and @xmath14 function for any bims channel ( see figure [ fig : extremes_f0_c ] ) . in the next section
, we apply this theorem and prove several analogous characterizations for other relevant quantities in the analysis of the error probability over bims channels .
[ th : extremes_f0 ] for any bims channel with capacity @xmath0 and function @xmath14 for @xmath35 , the following statements hold 1 . the function @xmath14 of the channel satisfies @xmath36 2 . the capacity @xmath0 of the channel satisfies @xmath37 the extremes in are attained by the bec and the bsc .
furthermore , for a given pair @xmath38 satisfying the inequalities in or , there exists a bims channel with capacity @xmath0 and function @xmath14 .
conversely , if the inequalities do not hold for the pair @xmath38 , there exists no such bims channel with capacity @xmath0 and function @xmath14 . the proof is built around the idea that every bims channel admits a decomposition into subchannels that are bscs .
this decomposition follows directly from gallager s definition of symmetric channels @xcite as used in this paper .
a formal description may be found e.g. in @xcite .
here we deem identical the bec with erasure probability 1 and the bsc with crossover probability @xmath39 . in this decomposition , each channel output @xmath40 is associated with an index @xmath41 which is independent of the input and depends on the channel output only .
we denote by @xmath42 the probability mass or density function of subchannel @xmath43 , and by @xmath44 the corresponding binary output alphabet of the bsc with index @xmath43 .
assuming such a decomposition , and since @xmath45 @xcite we have @xmath46}{p_{y|x}(y|x)^{\frac{1}{1+\rho}}}\right)^\rho\right]\\ & = { \mathbb{e}}\left[{\mathbb{e}}\left [ \left(\frac{{\mathbb{e}}\bigl[p_{y|x , a}(y|x',a)^{\frac{1}{1+\rho}}|y , a\bigr]}{p_{y|x , a}(y|x , a)^{\frac{1}{1+\rho}}}\right)^\rho\bigg| ~a\right]\right]\\ & = { \mathbb{e}}\left[f^{{\rm bsc}}(\rho;c(a))\right],\end{aligned}\ ] ] where @xmath47 denotes the capacity of subchannel @xmath43 .
the following lemma is proved in appendix [ app : f - is - concave ] .
[ lem : f - is - concave ] the function @xmath48 is concave in @xmath49 $ ] for any @xmath32 , non - decreasing for @xmath31 , and non - increasing for @xmath29 . noting that @xmath50 = c$ ] , and given the concavity of the function @xmath48
, we apply jensen s inequality to obtain @xmath51 \\ & \leq f^{{\rm bsc}}(\rho ; { \mathbb{e}}[c(a ) ] ) \\ & = f^{{\rm bsc}}(\rho;c ) .
\end{aligned}\ ] ] the bound is obviously achieved when the channel is a bsc .
since @xmath48 is concave , we can lower - bound it by a straight line joining the points @xmath52 ( @xmath53 ) and @xmath54 ( @xmath55 ) ( see figure [ fig : extremes_f0_c ] ) , and then evaluate the expectation , i.e. , @xmath56 this bound is obviously achieved when the channel is a bec , thus proving eq . . for @xmath57 .
the upper curves correspond to the bsc and the lower straight lines to the bec . ]
eq . determines the boundaries of the region of feasible pairs @xmath38 .
since @xmath48 is concave and @xmath58 is convex , the region of feasible pairs is convex .
moreover , the functions @xmath48 and @xmath58 are non - decreasing for @xmath31 and non - increasing for @xmath29 . fixing the value of @xmath14
, the convexity of the region implies eq . .
we next prove that the region of feasible pairs @xmath38 is connected by constructing a bims channel with corresponding capacity @xmath0 and function @xmath14 .
consider a binary symmetric - erasure channel ( bsec ) with input alphabet @xmath59 , output alphabet @xmath60 , cross - over probability @xmath61 and erasure probability @xmath62 .
its transition probabilities are given by @xmath63 , @xmath64 , and @xmath65 .
the capacity @xmath66 and function @xmath67 are respectively @xmath68 for fixed @xmath0 , there exist several bsec channels with capacity @xmath0 , among them a bsc and a bec . each of them is characterized by a pair of probabilities @xmath61 and @xmath62
. the corresponding @xmath67 function is given by eq . .
since the function @xmath67 is continuous in @xmath61 and @xmath62 , one can always find a bsec with capacity @xmath0 whose function @xmath67 coincides with the desired @xmath14 . in the proof of theorem
[ th : extremes_f0 ] , we exploited the fact that the region of feasible pairs @xmath38 is convex and connected to characterize the extreme values of the capacity @xmath0 or the function @xmath14 . in this section
, we apply the theorem to provide extreme values for other relevant quantities in the error probability analysis of channel coding .
a simple extension to channel parameters @xmath69 given by @xmath70 , where @xmath71 is a monotonic continuous function , will prove convenient .
[ th : extremes_general ] let the channel parameter @xmath69 be given by @xmath70 , where @xmath71 is a monotonic strictly increasing continuous function .
for any bims channel , we have that 1 . the channel parameter @xmath72 satisfies @xmath73 2 .
the channel capacity @xmath0 satisfies @xmath74 the inequalities are reversed if @xmath71 is monotonic , strictly decreasing and continuous .
[ [ gallagers - function ] ] gallager s function + + + + + + + + + + + + + + + + + + + by letting @xmath75 , the previous theorem readily gives the extremes of gallager s function @xmath76 for a fixed capacity , and the extremes of the capacity for a fixed @xmath1 .
[ [ cutoff - rate ] ] cutoff rate + + + + + + + + + + + a particular case of the @xmath1 function is the cutoff rate , given by @xmath77 .
thus , the above result also gives the extremes of the cutoff rate .
[ [ bhattacharyya - parameter ] ] bhattacharyya parameter + + + + + + + + + + + + + + + + + + + + + + + a related quantity is the bhattacharyya parameter @xmath78 , given by @xmath79 the bsc / bec have the largest / smallest possible bhattacharyya parameter for bims channels of capacity @xmath0 , interestingly giving the reverse extremes of the uncoded error probability @xcite .
this result recovers sason s @xcite and arikan s @xcite bound for the bec , and provides the extreme in the other direction attained by the bsc .
[ fig : extremes_cz ] shows the bounds to @xmath0 for a given value of @xmath78 from theorem [ th : extremes_general ] , as well as arikan s generic bounds for binary - input discrete memoryless channels ( * ? ? ?
( 1 ) , ( 2 ) ) , illustrating some improvement . as a function of the bhattacharyya parameter @xmath78 .
s upper and lower bounds ( * ? ? ?
( 1 ) , ( 2 ) ) and the biawgn channel curve ( dashed line ) are also shown for reference . ]
[ [ random - coding - exponent ] ] random coding exponent + + + + + + + + + + + + + + + + + + + + + + the random coding exponent @xmath80 ( * ? ? ? * sect .
5.6 ) , given by @xmath81 provides an upper bound to the error probability of codes of rate @xmath82 .
this exponent involves a maximization of a function that , for fixed @xmath83 falls under the conditions for applicability of theorems [ th : extremes_f0 ] and [ th : extremes_general ] .
therefore , the exponent @xmath80 satisfies @xmath84 figure [ fig : error_exponents_extremes_c ] illustrates the extremes of random - coding error exponents @xmath80 .
the random - coding error exponent of an arbitrary bims channel must lie in the shaded area , two such examples are the biawgn channel of the same capacity ( with and without fading ) . [
[ expurgated - error - exponent ] ] expurgated error exponent + + + + + + + + + + + + + + + + + + + + + + + + + for rates below the channel critical rate , the expurgated error exponent @xmath85 ( * ? ? ? * sect .
5.7 ) , given by @xmath86 provides a tighter estimate of the error probability of good codes than the random - coding exponent .
the function @xmath87 is expressed in terms of the bhattacharyya parameter @xmath78 as @xmath88 theorem [ th : extremes_general ] provides the extremes of the expurgated exponent .
[ [ strong - converse - exponent ] ] strong converse exponent + + + + + + + + + + + + + + + + + + + + + + + + in @xcite , arimoto lower - bounded the error probability of block codes at rates above capacity in terms of the function @xmath89 given by @xmath90 theorem [ th : extremes_general ] also provides the extremes of this exponent .
[ [ sphere - packing - exponent ] ] sphere - packing exponent + + + + + + + + + + + + + + + + + + + + + + + the error probability of codes of rate @xmath82 is lower - bounded by a bound that depends on the sphere - packing exponent @xcite @xmath91 , given by @xmath92 again , theorem [ th : extremes_general ] provides the extremes of this exponent .
[ [ threshold - decoding - error - exponents ] ] threshold - decoding error exponents + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the exponent of random - coding bounds based on threshold decoding can also be expressed in closed form .
shannon @xcite derived the exponent of feinstein s bound to the error probability @xcite .
more generally , the exponent corresponding to a generalized form of feinstein s bound @xcite can be expressed as @xmath93 theorem [ th : extremes_general ] directly gives the error exponent extremes for the generalized feinstein s bound .
the exponent of the dependence - testing bound @xcite is @xcite @xmath94 where @xmath95 , for @xmath96 , and @xmath97}{p_{y|x}(y|x)^s}\right)^\rho\right].\ ] ] following similar and somewhat simpler steps to those in the proof of lemma [ lem : f - is - concave ] , one can prove that @xmath98 , evaluated for a bsc with capacity @xmath0 , is concave in @xmath0 .
therefore , theorem [ th : extremes_general ] holds and shows that the exponent of the dt bound has similar extreme values .
[ [ channel - dispersion ] ] channel dispersion + + + + + + + + + + + + + + + + + + recently , the gaussian approximation to the error probability @xmath99 of length-@xmath100 codes at rates close to the capacity has received renewed attention . in this approximation ,
a critical channel parameter is the dispersion @xmath101 , which for bims channels @xcite is given by @xmath102 moreover , it can be proved that one can choose either the @xmath13 function or the simpler @xmath103 to compute the latter derivative , that is @xmath104 .
as proved in appendix [ app : bounded_i3 ] , the third derivative of @xmath103 at @xmath105 is bounded for bims channels .
thus , a second - order taylor expansion of @xmath103 around @xmath106 shows that @xmath107 has the same extremes as @xmath103 . as illustration ,
figure [ fig : extremes_dispersion_v ] depicts the possible values of channel dispersion as a function of the capacity @xmath0 of the bims channel . the dashed line , which lies within the shaded area indicating the feasible region of pairs capacity / dispersion , corresponds to the biawgn channel . .
] [ [ error - probability - of - specific - codes ] ] error probability of specific codes + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + our theorems may also be applied to specific codes @xmath108 with a given distance spectrum by means of the shulman - feder bound @xcite ( see also @xcite ) given by @xmath109 where @xmath110 is a function of the distance spectrum of the code that quantifies how far the distance spectrum of @xmath108 is from that of the ensemble average .
[ [ exact - error - probability ] ] exact error probability + + + + + + + + + + + + + + + + + + + + + + + one might wonder whether our extremal results extend to the actual error probability .
the answer is not immediately obvious . for uncoded transmission ( a code of length @xmath111 and rate @xmath112 ) over a given bims channel of capacity @xmath0
the error probability @xmath113 is upper- and lower - bounded by that of the bec and the bsc , respectively , @xmath114 @xcite .
in contrast , the extremes of the exponential bounds to the error probability , including the bhattacharyya parameter , are reversed .
this phenomenon suggests the existence of a pair @xmath115 such that a crossing point occurs , in the sense that for rates above ( resp . below )
@xmath116 the extremes may be those of uncoded transmission ( resp .
the error exponents ) .
[ [ connection - with - arikan - telatar - and - alsan ] ] connection with arikan , telatar @xcite , and alsan @xcite + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + unpublished work by arikan and telatar @xcite uncovered results of similar nature to those reported in this paper , showing that for channels with a fixed rate @xmath117 , for @xmath118 , the random coding exponent satisfies @xmath119 for @xmath120 we have that @xmath121 and we obtain the trivial result that @xmath122 . instead
, our results compare channels of a fixed capacity and provide the extremal values of the random - coding exponent and other quantities .
the suitability of either of these two approaches to the problem may depend on the specific application .
a more recent result by alsan @xcite recovers both theorem [ th : extremes_f0 ] in this paper and the results in @xcite as particular cases , for bims channels in the interval @xmath123 .
we aim at proving the concavity of the function @xmath124 where @xmath125 is itself a function of @xmath0 , namely @xmath126 . without loss of generality ,
we limit our attention to the interval @xmath127 $ ] .
the function is concave if @xmath128 .
applying the chain rule of derivation , we have that @xmath129 direct computation gives @xmath130 an application of the inverse function theorem yields @xmath131 the derivatives with respect to @xmath0 are therefore given by @xmath132 since we have that @xmath133 for @xmath31 and @xmath134 for @xmath29 , we conclude that @xmath135 is increasing and decreasing in the respective ranges of @xmath83 .
let @xmath140 $ ] . with this change of variables we obtain @xmath141 we wish to show that @xmath142 .
the partial derivative with respect to @xmath83 is given by @xmath143 we are interested in the sign of @xmath144 , whose derivative is in turn given by @xmath145 we readily see that @xmath146,\\ \frac{\partial g_0(z,\rho)}{\partial \rho } & \leq 0,~~~~~\rho\in[1,+\infty).\end{aligned}\ ] ] * @xmath148 in @xmath149 $ ] , since @xmath144 is non - decreasing and @xmath150 , * @xmath151 in @xmath152 $ ] , since @xmath144 is non - decreasing and @xmath153 , * @xmath151 in @xmath154 , since @xmath144 is non - increasing and @xmath155 .
we wish to prove that the partial derivative @xmath157 is bounded . to this end
, we first note that the function @xmath158 can be expressed as @xmath159,\end{aligned}\ ] ] where @xmath160 is the information density , defined as @xmath161 the function @xmath158 is a cumulant generating function .
its third derivative evaluated at @xmath120 gives the third - order cumulant , that is the third - order central moment , @xmath162 ( \ln 2)^2.\ ] ] [ lemma : mink ] consider a memoryless channel with discrete input alphabet @xmath164 and arbitrary output alphabet @xmath5 .
then , with equiprobable inputs we have @xmath165\leq \left(2\log|{{\cal x}}| + \frac{k}{\ln2}\bigl(1 + |{{\cal x}}|^{\frac{1}{k}}\bigr)\right)^k.\ ] ] we will make use of minkowski s inequality @xmath166 where @xmath167)^{\frac{1}{k}}. $ ] using the definition of @xmath168 , we now have that @xmath169\right)^\frac{1}{k}\\ & \leq 2\log |{{\cal x}}| + \frac{k}{\ln2 } + \frac{k}{\ln2 } |{{\cal x}}|^\frac{1}{k}\end{aligned}\ ] ] where we have used that @xmath170 ( * ? ? ?
* eq . ( 4.1.37 ) ) .
a. martinez and a. guilln i fbregas , `` random - coding bounds for threshold decoders : error exponent and saddlepoint approximation , '' in _ 2011 ieee int .
theory , saint petersburg , russia _ , jul .- aug . 2011 . | this paper determines the range of feasible values of standard error exponents for binary - input memoryless symmetric channels of fixed capacity @xmath0 and shows that extremes are attained by the binary symmetric and the binary erasure channel .
the proof technique also provides analogous extremes for other quantities related to gallager s @xmath1 function , such as the cutoff rate , the bhattacharyya parameter , and the channel dispersion . |
a red blood cell ( rbc ) has been a standard test bed of membrane physics due to its simple structure lacking a nucleus and organelles .
early studies , treating a rbc as a lipid bilayer bag , already rendered insightful understanding of various properties , to name a few , biconcave shapes@xcite , flickering phenomenon @xcite , and tank - treading motion @xcite .
still there are certain features signifying the role of other substructure ; a spectrin network ( cytoskeleton ) coupled to the bilayer is found to be responsible for the shape transformation sequence @xcite , large - scale shape changes under shear flow @xcite , and crenated shapes @xcite .
many attempts have been made in order to provide simplified descriptions of the cytokeleton as for its effects on non - trivial shapes and fluctuation of rbcs @xcite . meanwhile , theoretical works @xcite analyzed the fluctuation spectrum measured by zilker _
et al . _
@xcite and introduced a picture of a composite membrane , that is , a lipid bilayer sparsely connected to the cytoskeleton . in ref .
@xcite , through an empirical approach , it was claimed that the coupling to the cytoskeleton induces surface tension in such a way that the effective bending modulus of the lipid bilayer undergoes an abrupt jump at the crossover length @xcite .
this claim was elaborated by considering the cytoskeleton explicitly as spring meshwork @xcite , where the authors examine the elastic energy of the meshwork as a function of the membrane area coarsely grained at the mesh size @xmath0 and thereby conclude that the tension contribution vanishes suddenly at length scale @xmath0 .
also , a model hamiltonian was proposed , assuming infinitely strong coupling of the membrane at sites linked to the cytoskeleton , so as to explain the induced tension in the long wavelength fluctuations @xcite . in this
regard it seems that a rbc as a composite membrane necessarily acquires surface tension at long length scales .
however , clear signature of the surface tension was not always observable , and also there exists subtle variance between recent observations .
et al . _
@xcite reported the shape dependence of the fluctuation spectrum where the tension contribution is fairly pronounced for spherocype - shaped rbcs but almost unnoticeable for discocytes .
in contrast , popescu _ et al .
_ observed clearly the induced surface tension irrespective of shapes @xcite . using similar interferometric technique in ref .
@xcite , many red blood cells showed bending dominant fluctuations at short wavelengths and confining dominant long wavelength fluctuations , but the intermediate tension dominant region could be seen from few number of the samples @xcite . in all those experiments ,
the notion of tension jump seems to be not evident as derived from the theories @xcite .
the origin of such diversity is left unknown , and more experimental evidences together with theoretical models are anticipated in order to draw a unified picture of a rbc as a composite membrane . in this study , we propose an exactly solvable model to describe elastic properties of a rbc membrane , positing that the cytoskeleton is spring network and the membrane is sparsely and regularly linked to the network via harmonic coupling .
unlike assumed in ref .
@xcite , the coupling strength is finite and may even be weak due to flexibility of protein complex linking the bilayer and the cytoskeleton and also due to fast dissociation / association process of the cytoskeleton controlled by atp concentration @xcite .
although governing factors are not known precisely , it is reasonable to think that the coupling strength can vary depending on experimental conditions , and observation results should intrinsically be diverse . in our consideration
, the coupling strength is introduced as an essential model parameter which interplays with other involved energy scales to determine the tension emergence .
this paper is organized as follows . in sec .
ii , we introduce a model describing energies associated with fluctuations of the membrane , deformation of the spring network , and the coupling between them .
the effective hamiltonian for the equilibrium fluctuation of the membrane is obtained in sec .
iii , and analytic expression of the fluctuation spectrum is derived in sec .
iv . section v consists of three subsections . in sec . v. a
, various kinds of the qualitatively distinctive fluctuation spectra are analyzed .
we find that at length scale shorter than the network mesh size , the cytoskeleton elasticity is negligible irrespective of the coupling strength , and the fluctuation spectrum of the membrane exhibits the fluid - like behavior . at long wavelengths ,
the coupling to the cytoskeleton imparts elasticity to the membrane , and yet resulting features in the fluctuation spectrum thereof are strongly dependent on the system parameters . the effect of confining potential is discussed in sec . v. b. in the last subsection v. c , we compare our results with previous studies , in particular related to the tension jump " stated in ref .
@xcite , and conclusion follows in sec .
we consider a lipid bilayer of a rbc as an almost flat , symmetric(zero spontaneous curvature ) fluid membrane . assuming small deformations without overhangs , the membrane height can be represented by a single - valued function @xmath1 where @xmath2 spans the flat reference plane at @xmath3 .
the corresponding hamiltonian reads as : @xmath4=\int d{\mathbf r } ~\kappa ( \nabla^2 h)^2 + \gamma h^{2 } ~.\ ] ] the second term , so called the confining potential , is introduced phenomenologically to describe the long wavelength fluctuations of red blood cell membranes @xcite .
closed geometry @xcite , periodic pinning @xcite and excluded - volume effect @xcite were suggested as the physical origin of the confinement , yet no consensus has been met . although the term has been widely used for many experimental studies @xcite , there are also skeptical views on the necessity of the confining potential @xcite .
to embrace every possible situations , we do not discard this phenomenological term from the model .
however , overall discussion in this paper will be made with very weak confinement so that the effect of the confinement is marginal to the fluctuation spectrum ; the limit @xmath5 would not hurt the key lessons of our research .
more subtle issues on the confining potential would be discussed in sec . v.
b. we also mention that in order to reflect area conserving property of a bilayer , the bending energy usually includes effective bare surface tension , which is dependent on the excess area of the bilayer and ambient temperature @xcite .
however , we intentionally discard the bare tension in the hamiltonian @xmath6 in order to explicate the induced tension by the cytoskeleton .
another ingredient comprising the system is the cytoskeleton which is two - dimensional regular network of spectrin tetramers .
we consider the cytoskeleton as spring network @xcite and model it as a sparse square array of beads connected by springs . since we assumed the small deformations , the in - plane stretching and the out - of - plane deformation
is approximately decoupled @xcite .
in addition , the single particle tracking experiment also showed that the diffusion of the band-3 proteins which are connected to the spectrins are significantly restricted @xcite .
hence , we assume that the in - plane position of the @xmath7th bead is fixed to form a square lattice with uniform spacing @xmath0 and represented by the position vector @xmath8 with integers @xmath9 and @xmath10 . on the other hand , we allow out - of - plane displacements of the beads and express the displacement of the @xmath7th bead by @xmath11 .
the imbalance in the displacements between the nearest neighboring beads brings about either contraction or dilation of the connecting springs . assuming the identical spring constant @xmath12 , the elastic energy cost for such deformation of the spring network is , @xmath13=\sum_{\langle { \bf x}_{i},{\bf u}\rangle}k[\phi({\mathbf x}_{i})-\phi({\mathbf x}_{i}+{\bf u})]^2~,\ ] ] where @xmath14 or @xmath15 are the primitive vectors for the square lattice .
in fact , the cytoskeleton is closer to triangular ( or hexagonal ) network .
yet we expect that effects of the symmetry details must be weak on the qualitative properties of the membrane fluctuation and hence pursue simple analytic approach , remaining in the picture of the four - fold network .
we finally introduce interaction between the membrane and the spring network .
the coupling between the two systems is mediated by protein inclusions in the membrane and adaptor proteins ( for example , ankyrin ) bridging the membrane inclusion and the spectrin network .
hence , the coupling must be the consequence of various kinds of microscopic interactions among the membrane , inclusions , adaptor proteins , and spectrins . rather than deriving the coupling energy from all these complex origins , here we approximate it under the assumption that the separation between the membrane and the coupled cytoskeleton @xmath16 is small .
as the leading order of taylor expansion , we can then model the coupling energy by harmonic potentials between the out - of - plane displacement of the bead and the height undulation of the membrane at the bead location @xmath17 : @xmath18=\sum_{{\mathbf x}_{i}}v[h({\mathbf x}_{i})-\phi({\mathbf x}_{i})]^2~,\ ] ] where @xmath19 measures the coupling strength . since recent numerical research based on the harmonic membrane - cytoskeleton interaction showed good agreement with experiments on red blood cell membranes @xcite
, we expect the harmonic approximation is sufficient and ignore the higher order terms @xcite . in total , the system hamiltonian of our interest is composed as @xmath20 . in ref .
@xcite , the authors considered the bilayer - cytoskeleton interaction by imposing a hard constraint that the membrane height at the coupling site is identical with the bead location , i.e. , @xmath21 .
in our model , this corresponds to an extreme situation of infinite coupling ( @xmath22 ) , and the behaviors at general coupling strengths are still to be investigated .
even though steric interaction can also be involved to correlate the membrane and the cytoskeleton @xcite , this research neglects the excluded - volume effect and assumes the phantom " membrane and network .
equilibrium properties of the system can be extracted from evaluating the canonical partition function : @xmath23}~ \\ \nonumber & = & \int { \cal d}h ~e^{-\beta { \cal h}_{\mathrm{eff}}[h ] } ~,\end{aligned}\ ] ] where the second line defines the effective hamiltonian for the height deformation of the membrane , @xmath24 $ ] , which can be obtained by integrating the boltzmann factor over the fields @xmath25 . for mathematical convenience ,
let us obtain the hamiltonian in the momentum space .
for the height field @xmath26 of the membrane , we employ the fourier transform in two dimensions , @xmath27 here we discretize the space in units of the length @xmath28 which is a microscopic length scale , e.g. , molecular size of a lipid molecule consisting the membrane , and let the momentum @xmath29 take values to satisfy the periodic boundary conditions , @xmath30 for @xmath31 with @xmath32 being the lateral dimension of the reference plane .
we then obtain the bending energy of the membrane in terms of @xmath33 as @xmath34=\sum_{{\bf q}}e_{m}({\bf q } ) |h_{{\bf q}}|^{2}~,\ ] ] where the bending energy spectrum in the discrete space is given by @xmath35^{2}+\gamma$ ] .
since we are interested in the length scale much longer than the size of the monomer , we take the continuum limit and describe the bending energy of the bilayer as @xmath36 .
the remaining components of the hamiltonian are given in terms of the fields defined only at the link sites @xmath17 s , for which we introduce the fourier transform as @xmath37 this is again the discrete fourier transform but the length unit is given by @xmath0 instead of @xmath28 . in order to discern the fourier transform , eq .
( [ ft1 ] ) , we give the tilde symbol to the function in the momentum space . here a commensurate condition @xmath38 with an integer @xmath39 is assumed , and @xmath40 is the number of link sites along one direction . for the periodic boundary conditions , the values of @xmath41 are given by integer multiples of @xmath42 but lie in the restricted brillouin zone ( rbz ) , @xmath43 where the zone boundary @xmath44 reflects the discrete nature of the spring network in our theory .
with eq .
( [ ft2 ] ) applied , the elastic energy of the spring network , eq .
( [ h_e ] ) , is diagonalized into @xmath45=\sum_{{\bf q}\in \mathrm{rbz}}e_{e}({\bf q})|{\widetilde \phi}_{{\bf q}}|^{2}~,\ ] ] where @xmath46 denotes the summation over the range given in eq .
( [ rbz ] ) , and the energy spectrum is given by @xmath47 $ ] , which gives @xmath48 behavior only in the long - wavelength limit specified by the length scale @xmath0 : @xmath49 .
on the other hand , the coupling energy , eq .
( [ v ] ) , is transformed as @xmath50= \sum_{{\bf q}\in \mathrm{rbz } } v|{\widetilde \phi}_{{\bf q}}-{\widetilde h}_{{\bf q}}|^2~.\ ] ] given the hamiltonian components in the momentum space by eqs .
( [ ft1 h_m ] ) , ( [ ft2 h_e ] ) , and ( [ ft2 v ] ) , we perform integration in eq .
( [ z ] ) over the field @xmath51 and obtain the effective hamiltonian governing the membrane height fluctuations as follows : @xmath52=\sum_{{\bf q}}e_{m}({\bf q})|h_{{\bf q}}|^{2}+\sum_{{\bf q}\in \mathrm{rbz } } \frac{ve_{e}({\bf q})}{v+e_{e}({\bf q})}|{\widetilde h}_{{\bf q}}|^{2}~.\ ] ] the second term induced by the coupling to the spring network describes the energy for the height deformations occurring over length scales longer than @xmath0 . in the infinite coupling limit
, one can find that the coefficient of @xmath53 becomes @xmath54 , which yields harmonic couplings in the real space representation , @xmath55 ^ 2 $ ] , as considered in ref .
@xcite . for finite coupling strength , expanding the coefficient as a series of @xmath56 gives the first - order term , @xmath57 , and the corresponding energy is given by a lattice laplacian , @xmath58 ^ 2 $ ] , where the negative sign indicates energy gain to create large - scale ( @xmath0 ) curvature .
now we investigate the fluctuation spectrum of the membrane in equilibrium , @xmath59 .
the angular bracket denotes the equilibrium average defined as @xmath60}h_{{\bf q}}h_{{\bf q}}^{*}~,\ ] ] where the partition function @xmath61 and the effective hamiltonian @xmath62 $ ] are given in eqs .
( [ z ] ) and ( [ h_eff1 ] ) , respectively . in order to perform the integration , we need to express @xmath63 in the effective hamiltonian ( [ h_eff1 ] ) in terms of @xmath33 .
reciprocal lattice vectors @xmath64 with integral coefficients @xmath65 and @xmath66 are defined by the orthogonality relation for the discrete lattice @xmath67 : @xmath68 which gives the inverse transform of eq .
( [ ft2 ] ) , @xmath69 here we can replace @xmath70 with the right hand side of eq .
( [ ft1 ] ) with @xmath71 .
then , from eq .
( [ ortho ] ) it follows that @xmath72 where @xmath29 is restricted to the range in eq .
( [ rbz ] ) . using this relation ,
we write the effective hamiltonian : @xmath73=\sum_{{\bf q}\in \mathrm{rbz}}\sum_{{\bf t},{\bf t } ' } h^{*}_{{\bf q}+{\bf t}}m_{{\bf q}}({\bf t},{\bf t}')h_{{\bf q}+{\bf t}'}~,\ ] ] where the summation over @xmath29 in the bending energy term is folded into the restricted brillouin zone ( rbz ) , and the interaction kernel is given by @xmath74 with @xmath75 note that @xmath76 induced by the bilayer - network coupling describes the all - to - all coupling between any deformation fields with their wave vector difference given by the reciprocal lattice vector of the spring network , reflecting the fact that length scales shorter than @xmath0 are indiscernible as far as the network effect is concerned . with the effective hamiltonian given by eq .
( [ h_eff2 ] ) , the integration in eq .
( [ define psd ] ) is performed to give @xmath77 with @xmath78 which is the main results of our model for red blood cell membranes .
two factors enter the characteristics of the fluctuation spectra : the all - to - all coupling energy kernel @xmath79 and the bending energy kernel for wave vectors in the high - order brillouin zone @xmath80 .
because the effective cytoskeleton elastic energy @xmath81 is the harmonic mean of these two terms , the smaller energy contributes more to the fluctuation spectrum .
this competition of two factors can result in new kinds of fluctuation spectra of the coupled membrane , which will be discussed in the next section .
although the phenomenological model proposed by gov _ et al_. explains the long wavelength fluctuation data of ref .
@xcite , the bilayer coupled to the two - dimensional meshwork may not follow such description for all possible set of parameters @xmath82 , @xmath12 , and @xmath19 , and it may possess potential diversity in its fluctuation spectrum . as given in eq .
( [ psd ] ) , the characteristics of the fluctuation spectrum are determined by three competing energies : @xmath83 , @xmath84 , and @xmath76 . roughly speaking , for given @xmath85 and the system parameters , one of the energy functions , which satisfies @xmath86~,\ ] ] determines the main properties of the fluctuation spectrum . here
the coupling strength @xmath19 regulates the form of @xmath76 .
it should be also mentioned that the fluctuation spectrum has directional dependence which can be removed by angular average .
we have confirmed that the average does not show any qualitatively different properties apart from a constant factor .
every different type of fluctuations ( e.g. , non - monotonic fluctuation spectrum ) can also be observed from the angular averaged fluctuation spectra .
we will thus let @xmath87 henceforth .
_ strongly coupled membrane _ : if the coupling energy @xmath19 is sufficiently larger than the cytoskeleton elastic energy , i.e. , @xmath88 , the coupling energy function is given by @xmath89 where the last approximation is valid in the long wavelength limit .
figure 1 ( a ) shows the relevant energy functions for various @xmath12 values .
note that in the region of interest , among @xmath90 , the term with @xmath91 makes the best contribution to the summation in eq .
( [ c_eff ] ) .
we only consider @xmath92 which crosses with @xmath93 at the zone boundary @xmath94 , as displayed in fig .
the magnitude of these bending energies in relative to the elastic energy depends not only on @xmath85 but also on the spring network rigidity @xmath12 , which according to eq .
( [ cri ] ) results in the variety of fluctuation spectrum as presented in fig .
we add the inset displaying the power - law exponent @xmath95 , @xmath96 in order to see more clearly the wavenumber dependence of the fluctuation spectrum .
one common feature among fluctuation spectra for various network rigidities is that in the short wavelength region ( @xmath97 ) , the bending energy @xmath93 is dominant over @xmath92 and @xmath98 , and the fluctuation spectrum shows the fluidic behavior @xmath99 . on the other hand , qualitative behavior of the long wavelength fluctuation ( @xmath100 ) is crucially altered by the elasticity @xmath12 . for small @xmath101 ( the dashed lines )
, we have that @xmath102 and @xmath103 ( see fig . 1(a ) ) .
reminding of the criterion eq .
( [ cri ] ) , we expect the fluctuation spectrum to be well described by the bare bending energy @xmath93 , which is indeed so as shown in fig .
while such soft network does not influence the mechanical properties of the lipid bilayer , for large @xmath12 there exists a wavenumber region satisfying @xmath104 ( see the curves for @xmath105 in fig .
1(a ) ) . in this case
, the elastic energy contribution from the cytoskeleton is observable in the fluctuation spectrum with @xmath106 behavior in fig .
1(b ) . note that for @xmath107 , near the zone boundary @xmath108 , the higher order bending energy @xmath92 is comparable or smaller than @xmath98 and yields the non - monotonic fluctuation spectrum . in a recent paper @xcite a fluid membrane
discretely coupled to a very hard fluid membrane is found to show non - monotonic fluctuations , and its origin was interpreted by negative surface tension . from our analysis
it becomes clear that the non - monotonicity arises from the bending energy for creating short wavelength curvature which is less costly than the deformation of the extremely rigid cytoskeleton , that is , @xmath109 .
the non monotonic behavior is also reflected on the power law exponent @xmath95 which has a deep near @xmath110 , while for other @xmath12 values it monotonically decreases from @xmath111 to @xmath112 with @xmath113 plateau for @xmath114 .
_ weakly coupled membrane _ :
let us consider that the bilayer - cytoskeleton coupling is not strong enough , i.e. , @xmath115 . defining a wavenumber @xmath116 at which @xmath117
, we obtain the approximate form of the coupling energy function @xmath98 : @xmath118 which well describes the behavior illustrated in figure 2(a ) .
this coupling energy should be compared with the bending energy costs , and the fluctuation nature is again determined by the dominant function which follows eq .
( [ cri ] ) . in the short wavelength region ( @xmath97 ) , the bare bending energy @xmath93 is dominant over the other relevant energy functions and results in @xmath99 dependent fluctuation spectrum , as displayed in fig .
2(b ) . in the region
@xmath100 , the network rigidity comes into play . for small @xmath12 ( see the dashed lines for @xmath119 ) , @xmath93 is still a dominant energy factor , which explains well the fluctuation spectrum presented in fig .
2(b ) . for rather rigid network , at intermediate wavenumber
the coupling energy @xmath98 can be less than @xmath92 and yet larger than @xmath93 . in the case
, @xmath85-independent fluctuation can be observable due to the coupling function behavior for @xmath120 in eq .
( [ wcc ] ) .
this coupling induced confining effect is more pronounced for large @xmath12 , for example , @xmath121 .
such rigid network imparts the elasticity for long wavelength ( @xmath122 ) undulation of the membrane yielding @xmath106 behavior .
this is a unique consequence of the finite coupling strength , which was not expected from the gov s model @xcite and other previous studies @xcite .
fluctuation spectrum can also be influenced by the confining potential @xmath123 in eq .
( [ h_m ] ) . up to now , although we have considered only very small confining strength @xmath124 relative to the bending energy @xmath125 , the confining strength of red blood cell membranes is not known precisely . in this section ,
we discuss the effect of the confining potential on the fluctuation spectrum . since the confining potential provides a ( positive ) constant shift to the bending energy @xmath126 , the role of the coupling can be insignificant when the confinement is too strong compared to the cytoskeleton elasticity . in order for the elastic contribution from the spring network to be visible in the spectrum , in particular , the @xmath106 dependence , the confining should be sufficiently weak , for which rough estimation gives a condition @xmath127 note here that @xmath128 is the wavenumber at which the elastic energy @xmath129 in the long wavelength limit equals to the confining potential , and @xmath130 and @xmath116 are roughly the upper boundaries that @xmath129 dominates the bending energies for the strong coupling ( sc ) and the weak coupling ( wc ) case , respectively .
different microscopic models have been proposed to understand the physical origin of the cytoskeletal confinement .
inhomogeneous pinning due to sparse connection with the cytoskeleton @xcite and the closed geometry of rbcs @xcite are two major candidates .
our results suggest that the two pictures entail very different physical situations . in our model the discrete coupling , @xmath131 in eq .
( [ v ] ) , indeed leads to the @xmath85-independent fluctuation at long wavelength .
this coupling induced confinement is apparent only if the membrane is weakly coupled to very rigid cytoskeleton ( see fig .
2(b ) ) , and in the longer wavelength region ( @xmath132 ) , one should observe the tension signature as well .
for the closed geometry of rbcs , the translational symmetry of the elastic network is broken .
this can be effectively described by adding constant to the elastic energy : @xmath133 , which modifies only the all - to - all coupling kernel @xmath98 .
since the geometry induced confining @xmath134 is now part of @xmath98 , for its observability it requires the strong coupling and soft cytoskeleton .
such distinctive features depending on the origin of the confinement would help to verify which description is closer to reality .
we discuss the tension emergence , comparing our results with previous studies . in refs .
@xcite , the authors analyzed fluctuation spectra observed in ref .
@xcite , and focused on the quantity @xmath135 where @xmath136 , the effective bending rigidity , is defined as @xmath137 . assuming abrupt increase of surface tension in the long wavelength regime , they pointed out that jump occurs in @xmath135
. our model does produce neither sudden change of tension nor jump in @xmath135 curve .
instead , the tension signature only gradually appears as @xmath106-dependence in @xmath138 . in order to explain the tension jump in our model , the coupling energy function @xmath98 in eq .
( [ c_q ] ) should vanish at @xmath110 .
however , in our approach based on the microscopic model , the coupling to the spring network causes all - to - all correlations between membrane deformations , as given in eq .
( [ h_eff2 ] ) , yielding the effective coupling energy @xmath98 which does not vanish at @xmath97 but becomes less significant than the bare bending energy . therefore , introducing a sudden change of tension at @xmath139 in refs .
@xcite can not be justified within the physics explained in our model
. induced surface tension can also be observed in @xmath138 rather than @xmath135 .
the parameters in refs .
@xcite , estimated from the fit to the experiment @xcite , is marginal or discordant for observing even the gradual tension emergence in the fluctuation spectrum .
when the bending modulus @xmath140 @xcite , the estimated parameter values are @xmath141 and @xmath142 , which gives @xmath143 .
this does not meet the condition , eq .
( [ conf ] ) , and the signal of induced tension in the fluctuation spectrum should be either weak or absent .
when we use an order of magnitude larger value of the bending modulus , @xmath144 , which is also used in recent computer simulations @xcite , the non - trivial fluctuation spectrum due to the coupling with the cytoskeleton becomes less visible . for a given @xmath12 and @xmath124
, the effect of different values of @xmath82 is absorbed to the dimensionless parameters @xmath145 and @xmath146 , i.e. , increase of @xmath82 is equivalent to decrease of @xmath145 and @xmath146 . in figure
[ fig : sclog ] , the decrease of @xmath147 makes the strongly coupled membrane s peculiar properties like non - monotonic fluctuation near @xmath148 and the signature of the induced tension be less visible . with larger bending modulus
, @xmath149 gets greater so that the tension signature becomes more difficult to be observed .
similarly , the figure [ fig : wclog ] shows that decrease of @xmath147 let the size of the region for wave vector independent fluctuation decrease .
this suggests that , in order to observe the peculiar influences of the cytoskeleton on the fluctuation spectrum , the membrane with soft bending rigidity should be used . in order to observe @xmath106 behavior
, @xmath12 values should be relatively large .
in fact , the spring constant can be estimated in various ways . from the measured shear modulus @xmath150 , adopting the continuum model @xcite , we obtain the spring constant as @xmath151 @xcite , and correspondingly @xmath152 . if treating the spectrin tetramer as an ideal entropic spring , the spring constant is given by @xmath153 .
the persistence length of the spectrin @xmath154 @xcite and the contour length @xmath155 @xcite give @xmath156 and @xmath157 .
these spring constants are orders of magnitude larger than the fitting values and satisfy eq .
( [ conf ] ) .
we investigate fluctuation properties of red blood cell membranes through the gaussian model which takes account of the finite coupling strength between the bilayer and the cytoskeleton . focusing on the role of the coupling and discrete nature of the spectrin meshwork
, we obtain different types of fluctuations for the coupled membrane .
a membrane strongly coupled to rigid cytoskeleton presents non - monotonic fluctuations because of the membrane curvature deformation at length scale smaller than the cytoskeleton meshsize .
when soft cytoskeleton is coupled to the membrane , the system exhibits the well - studied fluctuation spectrum described by surface tension and the confining potential .
weakly coupled membranes have an extra crossover and present wave vector independent fluctuation at intermediate length scales .
the essential source of this diversity is the competition among the elastic energy of the spring network , the bilayer - cytoskeleton coupling , and the bending energy of the membrane at short length scales .
direct measurement of such diverse spectra must be a difficult task , depending on instrument resolution , sample availability , and control technique . yet , as shown in this work , the fluctuation spectrum is dependent on the ratio @xmath158 , and then controlling the bending modulus of red blood cell membranes may allow us to observe non - trivial spectrum like non - monotonic fluctuations for rigid cytoskeleton .
atp concentration can also change the elastic property of the cytoskeleton as well as the coupling strength of the anchoring proteins .
more direct evidences should be provided by such controlled experiments for better understanding the physical properties of rbcs . moreover ,
the nonequilibrium nature of rbc flickering in relation to active coupling between the bilayer and the cytoskeleton is also an important issue . in future studies
, we will pursue to analyze nonequilibrium spectrum in the presence of active coupling , based on our model with a certain type of time dependence introduced in the coupling strength of eq .
( [ v ] ) , which , we believe , is a minimal approach to grasp the essential characteristic of the system dynamics .
we thank y. park for helpful discussion and providing us unpublished results .
this research was supported by basic science research program through the national research foundation of korea(nrf ) funded by the ministry of education , science and technology(grant no .
nrf-2013r1a1a2013137 ) .
p. b. canham , j. theor . biol . * 26 * , 61 ( 1970 ) . h. j. deuling and w. helfrich , biophys .
j. * 16 * , 861 ( 1976 ) .
f. brochard and j. f. lennon , j. phys .
* 36 * , 1035 ( 1975 ) .
s. r. keller and r. skalak , j. fluid mech .
* 120 * , 27 ( 1982 ) .
g. lim h. w. , m. wortis , and r. mukhopadhyay , proc .
sci . u.s.a * 99 * , 16766 ( 2002 ) .
m. a. peterson , phys .
a * 45 * , 4116 ( 1992 ) .
r. mukhopadhyay , g. lim h. w. , and m. wortis , biophys .
j. * 82 * , 1756 ( 2002 ) .
y. park , et al .
sci . u.s.a * 107 * , 6731 ( 2010 ) .
j. paulose , g. a. vliegenthart , g. gompper , and d. r. nelson , proc .
sci . u.s.a * 109 * , 19551 ( 2012 ) .
j. li , m. dao , c. t. lim , and s. suresh , biophys .
j. * 88 * , 3707 ( 2005 ) .
s. b. rochal and v. l. lorman , phys .
* 96 * , 248102 ( 2006 ) .
t. auth , s. a. safran , and n. s. gov , phys .
e * 76 * , 051910 ( 2007 ) .
n. gov , a. g. zilman , and s. safran , phys .
lett . * 90 * , 228101 ( 2003 ) . j .-
b . fournier , d. lacoste , and e. raphal , phys .
lett . * 92 * , 018102 ( 2004 ) .
a. zilker , h. engelhardt , and e. sackmann , j. physique * 48 * , 2139 ( 1987 ) . c. dubus and j .- b .
fournier , europhys .
* 75 * , 181 ( 2006 ) .
y - z . yoon , et al .
, biophys .
j. * 97 * , 1606 ( 2009 ) .
g. popescu , et al .
lett . * 97 * , 219101 ( 2006 ) .
y. park , _ private communication .
_ s. manno , y. takakuwa , and n. mohandas , j. biol .
chem . * 280 * , 7581 ( 2005 ) .
t. betz , m. lenz , j - f .
joanny , and c. sykes , proc .
sci . u.s.a
* 106 * , 15320 ( 2009 ) .
n. gov and s. a. safran , phys .
e * 69 * , 011101 ( 2004 ) .
o. farago , phys .
e * 78 * , 051919 ( 2008 ) .
s. t. milner and s. a. safran , phys .
a * 36 * , 4371 ( 1987 ) .
u. seifert , z. phys .
b * 97 * , 299 ( 1995 ) . j. c. hansen , r. skalak , s. chien , and a. hoger , biophys . j. * 72 * , 2369 ( 1997 ) ; n. s. gov and s. a. safran , biophys .
j. * 88 * , 1859 ( 2005 ) ; j. li , g. lykotrafitis , m. dao , and s. suresh , proc .
sci . u.s.a * 104 * , 4937 ( 2007 ) .
m. tomishige , y. sako , and a. kusumi , j. cell biol .
* 142 * , 989 ( 1998 ) .
z. peng , et al .
usa * 110 * , 13356 ( 2013 ) .
if the fluctuations are not small as we assumed , there exists a coupling between the in - plane stretching and the bending . generally the gaussian approximation is no longer valid for large elastic deformations , and the higher order terms would be necessary .
merath and u. seifert , phys .
e * 73 * , 010401(r ) ( 2006 ) ; r .- j . merath and u. seifert , eur .
j. e * 23 * , 103 ( 2007 ) .
h. nochuchi and g. gompper , proc .
* 102 * , 14159 ( 2005 ) ; d. a. fedosov , b. caswell , and g. e. karniadakis , biophys .
j. * 98 * , 2215 ( 2010 ) .
d. boal , _ mechanics of the cell _ , 2nd ed .
( cambridge university press , ny , 2012 ) .
d. e. discher , d. h. boal , and s. k. boey , biophys .
j. * 75 * , 1584 ( 1998 ) .
t. byers and d. branton , proc .
usa * 82 * , 6153 ( 1985 ) . | we theoretically investigate the membrane fluctuations of red blood cells with focus laid on the role of the cytoskeleton , viewing the system as a membrane coupled to sparse spring network .
this model is exactly solvable and enables us to examine the coupling strength dependence of the membrane undulation .
we find that the coupling modifies the fluctuation spectrum at wavelengths longer than the mesh size of the network , while leaving the fluid - like behavior of the membrane intact at shorter wavelengths .
the fluctuation spectra can be markedly different , depending on not only the relative amplitude of the bilayer bending energy with respect to the cytoskeleton deformation energy but also the bilayer - cytoskelton coupling strength . |
much of the theoretical and experimental interest has been recently devoted to the study of the production of heavy quarks in hadronic collisions . on the theoretical side the calculation in perturbative qcd of the differential and total cross sections to order
@xmath6 has been performed by two groups @xcite , providing a firm basis for a detailed study of the properties of the bottom and charm quarks , and leading to reliable predictions for the production rate of the top quark @xcite .
these results do however present a non - negligible residual renormalization / factorization scale dependence , particularly at large @xmath0 .
furthermore , the validity of this next - to - leading ( nlo ) @xmath4 calculation is limited when @xmath7 , @xmath8 being the large quark mass , by the appearance of potentially large logarithms of the type @xmath2 , which have to be resummed to all orders .
the physical reason for that is quite clear .
for example , terms of order @xmath9 or @xmath10 are simply related to the mass singularities originating from collinear configurations when @xmath11 for fixed @xmath0 .
the theoretical uncertainty associated to those corrections has been roughly estimated in @xcite . whereas for top quark production this uncertainty is irrelevant , this is not the case for the production of bottom and charm quarks at large @xmath0 , leading to relevant phenomenological consequences . in the present paper
we consider a solution to this problem , namely we study the behaviour of the differential cross section for the production of heavy quarks in the limit @xmath12 , to nlo accuracy . to this aim
we follow an approach based on the properties of fragmentation of a generic parton @xmath13 in the heavy quark @xmath14 , after the parton has been produced inclusively in the hard collision of the two initial hadrons .
the basic formula is represented by eq .
( [ eq : fact ] ) , where the partonic cross sections @xmath15 at @xmath4 have been given in ref .
@xcite in the massless quark limit .
@xmath15 introduces an explicit dependence on @xmath0 and on renormalization / factorization mass scales . the dependence on the heavy quark mass
is then obtained through the fragmentation function of the parton @xmath16 , evolved at nlo accuracy from an initial scale @xmath17 ( see below ) to @xmath18 .
this approach explicitly resums potentially large terms ot the kind @xmath19^n$ ] , giving a better description of the theoretical predictions at large @xmath0 .
indeed the corresponding uncertainty is quite reduced in this region with respect to the perturbative result , due to a significantly smaller sensitivity to the relevant scales . on the other hand , because of the massless limit used for the @xmath4 kernel cross sections @xmath15 , this approach does not allow to recover in a simple way the limit @xmath20 of the perturbative calculation . in this paper
we restrict our approach to @xmath3-quark production , the application to charm production will be given elsewhere . in sect .
[ sect2 ] we give the general formalism . in sect .
[ sect3 ] we first discuss the basic ingredients necessary to obtain our nlo cross sections and then we present our results .
we compare with the perturbative calculation of ref .
@xcite and study in detail the dependence on the various mass scales entering the problem .
our conclusions are finally given in sect .
[ sect4 ] .
according to factorization theorems @xcite the cross section for the inclusive hadroproduction of a hadron at high transverse momentum , i.e. for the process @xmath21 can be written as = _ i , j , kx_1
x_2 x_3 f_h_1^i(x_1,_f ) f_h_2^j(x_2,_f ) _
ijkx(x_1,x_2,_r,_f ) d_k^h_3(x_3,_f ) [ eq : fact ] as usual , the @xmath22 s are the distribution functions of the partons in the colliding hadrons , @xmath23 is the kernel cross section and @xmath24 is the fragmentation function of the observed hadron . the factorization mass scales @xmath25 of the structure and fragmentation functions are assumed to be equal for the sake of simplicity .
@xmath26 is the renormalization scale . due to the presence of collinear singularities both in the initial and final state
this process is not fully predictable by qcd itself .
we can actually calculate the kernel cross section and the evolution of the structure and fragmentation functions , but we have to rely on some phenomenological input to obtain the latter at some given initial scale .
this situation changes drastically when we come to consider the inclusive production of a heavy quark . in this case
its mass , being finite and considerably greater than @xmath27 , makes the perturbative expansion feasible and prevents collinear singularities from appearing in the splitting vertices which involve the heavy quark . having this in mind
two approaches can be pursued in the calculation of heavy quark production .
the first one is to directly calculate in perturbation theory the process @xmath28 , q being the heavy quark and @xmath29 the initial state _ light
_ partons ( i.e. light quarks and gluons ) .
this kernel cross section will then be convoluted with initial state structure functions only , the final state showing no singularities of any kind .
this approach has been followed in the past @xcite , providing a full perturbative @xmath30 ) calculation . in the following we shall use for comparisons the results of nason , dawson and ellis , and refer to them as nde . in this perturbative approach ,
as stated earlier , terms of the kind @xmath31 will appear .
they are remnant of the collinear singularity screened by the finite quark mass . as noted in ref .
@xcite they can grow quite large at high tranverse momenta , thereby spoiling the validity of the expansion in @xmath32 . therefore they have to be summed to all orders . the alternative way is to consider that when a quark , of whichever flavour , is produced at very high transverse momentum @xmath7 its mass plays almost no role at all in the scattering process .
this is to say that mass effects in the kernel cross section are suppressed as power ratios of mass over the scale of the process .
we can therefore devise a picture in which all quarks are produced in a massless fashion at the high scale @xmath33 and only successively , as their virtuality decreases , they can _ fragment _ into a massive heavy quark .
the cross section can therefore be described by a formula analogous to eq .
( [ eq : fact ] ) , with @xmath34 .
the key difference to the hadron production case considered in eq .
( [ eq : fact ] ) is that initial state conditions for the heavy quark fragmentation functions are now calculable from first principles in qcd ( hence the definition of `` perturbative '' fragmentation functions , pff ) and do not have to be taken from experiment .
actually , the following set of next - to - leading initial state conditions can be obtained @xcite in the @xmath35 scheme for the fragmentation function of a heavy quark , gluon and light quark respectively , in the heavy quark @xmath14 & & d_q^q(x,_0 ) = ( 1-x ) + ( _ 0 ) c_f_+ [ dqq ] + & & d_g^q(x,_0 ) = ( _ 0 ) t_f(x^2 + ( 1-x)^2 ) [ dgq ] + & & d_q,|q,|q^q(x,_0 ) = 0 [ dqq ] where @xmath36 must be taken of the order of the heavy quark mass .
then using the usual altarelli - parisi evolution equations at nlo accuracy one finds the fragmentation functions set at any desired factorization scale @xmath25
. an important feature of this approach can now be appreciated .
the `` almost - singular '' logarithmic term @xmath37 splits into two , as follows . a @xmath38 will be found in the kernel cross section @xmath23 which has no dependence on the heavy quark mass , according to the assumption that it is produced in a _ massless _ way .
moreover , by choosing @xmath39 it will not contain large logarithms and its perturbative expansion will behave correctly .
the remaining part of the log will instead be lurked into the fragmentation function @xmath40 .
the large @xmath41 is resummed to all orders by the evolution equations , and only the small @xmath42 provided by the initial state condition is treated at fixed order in perturbation theory .
therefore one expects a better control of the theoretical uncertainty at large @xmath0 . on the other hand , for @xmath43 the fragmentation approach does not allow to recover easily the perturbative result , which , of course , holds exactly .
in order to implement the `` perturbative fragmentation function ( pff ) approach '' at a numerical level we need four ingredients , which are all available at the next - to - leading level : * the distribution functions of any parton ( including the heavy flavour in question ) in the hadrons ( proton or antiproton ) , evolved at nlo accuracy .
all modern sets satisfy this requirement .
an important point must however be made clear .
a heavy quark present in the initial state can be directly brought to the final state where it is fragmented to the detected heavy flavour through the @xmath44 , and therefore with high probability ( see below ) .
this means that the resulting cross section is particularly sensitive to the overall heavy flavour content of the colliding hadrons . in the parton distribution functions ( pdf )
sets available this content is generated through perturbative gluon splitting above a given threshold .
the total yield will therefore depend on the choice made .
for instance , the hmrs set @xcite takes @xmath45 as initial condition , whereas the mt @xcite and the cteq @xcite ones choose , according to ref .
@xcite , @xmath46 .
the resulting different bottom distributions are plotted in figure [ bpdf ] : the mt and cteq bottom contents are consistently higher than the hmrs one , the discrepancy growing larger as the hmrs threshold value is approached .
* the kernel cross section for the scattering of any two massless partons into another massless parton .
this calculation is provided at the nlo in various renormalization / factorization schemes in ref .
the massless limit approximation used in these @xmath4 kernel cross sections will not allow to evaluate the heavy quark cross section for @xmath20 . *
the next - to - leading expression for the strong running coupling : @xmath47 \nonumber\\ \vbox{\vspace{.4 cm } } & & b = \frac{33 - 2n_f}{12 \pi } , \ ; \ ; b'= \frac{153 -19 n_f}{24 \pi^2}\nonumber\end{aligned}\ ] ] the qcd scale @xmath27 is set at the value fixed by the structure function set used and the number of active flavours @xmath48 is set at 5 above the appropriate threshold . *
the fragmentation functions of any parton into the heavy flavour .
they are obtained by evolving the initial conditions given above ( eqs .
[ dqq],[dgq],[dqq ] ) with nlo accuracy @xcite .
this is done through numerical inversion of the mellin moments of the evolved distributions , and a typical result for bottom fragmentation is depicted in figure [ bfrag ] , together with the uncertainty given by the choice of the initial condition scale @xmath36 around @xmath49 .
the implication of these ideas to lep hadronic data is discussed in ref .
@xcite . with these four ingredients at our disposal we can now evaluate the cross section for the high @xmath0 inclusive hadroproduction of a bottom quark .
this will be done by performing numerically the triple integration of eq .
( [ eq : fact ] ) .
it must be noted that this integration is not as straightforward as the one usually performed when calculating hadron distributions .
indeed while the hadronic fragmentation functions fall smoothly to zero as @xmath50 approaches one , now the @xmath44 is instead singular in the @xmath51 limit , due to unresummed sudakov terms .
this makes the most usual integration routines to fail in calculating both the integral and the associated numerical error , the latter being heavily underestimated .
more in detail this problem has been overcome by suitably rearranging the integration in such a way that the numerical integration could be reliably performed . by rewriting eq .
( [ eq : fact ] ) in a more compact form , and dropping all unnecessary symbols , we have & = & ^1_x_3^minx_3x_1 x_2 _ ijk f_i f_j _
ijk(x_3 ) d_k(x_3 ) = + & = & ^1_x_3^minx_3x_1 x_2 _
ijk(x_3 ) d_k(x_3 ) + + & + & ^1-_x_3^minx_3x_1 x_2 _ ij f_i f_j _ ijq(x_3 ) d_q(x_3 ) + + & + & _ 1-^1x_3
x_3 d_q(x_3 ) + + & + & x_1 x_2 _ ij f_i f_j _
ijq(1)_1-^1x_3 x_3d_q(x_3 ) the last integral over @xmath52 is then evaluated by resorting to the mellin moments ] ] of the evolved fragmentation function : _
1-^1x_3 x_3d_q(x_3 ) = d_q(2 ) - _ 0 ^ 1-x_3 x_3d_q(x_3 ) [ coeff ] the result should of course be independent of the choice of @xmath53 , and we have verifyed that this is indeed the case for @xmath54 . figure [ 1800 ] shows a comparison of the perturbative result of ref .
@xcite with our calculations for two different pdf sets , at the fermilab energy of 1800 gev , for @xmath55 .
we can see that in the high @xmath0 region the perturbative cross section is quite sensitive to the structure function set choice , the mt one giving a markedly lower result .
the opposite happens in the pff approach , which becomes very sensitive in the low @xmath0 region which , as stated above , is plagued by the neglect of the heavy quark mass terms in the partonic cross sections .
this pattern is due to the fact that the nde calculation only uses the light quark content from the structure functions , since the heavy one is produced at the purely perturbative level .
this introduces a mismatch ( albeit of higher order ) as the evolution had been performed by taking into account all flavours , which get mixed through gluon exchange .
our calculation , on the other hand , treats all quarks on the same ground and uses the full information of the pdf set .
this leads to substantially identical predictions obtained in the high @xmath0 region with the two pdf sets . on the contrary , at low @xmath0
the different threashold behaviour of the two sets in the heavy quark content is responsible for the discrepancy between the two results .
figure [ 630]a , b show the same kind of comparison at 630 gev and 16 tev .
the behaviour is practically identical , once a rescaling on the @xmath0 axis has been considered to account for the different cm energies .
one further comparison can also be made between the mt - b2 and the cteq1 m sets , which have a similar threshold behaviour for the heavy quark flavour .
as can be seen in figure [ 1800cteq]a , b the theoretical uncertainty due to the structure functions set choice is once again smaller in the fragmentation function approach .
next we consider the dependence on the choice of the renormalization / factorization mass scale @xmath56 .
figure [ 1800sc_mt]a(b ) shows , at 1800 gev and with the mt - b2 ( hmrs - b ) set , the theoretical uncertainty resulting from the variation of the factorization and renormalization mass scales between @xmath57 and @xmath58 , where @xmath59 is defined as @xmath60 and we have taken @xmath61 . as expected the band of the perturbative nde calculation is sensibly larger than ours , showing the improvement brought by the resummation of the large logarithms of @xmath62 .
the drop at low @xmath0 in the prediction of the hmrs - b set with @xmath63 , which considerably broadens our band in that region , is due to the very poor heavy quark content of this set at such a low scale ( see fig .
[ bpdf ] ) .
all these features can be better appreciated in figure [ scale ] , where the cross section at 1800 gev with the mt - b2 set is plotted , at fixed @xmath64 and @xmath0 , as a function of @xmath65 , for @xmath66 varying between 0.25 and 4 .
this figure also shows a comparison with the factorized calculation with a born ( i.e. lo ) cross section kernel ( but with two - loop @xmath32 and nlo structure and fragmentation functions ) .
as expected , the lack of the next - to - leading terms strongly enhances the scale dependence .
the similarity between the nde result and the fragmentation function approach with born kernel cross section is striking .
the small scale sensitivity of our full nlo calculation shows that the fragmentation / renormalization scale dependence is a real @xmath67 effect , whereas in the perturbative approach ( e.g. nde ) the presence of large @xmath2 results in an effective @xmath4 dependence . once again , the results at 630 gev greately mimick those at 1800 gev ( see figure [ 630sc_mt ] ) .
one further theoretical uncertainty must be considered , namely the one related to the freedom in the choice of the initial condition scale @xmath36 in the fragmentation functions evolution .
figure [ 1800mu0 ] shows that the cross section is marginally affected by a change of @xmath36 in the range @xmath68 , particularly in the high @xmath0 region .
this leads to the conclusion that the pff approach has still a smaller overall theoretical uncertainty also after this qualitatively new feature has been taken into account .
finally , we show in figs .
[ channels]a , b the relative importance of the various partonic subprocesses contributing to the cross section for two sets of pdfs . in the case of the mt - b2 set (
similar results are obtained with the cteq1 m one ) the @xmath69 scattering clearly dominates al low @xmath0 , whereas at large @xmath0 it becomes comparable to other channels . comparing with the hmrs - b set we see that the different threshold behaviour of the heavy quark structure function in this latter set , repetedly pointed out before , is responsible for the lower contribution of the @xmath69 and @xmath70 channels .
we can now turn to quantities of more direct experimental interest and see how the advantages of our approach are reflected onto them .
namely , we will consider the total cross section for one particle inclusive heavy quark production , integrated above a given @xmath71 and within a rapidity region @xmath72 . only the variation of the theoretical prediction due to changes in the factorization / renormalization scale and in the pdf set used will be studied .
other possible sources of uncertainties , aside the change of @xmath36 which has been shown to be almost neglibible , are the value of the qcd scale @xmath27 and of the bottom mass @xmath49 .
they should however be common to both the perturbative and the fragmentation function approach , and have been studied in detail in ref .
@xcite .
figures [ 1800_hmrs_tot ] , [ 1800_mt_tot ] and [ 1800_cteq_tot ] show the total cross section @xmath73 at a cm energy of 1800 gev .
the hmrs - b , mt - b2 and cteq1 m sets have been used respectively , and the curves obtained with @xmath74 and @xmath63 are plotted .
we can see that the features of the differential cross sections previously discussed translate almost directly into the total cross section .
in particular , the scale dependence of our result is once again remarkably smaller than the one of the perturbative calculation , especially in the high @xmath0 region .
the broadening of the band in the low @xmath0 region than can be observed when the hmrs - b or the cteq1 m set is used with the scale choice @xmath63 is once again due to the different heavy quark content at low scales ( see fig .
[ bpdf ] ) .
the overall smaller theoretical uncertainty of the pff result can finally be appreciated in fig .
[ final ] , where the highest and lowest predictions of the two approaches , out of the six curves previously considered , have been plotted .
note that we have not considered in detail the uncertainties for @xmath75 gev .
the cdf experimental data @xcite are also shown .
the same kind of result can also be produced at 630 gev .
figure [ final630 ] compares the band of nde to our one and to experimental data by ua1 @xcite .
also in this case we find a sizeable reduction in the uncertainty of the theoretical prediction .
in this paper we have discussed the production of heavy quarks at large @xmath0 ( @xmath7 ) in hadronic collisions , in order to make the theoretical predictions for the differential distributions more reliable .
we have studied this problem in the framework of the perturbative fragmentation functions , which allow a nlo evaluation of the potentially large logarithms of the kind @xmath2 , which are resummed to all orders .
our analysis for the @xmath3-quark leads to much more stable results with respect to changes of the factorization / renormalization scales compared to what is obtained in the @xmath4 perturbative calculation . also the theoretical uncertainty related to different choices of pdf sets is reduced .
other possible sources of uncertainties , like the scale @xmath36 of the initial state condition in the fragmentation functions evolutions , are negligible . a detailed discussion of the differential and integrated cross sections of experimental interest has also been presented
. 99 # 1#2#3nucl . phys . * b#1 * ( # 2 ) # 3 # 1#2#3phys .
b#1 * ( # 2 ) # 3 # 1#2#3 phys .
rev . lett . *
# 1 * ( # 2 ) # 3 # 1#2#3phys . rev .
* # 1 * ( # 2 ) # 3 # 1#2#3phys . rev .
* d#1 * ( # 2 ) # 3 # 1#2#3z . phys . *
c#1 * ( # 2 ) # 3 # 1#2#3comm .
* # 1 * ( # 2 ) # 3 p. nason , s. dawson and r.k .
ellis , nucl .
* b303 * ( 1988 ) 607 p. nason , s. dawson and r.k .
ellis , nucl .
b327 * ( 1989 ) 49 w. beenakker , h. kuijf , w.l .
van neerven and j. smith , phys . rev .
* d40 * ( 1989 ) 54 w. beenakker , w.l .
van neerven , r. meng , g.a . schuler and j. smith ,
* b351 * ( 1991 ) 507 g. altarelli , m. diemoz , g. martinelli and p. nason , e. laenen , j. smith and w.l .
van neerven , fermilab - pub-93/270-t , itp - sb-93 - 55 , thu-93/23 f. aversa , p. chiappetta , m. greco and j.ph .
guillet , nucl . phys .
* b327 * ( 1991 ) 105 see , e.g. , j.c .
collins , d.e . soper and g. sterman , in
_ perturbative qcd _
( singapore , world scientific , 1989 ) , p. 1 .
b. mele and p. nason , nucl . phys . *
b361 * ( 1991 ) 626 p.n .
harriman , a.d .
martin , w.j . stirling and r.g .
roberts , phys . rev .
* d42 * ( 1990 ) 798 j.g .
morfin and w.k .
tung , zeit .
c52 * ( 1991 ) 13 j. botts , j.g .
morfin , j.f .
owens , j. qiu , w.k .
tung and h. weerts , phys .
b304 * ( 1993 ) 159 j.c .
collins and w.k .
tung , nucl .
b278 * ( 1986 ) 934 g. curci , w. furmanski and r. petronzio , + w. furmanski and r. petronzio , b. mele and p. nason , + g. colangelo and p. nason , f. abe _
( cdf coll . ) , + s. vejcik ( cdf ) , presented at dpf meet .
soc . , batavia ( 1992 ) , to appear in the proceedings + b.t .
huffman ( cdf ) , presented at dpf meet .
soc . , batavia ( 1992 ) , fnal - conf-92/337-e ( 1992 ) , to appear in the proceedings + see also f. abe _ et al . _
( cdf coll . ) , and references therein c. albajar _ et al . _
( ua1 coll . ) , ; | the production of heavy quarks at large @xmath0 ( @xmath1 ) in hadronic collisions is considered .
the analysis is carried out in the framework of perturbative fragmentation functions , thereby allowing a resummation at the nlo level of final state large mass logarithms of the kind @xmath2 .
the case of @xmath3-quark production is considered in detail .
the resulting theoretical uncertainty from factorization / renormalization scales at large @xmath0 is found to be much smaller than that shown by the full @xmath4 perturbative calculation .
[ cols= " < " , ] large @xmath5 hadroproduction + of heavy quarks + + + ' '' '' e - mail addresses : = [email protected] + [email protected] |
various pieces of evidence support the idea that the x - ray emission from active galactic nuclei ( agn ; see * ? ? ?
* ; * ? ? ?
* ) and stellar - mass black - holes @xcite originates from an accretion disc and the surrounding corona .
these are thought to be situated near a central black hole , no more than a few tens gravitational radii from the event horizon , giving rise to the relativistic effects in 67 kev iron line complex and the underlying continuum .
the standard scheme of accretion discs @xcite captures the main properties of accreting black holes surprisingly well , nevertheless , the model omits some important aspects .
in particular , the radial profile of the source intrinsic emissivity is represented by a smooth function ( decaying as a power law @xmath2 at large distance ) , while the realistic profile is likely to be more complicated .
different approaches have been pursued in order to understand how accretion disc x - ray spectra are formed .
generally , these include the investigations of accretion disc instabilities as well as the interpretation of spectra to constrain the model parameters ( such as the black hole spin , the source orientation , and the location and size of the accretion disc ) , and to determine the radial profile of accretion disc emissivity @xcite .
recently , @xcite discussed an interesting approach to the inversion problem of determining the radial emissivity bulk profile of the relativistic broad iron line in seyfert galaxy 1h 0707 - 495 .
here we turn our attention to additional features superposed on top of the broad line .
the form of the inner accretion flow remains an open question . according to the standard scheme the flow proceeds down to the inner edge at the innermost stable circular orbit ( isco ) , i.e. , @xmath3 for schwarzschild black hole , and @xmath4 for an extreme kerr black hole ( gravitational radius @xmath5 with @xmath6 being the mass of the supermassive black hole in units of @xmath7 ) .
there are uncertainties about this assumption .
it has been argued that because of magnetic stresses the radiation edge of the emission may be somewhat off the isco , and its exact location may be different for the continuum and for the iron line @xcite .
also , the inner rim recedes further out from the black hole when the accretion flow ceases and the disc becomes truncated , as was demonstrated in several objects ( * ? ? ? * ; * ? ? ?
* and references cited therein ) . intermittent episodes of a localized disc irradiation can naturally lead to a radially stratified emission profile rather than monotonic , continuous dependence of a standard accretion disc .
we can approximate this configuration by radially constrained zones , which can be called `` rings '' .
let us remark that we are concerned with a spectral line emissivity , which is only partly related to the gas density ; the essential quantity here is the ionization state of matter and how this varies with radius .
localized coronal irradiation of the disc material enhances the line emission above the mean value in the neighbourhood of a certain point .
integration of detected signal over a period of time then effectively produces a ring - type source @xcite .
in other words , one need not imagine physically separate rings and gaps emerging within the disc , even if this possibility has been also discussed ; e.g.@xcite show a temporary density ring in an accretion disc surrounding a black - hole binary , while @xcite and @xcite examine the process of gap formation by a secondary satellite embedded within the disc .
the excess emission from a certain radius could be revealed by high - resolution spectroscopy of the broad iron line @xcite , although at present it appears to be rather unfeasible because of insufficient signal - to - noise in the available data . in this paper
we discuss whether the proposed large observatory for x - ray timing ( loft ; @xcite @xcite , @xcite ) will have the necessary capability , at least for bright enough sources . to this end
we produce artificial data with appropriate properties and then we analyze them by using a preliminary response file .
according to the standard ( stationary ) accretion disc scenario , most radiation is produced within just a few gravitational radii above isco ( up to @xmath8 ) . also various x - ray spectral features and the power - law component of the disc corona
are believed to originate in that region .
this main flux - producing area of the accretion disc is still rather wide in radius , hence it is thought to be the origin of the broad component of the line , which we include by employing the relativistic kyrline model @xcite .
some reports suggest transient structures that are localized in radius and exhibit themselves as narrow features in the spectrum ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* and references cited therein ) . in our picture of time - integrated spectra these are represented by rings .
although the significance of these features is still debated , their origin fits in the scenario of magnetic flares as sites of primary local illumination .
magnetic flares are thought to occur above the accretion disc @xcite ; in this context the line emission of the rings would arise not because the accretion disc develops such a physical structure , but instead the localized emission by reconnection flares illuminates the underlying disc at a certain place , affecting there its ionization and producing the observable reflection features in spectrum .
rotation of the disc matter and time integration of the signal during observation produce belts out of point - like sites of illumination .
in fact , we could then call these structures as reflection rings .
although the mentioned scheme is our preferred way of effectively producing the rings , similar effects can arise by different mechanisms operating near a black hole . they can be broadly distinguished in two categories .
firstly , during the period of intermittent accretion the flow of material varies and the density profile and other variable ( ionization state profile ) can change along ( e.g. * ? ? ?
* ; * ? ? ?
* ) . in another context
, global spiral waves were suggested as a possible source of light curve modulation of accreting black holes @xcite . unlike ideal rings these spirals structures have a non - neglibible radial extent , although to certain extent they could mimic rings in case of significant wounding in azimuthal direction @xcite .
the formation of separate annuli occurs also in some models of strongly magnetized plasma discs @xcite . finally , confined radii of enhance emission are a consequence of avalanche models ( pechek et al . , in preparation ) where multiple flares arise in mutually connected families and gradually propagate towards the center before decaying at some characteristic distance .
in contrast to the model of major flares mentioned formerly , the latter alternatives appear to be unable of modulating the outgoing spectrum in any significant manner for a required period of time @xcite .
nonetheless , the current modelling has not yet provided a definitive answer and different options should be investigated further .
in fact , the topic for debates is the expected level of the electromagnetic signal modulation above the noise and the ability to detect it by a given technology , rather than the existence of the modulation at a certain small level . aside from the mentioned speculations ,
one may ask if the actual presence of the rings can be tested . here
we thus presume that the structure develops and modulates the outgoing signal , and we want to verify whether the location of emission rings could be reconstructed from the observed spectrum . by _ assuming the existence _ of the modulation we investigate if these structures can be spectrally resolved in present - day or future x - ray spectra .
+ + to determine the relativistic shift ( doppler and gravitational ) of the photon energy , we adopt the method described in our recent paper @xcite , where we considered propagation of photons from the source in the limit of geometrical optics in kerr metric @xcite .
the extremal values of energy shifts , @xmath9 and @xmath10 , are of particular interest , as they determine the range of line energy observed from distance .
the main parameters are the emission radius @xmath11 , the dimensionless spin parameter @xmath12 ( @xmath13 ) , and the inclination angle @xmath14 ( @xmath15 stands for the rotation axis , while @xmath16 is for the edge - on view of the accretion disc ) . extending our previous work ,
we allow also the case of retrograde rotation ( @xmath17 ) because such a possibility is not excluded a priori , and it may lead naturally to rings that emerge at somewhat larger radii than in the prograde case ( the isco recedes as counter - rotation of the black hole accelerates toward @xmath18 ; * ? ? ?
( gradually growing with the spin and inclination angle ) can be attributed to the way how frame - dragging effects had been treated in the original version of the relline code .
it turns out that this difference has been meanwhile eliminated from the latest version of the code , so the results are now in full agreement with the kyrline @xcite version for retrograde spin .
naturally , in the latter case , isco can recede as far as @xmath19 , and so the relativistic effects become quite small compared to discs rotating in the prograde direction . ]
a simple prescription for the intrinsic emission of light allows us to explore the observed spectral features across the parameter space .
we assume a power - law continuum ( representing the primary x - rays originating from corona ) plus a spectral line ( representing the k@xmath20 emission line of iron at @xmath21 kev rest energy ) .
the line is intrinsically narrow ( in the local co - moving frame ) of the line - producing ring , although it becomes subsequently broadened by relativistic effects .
the line is produced in a range of radii over the inner disc , so the rings are just those radii where the line production is enhanced above the baseline model of the power - law component and the broad line .
the mentioned components of the model spectrum give us an opportunity to test the procedure of reconstructing the source emissivity .
light rays propagate along null geodesics in the curved spacetime , which brings significant energy shifts to the final spectrum and it spreads the observed profile over the whole @xmath22 interval with respect to the rest energy of the line .
figure [ fig : g ] presents the extremal energy shifts , @xmath9 vs.@xmath10 . these determine the expected energy of the local peaks of the composite broad line , depending on the radial profile of emissivity from the disc .
one can see a complex interplay between gravitational effects , doppler shift and light - bending .
the overall gravitational redshift dominates near the horizon ( although we restrict the emission radii to @xmath23 ) , while doppler broadens the line ( i.e. it increases @xmath24 ratio ) and it enhances the height of blue wing , progressively more for large inclinations .
the focusing effect further enhances the observed flux , especially at large @xmath14 .
[ fig : g ] provides several examples including both prograde and retrograde rotation , although we can not show here all combinations of the parameters .
more examples of the prograde case were examined in @xcite ; in particular ; in that work we show also the case of small inclination angles , which are relevant for type-1 agns but are more difficult to plot because the energy shifts are much smaller too . from the figure
one can also realize that the parameter degeneracy of the model can be expected where the lines converge to each other , quite independent of the parameter values ( we will come back to this issue in more detail later in sec .
[ sec : discussion ] ) . of accretion rings . clearly , @xmath25 and @xmath26 cases contain some unmodelled features which allow us to reject these models .
[ fig : residua ] ] it is interesting to compare our results with a similar scenario for the geometrical and kinematical properties of the reflection model by @xcite , who also based their discussion on the extreme frequency shifts of a spectral line , as determined by measured radiation fluxes in the iron line wings . in particular , they show contour maps of the extremal redshift in two limiting cases a non - rotating ( @xmath27 , schwarzschild ) and the maximally co - rotating ( @xmath28 , kerr ) black hole .
our graphs are given while keeping the view angle @xmath14 fixed , showing the expected energy of spectral - line wings as the black hole spin is varied ( even to negative values ) .
we note that there is some unavoidable degeneracy among the model parameters .
however , we will show that this degeneracy can be avoided in situations where the accretion disc emission is dominated by contributions from a small number of narrow rings located at well - defined radii .
because the gravitational redshift becomes increasingly important as the spin increases and the radius of the ring goes to @xmath29 , the rings could be potentially revealed as features on the wing of the underlying relativistically broadened line .
the assumed source of reflection spectrum is a set of relatively narrow accretion rings or belts ( @xmath30 ) representing the emission excess above the standard accretion disc spectral line around a rotating ( kerr ) black hole .
thanks to the large effective area of the proposed detector ( designed to reach @xmath31 , i.e. about two orders of magnitude greater than that of xmm - newton near the iron line rest energy ) , as well as a sufficient energy resolution about @xmath32@xmath33 ev , the accretion rings should be visible when setting realistic values of the model parameters in our test spectra .
for the modelling purposes , as the worst case we assume @xmath34 ev energy resolution . in order to demonstrate the feasibility of the mentioned scenario we use a preliminary response matrix as an example of presumed capability of a future large - collecting - area device .
this will help us to assess the performance of xmm - newton versus loft and our ability of constraining the model .
first , we produced the simulated spectrum by assuming the source flux of approximately 1.3 mcrab ( @xmath35 in the energy range @xmath36@xmath37 kev ) , i.e.typical of a nearby bright seyfert galaxy , such as mcg6 - 30 - 15 .
we assumed a photo - absorbed power law continuum ( photon index @xmath38 , @xmath39 ) on top of which the three rings produce a relativistically broadened spectral line ( rest energy @xmath40 kev ) .
we assumed the exposure time of @xmath41ks for the loft mission and compared the model spectrum with the same set - up for @xmath42ks of xmm - newton spectrum .
it is important to realize that the huge collecting area of loft will allow us to constrain the parameters on a significantly shorter observation time , roughly comparable with the orbital time at the corresponding radius .
therefore , the required exposure does not much exceed reasonable duration of the flares , whereas for xmm - newton it had to be significantly longer .
the configuration of three rings is a reasonable ( albeit somewhat arbitrary ) trial assumption , where we can test the case of several , but not too many , radially confined excesses in the disc plane .
also the assumed duration of the observation is somewhat arbitrarilly set to a value which appears to be reasonably long and realistically achievable at the same time .
we set the ring widths to be initially @xmath43 , i.e. comparable with the gravitational radius as a typical length - scale of the system .
other relevant model parameters were set to : @xmath44 ( rapidly spinning black hole in prograde rotation ) , @xmath45 ( moderate inclination typical of a seyfert 1 nucleus ) as the initializing values .
our fiducial model is therefore phabs*(powerlaw+4*kyrline ) , i.e. a photo - absorbed power - law continuum and four line components blurred by relativistic effects ( we used xspec v. 12.6.0 ) .
one of the kyrline components originates over the entire disc surface and it has been fixed to its defaults parameters ( @xmath46 , radial emissivity index @xmath47 ) ; it provides the main portion of the total line flux ( we set about 60% for definiteness ) .
the other components represent the emission excesses from three ring - type regions , each one giving a smaller fraction ( @xmath481015% ) of the line flux .
naturally , diminishing the line flux produced by the rings relative to the mean flux of the entire disc and the powerlaw component makes the determination of @xmath49 , @xmath50 harder and eventually impossible .
we think , however , that the adopted level of the excess emissivity is realistically possible . figure [ fig : ringspectrum ] shows the data to model ratio for the fiducial spectrum .
the geometrical location and other parameters of the model components will be now submitted to a standard rigorous fitting procedure .
one expects that a single ring would be easier to recognize in the observed spectrum , while with the growing number of rings and diminishing separation among them the total signal should resemble that of a spectral line smeared over the accretion disc .
can one estimate the model parameters directly , by locating the energy of the peaks in the spectrum ?
cccccclcc & & & & & + ' '' '' & & & @xmath51 & @xmath52 & & @xmath51 & @xmath52 + 1 & 0.36 & 0.81 & 3.1 & 2.8 & & 3.7 & 3.4 + 2 & 0.48 & 0.91 & 4.1 & 3.9 & & 4.9 & 4.7 + 3 & 0.59 & 0.98 & 5.8 & 5.6 & & 7.1 & 6.9 + there is a partial degeneracy of the parameter values . in our case
this exhibits itself by the fact that , in order to obtain the red peaks of the line in right positions , the spin has to be greater than the lower limit of @xmath51 , however , the upper bound remains undetermined ( up to @xmath53 ) . for @xmath54 , i.e.up to the maximum spin of the kerr black hole
, we can reproduce the peaks by rearranging the ring radii .
this is shown in the table by giving two possible values of @xmath55 and @xmath56 that are consistent simultaneously with the mention minimum and maximum values of spin .
one can see that the uncertainty in the inferred radii is below 10% , while for spin the relative error represents about 25% .
although the `` visual '' approach to infer the peaks of the line profile constrains the model quite well , it does not allow us to determine the errors and to perform the formal statistical confidence contours analysis , so that apparently accurate values of @xmath49 and @xmath50 can not be assigned a proper meaning .
the spectral fitting should result in a more reliable determination of the parameters because it employs the whole spectral shape .
this appears to be important especially in the situation when the large effective area allows to resolve the broad line . on the other hand
, the spectral fitting tends to be sensitive to the assumed spectral model . in order to clarify the situation we carry out the following test .
we subjected the fiducial model to the standard xspec fitting procedure with the aim of recovering the initial model parameters including their confidence contours .
in particular , we tested if the assumed number of rings can be recovered with a significant confidence . in figure
[ fig : residua ] we show the best - fit residuals with respect to the continuum model , i.e. , the plot was obtained by removing the kyrline component from the best - fit spectra ( we show unbinned data because any definitive information about the future detector the quoted energy resolution is still preliminary at present ) .
the number of rings was changed from the original @xmath57 to @xmath26 and @xmath25 .
as free parameters we re - fitted the spin @xmath12 , inclination @xmath14 , and the radial width of rings , @xmath58 .
the rest energy of the line was fixed ( it was allowed to vary in subsequent tests ) .
the position of the rings can be also allowed to change , although the procedure can not start too far from correct values because of local minima in the @xmath59 space . by setting the number of rings to the fiducial number
we recover correctly the initial values of the parameters as expected , while different numbers of rings produce worse fits and the parameters converging off the right values ( or not converging at all ) .
we examined also the case of the line emission spread over the entire disc surface ( this represents the standard disc - line model ) , but the resulting fit was bad . in order to compare the results expected from loft with those that can be reached with currently available data , we carried out the same procedure as described above also with the xmm - newton response file of the epic - pn camera .
figure [ fig : contours ] demonstrates the expected accuracy with which the model parameters are constrained .
we plot the confidence contours for the inner ring radius versus the black hole spin . with the same exposure and the source brightness for both instruments
, loft will allow us to set much tighter constraints on the best - fit values . in figure
[ fig : chi2 ] we show the resulting best - fit @xmath60 values . decreasing the number of rings from the right value @xmath57 obviously degrades goodness of the fit . for xmm - newton
the best - fit @xmath60 values are lower , meaning that the model parameters are constrained at deteriorated confidence compared to loft .
both @xmath25 and @xmath26 tests give ( wrongly ) acceptable fits for xmm - newton , while they can be clearly rejected with loft .
naturally , @xmath57 gives a good fit because our artificial data were created with this value , so we only reproduce the original input . increasing the number of rings above @xmath57 produces formally acceptable statistics , but the number of model parameters then exceeds the necessary minimum .
this indicates an encouraging improvement that could be achieved for this particular problem with the improved sensitivity of loft .
parameter constraints could be improved also in our test case of xmm - newton , if the observation duration can be made longer and the number of collected photons correspondingly larger . because in practice one can not prolong the duration above some reasonable limit
, the increase of the collecting area of the loft detector is the most important aspect which improves the parameter constraints .
the energy resolution of the detector is a less critical factor because the effect of relativistic broadening causes the line width to be much broader than this limit by a significant margin .
we remark that the special choice of @xmath57 was needed for illustration purposes described above , but it does not mean any crucial limitation that would be essential for the described idea or the method .
naturally , a lower number of rings is easier to be recognized in the spectrum and its parameters to be reconstructed correctly . on the other hand , in the case of large @xmath61
the magnitude of the resulting peaks above the baseline profile becomes relatively small and the situation turns into the problem discussed by @xcite , i.e. the disc - line emission imagined as consisting of infinitesimal rings extending all the way from the inner to the outer rim of the disc .
we asked ourselves whether the energy position of the observed peaks determines the system parameters @xmath62 , @xmath14 , and @xmath12 in any unique way , or if instead the proposed procedure of reconstructing the parameter values converges to several different results , depending on the initial guess of the parameter values .
in fact , the latter is true .
therefore we need to set addition constraints in order to help xspec to find the right answer .
we can demonstrate this complication in another way , by plotting the observed energy difference @xmath63 of the line peaks in figure [ fig : width ] : @xmath64 .
notice that the dependence of @xmath65 on spin is ambiguous in some parts of the parameter space . in some situations
namely , for small radii and large spins the energy separation of the two peaks the observed profile is not in one - to - one correspondence to model parameters ; it is instead a double - valued quantity with respect to @xmath12 ( as can be seen in the top left corner of the fig .
[ fig : width ] left panel ) .
naturally , at large radii and small inclinations the spin dependence on @xmath65 is weak , and so the practical use of the method becomes compromised . on the other hand , we do see the range of the parameter space ( typically for radii between @xmath66 and @xmath67 ) , where the spin dependence is non - negligible and the possibility of reconstructing the parameters appears quite promising .
we can also notice an interesting conspiracy when the contour - line takes almost straight vertical direction ( e.g. in the same plot for @xmath68 deg , @xmath69 ) , implying that spin is degenerate with respect to radius for that given inclination . also , a narrow fe line
so ubiquitous in the x - ray spectra of type-1 agn will be unresolved by loft due to the poor energy resolution of its large area detector ( lad ) .
this could lead to confusion with the blue wing of the relativistic line if it happens to come close to the rest energy of the neutral line .
however , it does not mean an unsurmountable complication because these narrow features are thought to originate via scattering by a distant reflector ( e.g. a torus of matter located at several thousand gravitational radii , or more ) .
therefore , the narrow component does not vary on short time - scales and its influence can be taken into account by estimating the flux from other , high - energy resolution spectroscopy , and including it in the model . indeed , most of candidate objects for a relativistic line were already observed by xmm - newton @xcite , although a concurrent detection by loft together with another high - energy resolution mission would be the best strategy .
naturally , the problem of confusion or non - uniqueness would appear if the broad line is produced farther out from isco ( several tens of @xmath70 and more ) , so that the observed width is only moderate and comparable with the detector intrinsic energy resolution ; for such cases the loft will not be suitable .
we remark that the aforementioned simulations were performed for the response matrix resolution of 300 ev .
it is currently expected that a better resolution of about @xmath32@xmath33 ev can be realistically achieved @xcite , and so the obtained constraints should come out as even more favourable compared to the results presented here . as an example
, we constructed the confidence contours of fig .
[ fig : contours ] , but with the energy resolution of @xmath71 ev ( an optimistic expectation ) , and we checked that the contour levels come out very similar to those shown for @xmath34 ev resolution in the right panel .
both sets of best - fit values are consistent with each other .
the above - given simulations suggest that the mentioned features should be detectable around the iron line rest energy in bright accreting supermassive black holes .
the foreseen energy resolution of the detector is sufficient to reveal moderately broad excesses emerging on top of the underlying continuum ( unless the resolution becomes significantly worse than the currently discussed value about 200 ev )
. the signal - to - noise level would be more of a concern especially if the background level were not determined with sufficient confidence .
our example demonstrates that bright agns ( such as mcg6 - 30 - 15 during a prominent flare of * ? ? ?
* ) should be accessible to this kind of study .
to this end , the background model is required to be stable to approximately @xmath72% accuracy .
the expected signal from agns will be background dominated with loft .
therefore , for the applications discussed in this paper , the precise determination of the background model is important , especially near the iron line rest energy .
more detailed studies of the background impact are needed , however , currently available estimations @xcite suggest sufficient stability of the background model on the time - scales exceeding the orbital periods in the inner regions of agn accretion discs .
promising targets can be picked up from the list by ( * ? ? ?
* and further references cited therein ) ; these objects have already exhibited the presence of transient narrow - line features that could be relevant for our study .
naturally , galactic black holes in x - ray binaries represent another category of potential targets , assuming one can identify a suitable source with the flare - induced reflection signatures , similar to those seen in agns .
according to the current specifications , the temporal resolution does not pose serious limitations ( thanks to the large collecting area and sufficiently fast response time ) , even though the characteristic time - scale decreases proportionally to the mass , i.e. , by 67 orders of magnitude shorter for galactic black holes compared to supermassive black holes . on the other hand , high ionisation levels are typical of galactic black holes , and these may prevent successful detections despite the favorable source brightness that exceeds typical agns .
the adopted scenario has clearly an interesting potential to discriminate between the reflection model of spectral features from the case when the obscuration by intervening material and the resulting absorption are the main agents modulating the power - law continuum and shaping the observed spectrum . in summary ,
we arrive at the conclusion that the fiducial values of the model parameters can be reconstructed with the anticipated capabilities of loft .
although in this paper we considered the spectral features formed by azimuth - integrated rings , our preferred scenario of forming these rings is by orbital motion of spatially localized flares .
we note that the loft mission is dedicated primarily to timing studies .
indeed , it should provide us with light curves of flares to much better time resolution than any of its predecessors .
temporal profiles of the flare onset and the subsequent decay exhibit some signatures of model parameters ( see fig . 7 in * ? ? ?
* ) , but these are beyond current observational capabilities
. once resolved , they will give us a complementary information that can be compared against spectral studies discussed in the paper .
we explored a possibility of studying relatively indistinct excesses on top of the relativistically broadened spectral line profile . unlike the main body of the broad line , thought to originate from a whole extended region of the accretion disc , we modelled these features as the emission - line components arising in well - confined radial distance in the accretion disc .
we also suggested that some energetically narrow features could be explained as a signature of spatially localized irradiation by magnetic reconnection flares above the underlying accretion disc .
because of prevailing rotational motion of the accreted material and unavoidably prolonged duration of observations necessary to collect enough photons , even such localized events should reveal themselves in the observed spectrum as emission rings .
we found that the presence of about 1015% excess emission of the line flux originating from a ring of a moderate width ( typically a fraction of gravitational radius ) close to the black hole could be tested by the proposed loft satellite .
the effective area of this mission is designed to be large enough to allow the model parameters to be constrained despite its limited energy resolution .
this will significantly reduce the degeneracy of the model parameters a notorious obstacle complicating interpretations of the current data .
one needs to note that the ability of converging to the correct parameter values could be compromised if the background level is significantly increased above our default estimation , so this issue will need further investigation .
in fact , bright galactic stellar - mass black - hole candidates may be more suitable sources , as agn will be always dominated by the background . in this paper
we concentrated ourselves to the supermassive black hole especially because in this case the base - line model can be assumed in a relatively simple form and the iron - line complex has been commonly detected in agn , nevertheless , bright galactic binaries with accurately modelled spectral components may be eventually a more suitable category of targets . at this stage
we could only verify that the anticipated background level does not pose a serious limitation for the example described above , assuming that the background model is itself known with the sufficient accuracy .
this issue will need further investigation especially because the background level is expected to vary with energy , and so , in principle , inaccuracy of the background model could arise .
a typical double - horn profile gives us an opportunity to determine the parameters by measuring the energy shifts of the features within the broad spectral line wings .
the observed energy of these features is well - defined ( assuming a sufficiently narrow radial extent of the rings ) .
naturally , the inferred parameters from different rings should be consistent with each other ( e.g. the same inclination , assuming that the rings reside in the same equatorial plane ) .
this will help us to distinguish the disc line geometry of the line emitting region from alternative options , such as the lines originating in outflows and jets @xcite , spiral waves @xcite , or geometrically thick non - keplerian tori @xcite . even in situations when the contribution of rings is only moderate and the energy shifts
can not be determined immediately from the secondary peaks merely by inspecting the spectrum , the fitting procedure can be employed to reconstruct the model parameters .
this requires employing a physically substantiated model of the spectrum and using the entire profile of the broad line and the continuum .
naturally , in the latter case the assumptions about the intrinsic spectrum are essential .
the concept seems to be well suited to study the effects which are often coined as `` strong gravity '' in the astrophysical context . despite a limited energy resolution and various convergence issues of the spectral fitting procedure discussed in the paper above
, the large effective area offers a significant progress over what can be presently achieved with xmm - newton not only in the case of timing studies but also for the spectroscopy of truly broad relativistic spectral features .
we thank dr . alessandra de rosa for advice concerning the background estimation , and an unknown referee for useful comments . this research is supported by the esa pecs project no .
vk and md gratefully acknowledge support from the czech science foundation grant 205/07/0052 and the ministry of education project me09036 .
vs is supported by the doctoral grant gar no .
js acknowledges support from the grant gar 202/09/0772 .
part of this work has been carried out within the cost action mp0905 black holes in a violent universe .
abramowicz m. a. , jaroszyski m. , kato s. , lasota , j .-
, raska a. , sadowski a. ( 2010 ) , a&a , 521 , a15 artymowicz p. , lin d. n. c. , wampler e. j. ( 1993 ) , apj , 409 , 592 ballantyne d. r. , mcduffie j. r. , rusin j. s. ( 2011 ) , apj , 734 , i d .
112 coppi b. , rousseau f. ( 2006 ) , apj , 641 , 458 cuadra j. , armitage p. j. , alexander r. d. , begelman m. c. ( 2009 ) , mnras , 393 , 1423 czerny b. , raska a. , doviak m. , karas v. , dumont a .- m .
( 2004 ) , a&a , 420 , 1 dauser t. , wilms j. , reynolds c. s. , brenneman l. w. ( 2010 ) , mnras , 409 , 1534 de la calle prez i. , longinotti a. l. , guainazzi m. , bianchi s. , doviak m. et al .
( 2010 ) , a&a , 524 , a50 de marco b. , iwasawa k. , cappi m. , dadina m. , tombesi f. et al .
( 2009 ) , a&a , 507 , 159 doviak m. , bianchi s. , guainazzi m. , karas v. , matt g. ( 2004a ) , mnras , 350 , 745 doviak m. , karas v. , yaqoob t. ( 2004b ) , apjs , 153 , 205 fabian a. c. , iwasawa k. , reynolds c. s. , young , a. j. ( 2000 ) , pasp , 112 , 1145 fabian a. c. , pounds k. a. , blandford r. d. ( 2004 ) , frontiers of x - ray astronomy ( cambridge university press : cambridge ) feroci m. , stella l. , vacchi a. , labanti c. , rapisarda m. et al .
( 2010 ) , proc .
spie , 7732 - 66 ( arxiv:1008.1009 ) feroci m. , and the loft consortium ( 2011 ) , experimental astronomy , in press fuerst s. w. , wu kinwah ( 2007 ) , a&a , 474 , 55 goosmann r. w. , mouchet m. , czerny b. , doviak m. , karas v. , raska a. , dumont a
( 2007 ) , a&a , 475 , 155 guainazzi m. ( 2003 ) , a&a , 401 , 903 hartnoll s. a. , blackmann e. g. ( 2001 ) , mnras , 332 , l1 kato s. , fukue j. , mineshige s. ( 1998 ) , black - hole accretion disks ( kyoto university press : kyoto ) karas v. , martocchia a. , ubr l. ( 2001 ) , pasj , 53 , 189 karas v. , sochora v. ( 2010 ) , apj , 725 , 1507 karas v. , ubr l. ( 2001 ) , a&a , 376 , 686 king a. r. , pringle j. e. ( 2006 ) , mnras , 373 , l90 markowitz a. g. , reeves j. n. ( 2009 ) , apj , 705 , 496 mcclintock j. e. , remillard r. a. ( 2006 ) , in : compact stellar x - ray sources , eds .
w. lewin & m. van der klis ( cambridge : cambridge university press ) , pp .
157213 miller j. m. , fabian a. c. , wijnands r. , remillard r. a. , wojdowski p.
et al . ( 2002 ) , apj , 578 , 348 nandra k , oneill p. m. , george i m. , reeves j. n. ( 2007 ) , mnras , 382 , 194 novikov i. d. , thorne k. s. ( 1973 ) , in black holes , eds . c. dewitt and b. s. dewitt ( new york : gordon and breach publishers ) , p. 343
page d. n. , thorne k. s. ( 1974 ) , apj , 499 , 191 pariev v. i. , bromley b. c. , miller w. a. ( 2001 ) , apj , 547 , 649 ponti g. , cappi m. , dadina m. , malaguti g. ( 2004 ) , a&a , 417 , 451 reynolds c. s. , fabian a. c. ( 2008 ) , apj , 675 , 1048 reynolds , c. s. , nowak , m. a. ( 2003 ) , phys . rep .
, 377 , 389 schnittman j. , krolik j. h. , hawley j. f. ( 2006 ) , apj , 651 , 1031 svoboda j. , doviak m. , goosmann r. , karas v. ( 2009 ) , a&a , 507 , 1 svoboda j. , guainazzi m. , karas v. ( 2010 ) , a&a , 512 , a62 tagger p. , varnire p. ( 2006 ) , apj , 652 , 1457 uzdensky d. a. , goodman j. ( 2008 ) , apj , 682 , 608 wang j .- m . , zhou y .- y . , yuan y .-
f . , cao x. , wu m. ( 2000 ) , apj , 544 , 381 wilkins d. r. , fabian a. c. ( 2011 ) , mnras , 414 , 1269 | spectral features can arise by reflection of coronal x - rays on a black hole accretion disc .
the resulting profile bears various imprints of strong gravitational field acting on the light emitting gas .
the observed shape of the reflection line is formed by integrating contributions over a range of radii across the accretion disc plane , where the individual photons experience different level of energy shifts , boosting , and amplification by relativistic effects .
these have to be convolved with the intrinsic emissivity of the line , which is a function of radius and the emission angle in the local frame .
we study if the currently discussed instruments on - board x - ray satellites will be able to reveal the departure of the line radial emissivity from a simple smooth power - law function , which is often assumed in data fitting and interpretation .
such a departure can be a result of excess emission occurring at a certain distance .
this could be used to study variations with radius of the line production or to constrain the position of the inner edge of the accretion disc . by simulating artificial data from a bright active galactic nucleus of a type-1 seyfert galaxy ( inclination @xmath0 deg , x - ray flux @xmath12 mcrab in kev energy band )
we show that the required sensitivity and energy resolution could be reached with large area detector of the proposed loft mission .
galactic black holes will provide another category of potentially suitable targets if the relativistic spectral features are indeed produced by reflection from their accretion discs .
[ firstpage ] black hole physics accretion , accretion disks galaxies : nuclei |
monitoring known radial velocity planets at predicted transit times presents an avenue through which to explore the mass - radius relationship of exoplanets into regions of period / periastron space beyond that which is currently encompassed @xcite .
this is particularly true for those planets in relatively eccentric orbits @xcite , as demonstrated by the the variation in transit probability with the argument of periastron ( see figure 1 , left panel ) . here
we describe techniques for refining ephemerides and performing follow - up observations @xcite .
these methods are being employed by the transit ephemeris refinement and monitoring survey ( terms ) .
the transit window as described here is defined as a specific time period during which a complete transit ( including ingress and egress ) could occur for a specified planet .
the size of a transit window depends upon the uncertainties in the fit parameters .
they also deteriorate over time , motivating follow - up of the transit window as soon as possible after discovery .
figure 1 ( right panel ) shows the size of the first transit window ( after the fit time of periastron passage , @xmath0 ) for a sample of 245 exoplanets .
the transit windows of the short - period planets tend to be significantly smaller than those of long - period planets since , at the time of discovery , many more orbits have been monitored to provide a robust estimate of the orbital period .
targets chosen for terms tend to be those which have small transit windows , medium - long periods , and a relatively high probability of transiting the host star . [
cols="^,^ " , ]
a considerable number of high transit probability targets are difficult to adequately monitor during their transit windows because the uncertainties in the predicted transit mid - points are too high ( months or even years ) .
the acquisition of just a handful of new radial velocity measurements at carefully optimised times can reduce the size of a transit window by an order of magnitude .
this is described in more detail by @xcite .
due to the wide range of stars monitored , both in sky location and brightness , terms collaborates with a variety of existing groups to take advantage of transit window opportunities .
note that the observations from this survey will lead to improved exoplanet orbital parameters and ephemerides even without an eventual transit detection for a particular planet .
the results from this survey will provide a complementary dataset to the fainter magnitude range of the _ kepler _ mission , which is expected to discover many transiting planets including those of intermediate to long - period .
kane , s.r .
, 2007 , mnras , 380 , 1488 kane , s.r . , von braun , k. , 2008 , apj , 689 , 492 kane , s.r . , von braun , k. , 2009 , pasp , 121 , 1096 kane , s.r . , mahadevan , s. , von braun , k. , laughlin , g. , ciardi , d.r . , 2009 ,
pasp , in press ( arxiv:0910.0010 ) | transiting planet discoveries have yielded a plethora of information towards understanding the structure and atmospheres of extra - solar planets .
these discoveries have been restricted to the short - period or low - periastron distance regimes due to the bias inherent in the geometric transit probability . through the refinement of planetary orbital parameters , and hence reducing the size of transit windows , long - period planets become feasible targets for photometric follow - up . here
we describe the terms project which is monitoring these host stars at predicted transit times . |
in this paper , we compute the @xmath2-invariant cone of effective divisors of the fulton - macpherson configuration space of @xmath0 points on @xmath1 .
this space is isomorphic to the moduli space @xmath5 of stable maps of degree one from genus zero curves with @xmath0 marked points to @xmath1 .
we also compute the effective cone of the generic fiber of the natural map @xmath6 our motivation is to provide a geometric explanation of a formula , obtained by duke , rudnick and sarnak , giving the asymptotic behavior of the number of binary forms of degree @xmath0 with fixed discriminant and bounded integral coefficients .
this fits into a larger program to predict and prove asymptotic formulas for the number of rational and integral points of bounded height on algebraic varieties .
we introduce a counting function for integral points on an algebraic variety as follows : given a variety @xmath7 over a ring of integers @xmath8 and functions @xmath9 , regular on @xmath7 , define @xmath10 where @xmath11 is a valuation on @xmath8 .
this is finite only when the functions @xmath12 give an embedding of @xmath7 .
it is most natural to interpret the functions @xmath12 as _ sections _ of a line bundle @xmath13 on a projective compactification @xmath14 defined over the fraction field @xmath15 of @xmath8 .
the fact that the sections embed @xmath7 implies that @xmath13 is _ big _ ,
i.e. , is contained in the interior of the effective cone @xmath16 of @xmath17 .
therefore , in order to describe all natural counting functions on open subsets of @xmath17 we need to compute its effective cone .
furthermore , in many cases it can be proved that the asymptotic properties of @xmath18 are intimately related to the structure of this cone .
let @xmath19 be a homogeneous polynomial of degree @xmath0 in @xmath20 variables .
a standard heuristic in number theory predicts that the number @xmath21 of integral solutions of the equation @xmath22 of `` height '' @xmath23 grows asymptotically like @xmath24 as @xmath25 . when the number of variables is @xmath26 , the affine variety @xmath27 defined by @xmath28 is smooth and there are no local obstructions , an asymptotic formula can be established using the classical circle method in analytic number theory ( see @xcite , @xcite and the references therein )
of course , there may be difficulties when the number of variables is small or the variety @xmath27 is singular .
the following example appeared in the paper by duke , rudnick and sarnak @xcite .
consider the vector space of binary forms of degree @xmath0 @xmath29 the algebraic group @xmath3 acts on this space by coordinate substitutions .
when @xmath30 , the _ discriminant form _
@xmath31 generates the ring of @xmath3-invariants .
then there exists a constant @xmath32 so that @xmath33 as @xmath25 . note that the exponent @xmath34 is larger than what is predicted by the standard heuristic . more generally , one has the [ thm : example ] @xcite fix a generic binary form @xmath35 of degree @xmath4 with integral coefficients .
let @xmath36 be the number of binary forms @xmath37-equivalent to @xmath35 with coefficients bounded by @xmath38 .
then there exists a @xmath32 such that @xmath39 as @xmath25 .
we give a geometric interpretation of the exponent @xmath40 in theorem [ thm : example ] . to this end , we refine the heuristics for counting integral points to take into account singularities of the relevant varieties ( see conjecture [ conj : a ] ) . we verify that conjecture [ conj : a ] is consistent with theorem [ thm : example ] in theorem [ thm : main ] .
its proof involves the computations of effective cones alluded to above .
* acknowledgments : * we are grateful to the alfrd rnyi institute of the hungarian academy of sciences for organizing the conference at which much of this work was done .
we work over a field of characteristic zero .
let @xmath17 be a normal projective variety with canonical class @xmath41 and let @xmath42 be a reduced effective weil divisor of @xmath17 .
[ defn : good ] a _ good _ pair @xmath43 consists of a smooth projective variety @xmath17 and a strict normal crossings divisor @xmath42 in @xmath17 .
this means that all irreducible components of @xmath42 are smooth and intersect transversally .
let @xmath43 be a good pair and let @xmath16 denote the closed cone of effective divisors classes of @xmath17 ; a divisor is big exactly when its class is in the interior of this cone .
define @xmath44 where we identify line bundles and their divisor classes .
note that @xmath45 is a positive real number whenever @xmath46 is big .
the constant @xmath47 is called the _ log - kodaira energy _ of @xmath13 ( see @xcite ) .
if @xmath43 is not good then resolution of singularities implies the existence of a _ good resolution _ @xmath48 .
precisely , @xmath49 is a good pair , @xmath50 a birational projective morphism , and @xmath51 is the union of the exceptional divisors of @xmath50 and the proper transform of @xmath42 . recall that @xmath43 is _ log - canonical _ if @xmath52 is @xmath53-cartier and @xmath54 where the @xmath55 are the exceptional divisors of @xmath50 and @xmath56 for all @xmath57 .
[ exam : lc ] when @xmath17 is a smooth surface , @xmath43 is log - canonical only when the curve @xmath42 is smooth or nodal .
if @xmath17 is smooth of arbitrary dimension , @xmath42 must have at worse nodes in codimension one . if @xmath13 is a line bundle on @xmath17 put @xmath58 where @xmath59 is the total transform of @xmath42 .
note that @xmath45 is computed on @xmath60 .
[ prop : ind ] let @xmath43 be a log - canonical pair and assume that @xmath61 has canonical singularities . if @xmath13 is a big line bundle on @xmath17 then @xmath62 in particular , @xmath45 does not depend on the choice of a desingularization .
choose a good resolution @xmath48 , so that @xmath63 where @xmath56 , and @xmath64 if @xmath65 .
in particular , each exceptional divisor not contained in the total transform @xmath66 has log discrepancy @xmath67 .
therefore , we have @xmath68 with each @xmath69 . for any @xmath70-cartier divisor @xmath71 on @xmath17 and
effective divisor @xmath72 supported in the exceptional locus of @xmath50 , @xmath71 is effective iff @xmath73 is effective .
[ prop : eff ] let @xmath74 and @xmath75 be log - canonical pairs , so that @xmath76 and @xmath77 have canonical singularities .
assume that @xmath78 is a finite dominant morphism so that @xmath79 let @xmath13 be a big divisor on @xmath80 .
then @xmath81 . in fact
, it suffices to assume that either @xmath74 or @xmath75 satisfies the singularity condition ( * ? ? ?
20.3 ) .
given a finite dominant morphism @xmath82 and a @xmath70-cartier divisor @xmath71 on @xmath80 , @xmath71 is effective iff @xmath83 is effective . indeed , the divisor @xmath84 is defined and equal to @xmath85 . combining this with proposition [ prop : ind ]
gives the result .
let @xmath43 be a log terminal pair so that @xmath61 has singularities which are _ not _ canonical .
then our definition of the kodaira energy differs slightly from fujita s @xcite . in applications to integral points , we are interested in invariants of the open variety @xmath61 . in fujita s definition , on passing from @xmath43 to a good resolution , any exceptional divisors over @xmath61 with negative discrepancy must be added to the boundary .
this changes the open variety .
retain the notation from the previous section and assume that @xmath17 and @xmath42 are defined over a number field @xmath15 .
let @xmath86 denote the ring of integers of @xmath15 , where @xmath87 is a finite set of nonarchimedean places of @xmath15 .
fix models @xmath88 and @xmath89 flat and proper over the ring of integers @xmath86 .
a @xmath90-integral point is an @xmath86-point of @xmath91 .
in particular , if @xmath92 an integral point is the same as a rational point on @xmath17 .
let @xmath93 be a very ample metrized line bundle on @xmath17 , @xmath94 a zariski open subset and @xmath95 a model of @xmath7 over @xmath86 .
let @xmath87 be a finite set of places in @xmath15 , including the archimedean places .
let @xmath96 denote the number of @xmath90 integral points on @xmath7 of @xmath93-height bounded by @xmath38 .
a natural extrapolation of vojta s conjecture about integral and rational points on varieties of ( log-)general type @xcite and batyrev - manin conjectures about rational points of bounded height on fano varieties @xcite would be : [ conj : a ] for any @xmath97 , there exists a dense zariski open subset @xmath94 such that @xmath98 as @xmath25 . if @xmath46 is big then @xmath99 as @xmath25 , at least after a suitable finite extension of @xmath15 and @xmath87
. the statement is independent of the choice of @xmath87 and the choice of a metrization on @xmath13 .
many precise results about asymptotics of rational and integral points are currently available ( see , for example , @xcite and the references therein ) . as far as we know , conjecture [ conj : a ] is compatible with all of them . however , to actually check this compatibility one has to compute the geometric invariants of ( some resolution of ) the pair @xmath43 .
in particular , one has to determine the effective cone .
this can be a formidable task even for rational varieties , e.g. , like the moduli space of pointed rational curves @xmath100 ( see @xcite ) .
let @xmath17 be a nonsingular projective variety , perhaps with an action by a finite group @xmath101 .
we review strategies for computing the @xmath101-invariant effective cone @xmath102 and thus the effective cone of the quotient @xmath103 ( cf .
@xcite ) .
a curve class @xmath104 \in n_1(x)$ ] is said to be _ nef _ if @xmath104.d\geq 0 $ ] for each @xmath105 .
a family of curves passing through the generic point of @xmath17 is automatically nef .
indeed , consider a family @xmath106 of integral projective curves in @xmath17 and an irreducible codimension - one subvariety @xmath107 . if , for generic @xmath108 , the fiber @xmath109 , we have @xmath110.d\geq 0 $ ] .
fix a collection of effective divisors @xmath111 which we expect to generate @xmath102 . to prove that @xmath112 generates the ( @xmath101-invariant ) effective cone , it suffices to find a collection of nef ( @xmath101-invariant ) curve classes @xmath113 so that the cone generated by @xmath112 contains the dual to the cone generated by @xmath114 .
in section [ sect : fm ] , we shall use a refinement of this method ( see @xcite , @xcite ) .
a divisor @xmath105 is _ moving relative to @xmath112 _ if some multiple of @xmath42 contains no element of @xmath112 as a fixed component .
every effective divisor is a sum @xmath115 where @xmath71 is moving relative to @xmath112 . to prove that @xmath112 generates the effective cone , it suffices to show that @xmath71 is an effective sum of the @xmath116 .
a curve class is _ nef relative to @xmath112 _ if @xmath104.m\geq 0 $ ] for each @xmath71 which is moving relative to @xmath112 .
any family of curves passing through the generic point of _ some _ @xmath116 is nef relative to @xmath112 . consequently , to show that @xmath112 generates the effective cone
, it suffices to find a collection @xmath114 of curve classes , nef relative to @xmath112 , so that the cone generated by @xmath112 contains the dual to the cone generated by @xmath114 .
let @xmath117 be a two - dimensional vector space with coordinates @xmath118 and @xmath119 , equipped with the standard @xmath3-action .
let @xmath120 be the space of binary forms of degree @xmath0 @xmath121 it carries an induced action of @xmath3 by substitution .
associating to each form @xmath122 its roots @xmath123 yields a map @xmath124 and an identification @xmath125 .
the _ discriminant _ of a polynomial @xmath35 is a homogeneous form in its coefficients @xmath126 and defines a divisor @xmath127 .
now we may state our main result : [ thm : main ] let @xmath35 be a generic bilinear form of degree @xmath0 , @xmath128 the closure of the @xmath3-orbit through @xmath35 , @xmath129 the intersection of the discriminant with @xmath130 , and @xmath13 the restriction of the standard polarization to @xmath130 .
then we have @xmath131 in particular , conjecture [ conj : a ] is consistent with theorem [ thm : example ] . to prove this
, we require a resolution ( i.e. , a partial desingularization ) of @xmath132 on which we may evaluate @xmath133 using proposition [ prop : ind ] .
this resolution will be induced by a natural resolution of @xmath43 .
example [ exam : lc ] shows that @xmath43 is far from being log - canonical .
when @xmath30 , the discriminant has cusps in codimension one : a transverse slice @xmath134 intersects the discriminant in the cuspidal curve @xmath135 our resolution of @xmath43 will be a @xmath2-quotient of a natural desingularization for @xmath136 , where @xmath137 is the diagonal , i.e. , the points lying over the discriminant .
both admit interpretations as moduli spaces of stable maps .
fix an integer @xmath4 .
let @xmath100 denote the knudsen - mumford moduli space of stable curves of genus zero with @xmath0 marked points @xcite @xmath138 let @xmath139 denote the kontsevich moduli space of stable maps of degree one from genus - zero curves with @xmath0 marked points to @xmath140 @xcite @xmath141 this is naturally isomorphic to the fulton - macpherson @xcite configuration space @xmath142 $ ] for @xmath0 points in @xmath140 ( see @xcite 0 ) .
however , for our purposes it is convenient to use the moduli space notation .
we have the following natural maps : 1 .
the evaluation map @xmath143 2 . forgetting the point @xmath144 @xmath145 3 .
taking projective equivalence classes @xmath146 @xmath147 is obtained from @xmath148 by ` collapsing ' the irreducible components which are destabilized when @xmath144 ( resp . , the polarization ) is removed .
finally , we enumerate the boundary divisors of these moduli spaces . for each partition
@xmath149 consider stable curves @xmath150 which form a divisor @xmath151 .
the union of these is denoted @xmath152 .
note that the @xmath2-orbits of @xmath153 correspond to the integers @xmath154 for each subset @xmath155 consider stable maps @xmath156 collapsing the first component and mapping the second isomorphically onto @xmath140 .
these form a divisor @xmath157 .
the @xmath2-orbits of @xmath158 correspond to integers @xmath159 and we define @xmath160:=\sum_{|s|=s } b_s\,\,\ , { \rm and}\,\,\ , b:=\sum_{s=2}^n b[s]\ ] ] [ thm : smooth ] the moduli spaces @xmath139 and @xmath100 are smooth projective algebraic varieties . moreover , the boundary is a divisor with strict normal crossings .
[ rem : lcc ] in particular , the pair @xmath161 is log - canonical .
we obtain a good resolution @xmath49 of @xmath43 using the above formalism .
consider the quotient map @xmath162 let @xmath163 $ ] and @xmath51 be the images of @xmath164 $ ] and @xmath38 under this map .
[ prop : ba ] the map @xmath165 is ramified only along the boundary @xmath38 . at the generic points of @xmath166
$ ] the ramification has order @xmath167 . for all @xmath168 ,
the map @xmath165 is unramified at the generic points of @xmath164 $ ] .
we have the formula @xmath169 and @xmath49 is log - canonical . the map @xmath165 ramifies at points corresponding to stable maps @xmath170 that admit an automorphism permuting the marked points .
the ramification order is the order of this automorphism group .
if the images of the @xmath0 points under @xmath171 are distinct then there is no automorphism of @xmath171 permuting them .
this proves the first assertion .
if marked points coincide there is an irreducible component @xmath172 which is collapsed by @xmath171 and which contains these points . if there are two such points this component admits an automorphism of order two exchanging the points and fixing the point of intersection with the rest of @xmath148 .
this proves the second assertion .
if there are more than two marked points then there is generally no such automorphism .
this proves the third assertion . the ramification formula and the fact that the pair @xmath49 is log - canonical follow from an easy local computation combined with remark [ rem : lcc ] ( see propositions 20.2 and 20.3 of @xcite ) .
take @xmath2-quotients of the point map to obtain a birational map @xmath173 assigning to @xmath174 a polynomial vanishing at these points . the boundary divisor @xmath175 $ ] is the proper transform of the discriminant @xmath42 under @xmath176 .
the boundary divisors @xmath163 $ ] ( for @xmath177 ) are the exceptional divisors for @xmath176 .
let @xmath178 be a set of distinct complex numbers and @xmath179 the binary form of degree @xmath0 with roots @xmath180 .
let @xmath181 be the corresponding pointed rational curve and @xmath182 the corresponding map .
the fiber @xmath183 contains @xmath184 .
let @xmath185 be the image of @xmath186 under the quotient map @xmath165 and @xmath187 its intersection with the boundary @xmath51 .
this coincides with the general fiber of the map @xmath188 the map @xmath176 induces a resolution @xmath189 with @xmath190 . to describe the @xmath186 explicitly
, we use the tower when @xmath30 , @xmath191 and @xmath192 , which is isomorphic to the product @xmath193 blown up along the small diagonal @xmath194 .
the boundary divisors correspond to the following stable maps @xmath195 = \begin{array}{c } \includegraphics{b2 } \end{array},\,\,\,\ , b[3 ] = \begin{array}{c } \includegraphics{b3 } \end{array}.\ ] ] in the above pictures the collapsed components are represented by vertical lines .
note that the normal bundle @xmath196 so that the exceptional divisor @xmath197\simeq \bp^1\times \bp^1 $ ] .
let @xmath198 be the projection to the cross ratio of the marked points and the node and @xmath199 the projection onto the image of the collapsed curve .
the divisor @xmath166 $ ] is the proper transform of @xmath137 , the large diagonal . for the arbitrary degree case
, we analyze the failure of the block squares in the tower to be fiber products . given a generic @xmath200 we compare the fibers @xmath201 using the forgetting map @xmath202 given a stable map @xmath203 there are three cases to consider : 1 .
@xmath204 ; 2 .
@xmath205 with the collapsed component containing @xmath206 ; 3 .
@xmath205 with the collapsed component containing @xmath207 but not @xmath208 .
case 1 : @xmath209 case 2 : @xmath210 case 3 : @xmath211 over the open subset of @xmath212 corresponding to the first two cases , @xmath213 induces an isomorphism between @xmath212 and @xmath214 . in the third case , we forget the image of the @xmath0-th marked point . the map @xmath213 blows up the locus in @xmath214 where @xmath207 are on the collapsed component and @xmath208 coincides with the node ( of attachment ) .
this is a curve isomorphic to @xmath215\cap y_{{\alpha}_1, ... ,{\alpha}_{n-1}})$ ] : the generic map takes the form : we summarize the above discussion in the following [ prop : all ] let @xmath216 be distinct complex numbers .
the forgetting maps induce a sequence of birational morphisms @xmath217 the moduli space of stable maps @xmath218 is isomorphic to @xmath193 blown up along the small diagonal with exceptional divisor @xmath219 . the map @xmath220 blows up the proper transform of @xmath221 . in particular , @xmath212 is smooth and its boundary has strict normal crossings , contained in @xmath222\cup b[n]$ ] .
[ rem : dis ] we are blowing up along disjoint curves , so the order of the blow - up does not matter .
[ prop : fff ] let @xmath35 be a generic binary form of degree @xmath4 with roots @xmath206 .
then the restriction of @xmath165 to @xmath130 is ramified only along the boundary @xmath223 . at generic points of @xmath224\cup
b[n-1])\cap y_{{\alpha}}$ ] , the restriction of @xmath165 is unramified .
we have the formula @xmath225\ ] ] and @xmath226 is log - canonical .
the argument is similar to the one in proposition [ prop : ba ] , and is omitted .
write @xmath228 with large diagonals @xmath229=2(g_1+g_2+g_3)-3e.\ ] ] by proposition [ prop : all ] , @xmath186 is obtained by blowing up the @xmath230 sections of @xmath231\ra \bp^1.\ ] ] let @xmath232 denote the corresponding exceptional divisors and identify @xmath233 and its proper transform .
relabel @xmath234\\ & = & g_i+g_j - e - f_4- \cdots - f_n , \end{array}\ ] ] so that @xmath235 acts on the @xmath236 , @xmath237 , in the obvious way .
note that @xmath233 and the @xmath236 generate @xmath227 .
[ prop : ky ] the @xmath235-stable boundary divisors @xmath238 & = & f_1+\cdots + f_n\\ a[n ] & = & e , \end{array}\ ] ] generate the @xmath235-invariant picard group of @xmath186 , and @xmath239=b[j]\cap y_{{\alpha}}$ ] .
the canonical class of @xmath186 is @xmath240 - 2a[n ] \end{array}\ ] ] [ lemm : comp ] the @xmath235-invariant effective cone of @xmath186 is generated by the classes @xmath241 $ ] and @xmath242 $ ] .
we apply the method of [ sect : compeff ] .
the class @xmath241 $ ] is exceptional and thus a generator of the effective cone . to show that @xmath242 $ ] is the second generator
, we exhibit a nef curve not intersecting @xmath242 $ ] .
consider the @xmath243-action on @xmath140 : @xmath244 we may assume that the points @xmath206 are not contained in the fixed point locus of @xmath245 . any singular element in the orbit closure is : where the point of attachment is 0 ( or @xmath246 ) and the other labelled point is @xmath246 ( resp . 0 ) .
this is disjoint from @xmath242 $ ] .
the kodaira energy for @xmath132 can be computed on @xmath226 , by propositions [ prop : ind ] and [ prop : fff ] . by propositions [ prop : eff ] and [ prop : all ]
, it suffices to compute the kodaira energy for @xmath247+a[n-1])$ ] .
recall there is a composed morphism @xmath248 the pull - back of the hyperplane class takes the form @xmath249= \frac{1}{2}\left ( ( n-2)a[n-1]+na[n]\right).\ ] ] let @xmath250 be the class of a curve in @xmath241 $ ] corresponding to with varying point of attachment on the _ collapsed _ component .
this is the proper transform of the generic fiber of the map @xmath251 .
then @xmath252|_r & = & n = \ # \left\ { \begin{array}{c } \includegraphics{ad } \end{array } , j=1,\ldots , n \right\ } \\ a[n]|_r & = & -1-(n-3)=2-n .
\end{array}\ ] ] for the second intersection number , note that @xmath241=e$ ] and apply the blow - up description of proposition [ prop : all ] . in @xmath218 we have @xmath253 and @xmath254 .
after blowing up @xmath230 further sections of @xmath255 the normal bundle is reduced to @xmath256 . we know that @xmath241=b[n]\cap y_{{\alpha}}$ ] is collapsed by the map @xmath257 , so @xmath258+na[n]\right)\ ] ] for some @xmath259 . since @xmath260+na[n]\\ = 2\left((n-2)(g_1+g_2+g_3)-(n-3)e-(n-2)(f_4+\cdots + f_n)\right).\end{gathered}\ ] ] the claim follows .
we have @xmath261&=&k_{y_{{\alpha}}}+a[n]+a[n-1]=-a[n]\\ k_{y_{{\alpha}}}+[y_{{\alpha}}\cap b]+\alpha l & = & \alpha \frac{n-2}{2}a[n-1 ] + ( \alpha \frac{n}{2}-1)a[n],\end{aligned}\ ] ] and by definition @xmath262 \in \lambda_{\rm eff}(y_{{\alpha}})\}.\ ] ] hence lemma [ lemm : comp ] yields @xmath263)=2/n.\ ] ] thus @xmath264 , as desired !
in this section , we compute the @xmath2-invariant part of the effective cone of @xmath266 , its canonical class , and the kodaira energy of the line bundle @xmath267 , where @xmath268 we will also compute the kodaira energy of @xmath269 .
we first recall some basic facts about @xmath270 $ ] , following @xcite .
in addition to the divisor classes @xmath271 introduced above , we shall also consider @xmath272 the cohomology @xmath273 is generated by the classes @xmath274 and @xmath271 , subject to the relations 1 .
@xmath275 ; 2 .
@xmath276 for @xmath277 ; 3 .
@xmath278 for @xmath279 ; 4 .
@xmath280 , for @xmath281 .
the generators of the @xmath235-invariant subspace are @xmath282=\sum_{|s|=s } b_s , \,\,\ , 2\le s \le n.\ ] ] after averaging over @xmath235 @xmath283.\ ] ] [ thm : efff ] the classes @xmath284 , ... , d[n]$ ]
generate the effective cone of @xmath285 .
the classes @xmath166 , ... ,
b[n]$ ] generate the @xmath235-invariant effective cone of the moduli space @xmath5 .
we implement the strategy of [ sect : compeff ] with @xmath286,\ldots , b[n ] \}.\ ] ] this entails finding curve classes that are nef relative to @xmath112 .
let @xmath287\ ] ] denote an @xmath2-invariant divisor class with no boundary divisors as fixed components .
recall the description of the boundary divisor @xmath271 : @xmath288 take @xmath177 and let @xmath289 be the class of the generic fiber of the map @xmath290 forgetting the attaching point .
since @xmath291 passes through the generic point of @xmath271 , averaging @xmath291 over @xmath265 yields a curve class which is nef relative to @xmath112 .
in particular , for each @xmath2-invariant divisor @xmath292 $ ] , moving relative to @xmath112 , we have @xmath293 we compute intersections of @xmath291 with the various elements of @xmath112 .
first , the map @xmath257 blows down the divisors @xmath271 for @xmath294 ; the data of the collapsed component is lost completely .
it follows that @xmath295 .
a simple combinatorial analysis gives @xmath296 which means that @xmath297=s$ ] .
relation [ relation ] gives @xmath298 so @xmath299 .
to summarize , we have @xmath300 = \left\ { \begin{array}{cl } s & \text { if } j = s-1 ; \\
-(s-2 ) & \text { if } j = s ; \\ 0 & \text { otherwise . } \end{array } \right.\ ] ] using this information , we extract inequalities on the coefficients of @xmath71 .
the condition @xmath301 yields @xmath302 so we get a chain of inequalities : @xmath303 if some @xmath304 then @xmath305 for each @xmath306 .
we consider another curve class in @xmath271 to get inequalities in the reverse direction .
fix @xmath307 and let @xmath308 denote the class of the generic fiber of @xmath309 induced by forgetting @xmath310 , one of the @xmath311 points not contained in @xmath87 .
again , @xmath308 passes through the generic point of @xmath271 , so averaging over @xmath265 yields a curve class such that @xmath312 we compute intersections as before .
the map @xmath257 sends @xmath308 to a line in @xmath313 , i.e. , the linear forms with @xmath314 fixed roots and one varying root .
it follows that @xmath315 .
the line @xmath308 intersects @xmath316 properly in the following cases @xmath317 summing over @xmath265-orbits gives @xmath318 = \left\ { \begin{array}{cl } 1 & \text { if } j = s+1;\\ n - s-1 & \text { if } j=2 ; \\
0 & \text { otherwise , unless } j = s .
\end{array } \right.\ ] ] applying relation [ relation ] , we find @xmath319 + ( s+1)s/2 + ( n - s-1),\ ] ] so @xmath320 = -1.$ ] we extract the inequalities @xmath321 in particular , if @xmath322 then each @xmath323 .
adding together the inequalities @xmath324 gives @xmath325 combining with inequality [ chain ] , we obtain @xmath326 hence @xmath322 .
[ thm : full ] on the moduli space of stable maps @xmath327 and its @xmath265-quotient @xmath328 , we have @xmath329 we proceed to calculate the canonical class : @xmath330(s-2)\ ] ] this follows from the explicit blowup realization of the fulton - macpherson configuration space @xmath331=\om_{0,n}(\bp^1,1)$ ] @xcite . the exceptional divisors @xmath271 ( for @xmath332 )
arise from blowing up centers in codimension @xmath333 .
we compute the log - kodaira energy of @xmath13 on @xmath266 with respect to the boundary @xmath38 .
we have @xmath334 \\ & = & ( a-2)\left ( \sum_{s=2}^n \frac{s(s-1)}{2(n-1)}b[s ] \right)+ \sum_{s=2}^n ( s-1)b[s ] \\ & = & \sum_{s=2}^n ( s-1 ) \left ( \frac{s(a-2)}{2(n-1)}+1 \right ) b[s]\end{aligned}\ ] ] which is effective if and only if @xmath335 the most restrictive inequality occurs when @xmath336 , where we obtain @xmath337 .
consequently , @xmath338 , as expected .
the kodaira energy for @xmath43 can be computed on @xmath49 , by propositions [ prop : ind ] and [ prop : ba ] . by proposition [ prop : eff ] and theorem [ thm : smooth ] ,
it is equal to the kodaira energy for @xmath327 .
\a ) the orbit closure @xmath130 depends on the form @xmath339 $ ] .
we get equivariant compactifications of @xmath340 depending on moduli .
this dependence is made abundantly clear in the blow - up description of proposition [ prop : all ] .
\b ) the pair @xmath341 is of log general type : @xmath342 is ample and log canonical ( see , for example , 7.1 of @xcite ) .
the map @xmath343 is a log - fano fibration onto a log - variety of general type ( or the point , when @xmath30 ) .
\c ) the pair @xmath344 satisfies vojta s conjecture .
we realize @xmath345 as an open subset of an algebraic torus with explicit complement .
fix @xmath314 points in @xmath346 in general position .
consider the set @xmath347 of @xmath348 hyperplanes spanned by @xmath349 of the fixed points .
kapranov @xcite has shown that @xmath350 .
the torus is obtained by excising the @xmath351 hyperplanes spanned by subsets of the first @xmath351 of the points .
\d ) we therefore expect that the asymptotic behavior of integral points of @xmath139 with respect to the boundary is obtained by summing the contributions of the integral points on the fibers of @xmath352 .
this explains why the kodaira energies for the moduli space of stable maps ( theorem [ thm : full ] ) and the fibers of @xmath352 ( theorem [ thm : main ] ) should coincide ( for the case of rational points , see @xcite ) .
w. fulton and r. pandharipande , notes on stable maps and quantum cohomology , _ algebraic geometry
santa cruz 1995 , _ 4596 ,
pure math .
, * 62 * , part 2 , amer .
soc . , providence , ri , 1997 .
b. hassett and yu .
tschinkel , on the effective cone of the moduli space of pointed rational curves , _ topology and geometry : commemorating sistag , _ a.j .
berrick , m.c .
leung , and x.w .
xu eds . , contemporary mathematics * 314 * , 8396 , american mathematical society , providence , 2002 . | consider the fulton - macpherson configuration space of @xmath0 points on @xmath1 , which is isomorphic to a certain moduli space of stable maps to @xmath1 .
we compute the cone of effective @xmath2-invariant divisors on this space .
this yields a geometric interpretation of known asymptotic formulas for the number of integral points of bounded height on compactifications of @xmath3 in the space of binary forms of degree @xmath4 . |
coevolution is often antagonistic , such that one species benefits from the other s loss .
classic examples include predators and their prey , and pathogens and their hosts .
antagonistic coevolution is commonly thought to result in rapid co - evolutionary arms races @xcite .
when participants in a coevolutionary arms race can tamper with their opponents control systems , as microbial pathogens do with host immune regulation @xcite , we might expect to see a series of subversion efforts and subsequent countermeasures deployed over evolutionary time
. thus one might expect rapid evolutionary divergence of the systems involved in controlling and regulating the attacks and defenses used in antagonistic interactions @xcite .
however , antagonistic coevolution need not always lead to rapid evolutionary change .
mechanisms that prevent subversion can slow down coevolutionary arms races ; the field of cryptography abounds with examples of such systems . in a prescient 1955 letter only recently declassified ,
john nash anticipated this result @xcite : _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ ... for almost all sufficiently complex types of enciphering , especially where the instructions given by different portions of the key interact complexly with each other in the determination of their ultimate effects on the enciphering , the mean key computation length increases exponentially with the length of the key .... as ciphers become more sophisticated the game of cipher breaking by skilled teams , etc .
, should become a thing of the past .
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ nash was right ; one important example is the rsa cryptosystem @xcite , in which two parties communication over a network can not be decoded by adversaries , unless they successfully find the prime factors of a large number .
prime factorization has been proven to be computationally difficult ( or in the parlance of computer science , has high time complexity ) and so the system is effectively secure .
the main insight for rsa was that a mechanism can be made secure against subversion by using intractability or complexity as a defense . in this paper
we explore how _ defensive complexity _ strategies can be generalized to domains beyond cryptography for example , immunology .
we review evidence showing that the overarching control logic of the adaptive immune system ( as revealed in the dynamics and structure of the immune system ) is phylogenetically conserved in mammalian lineages .
by contrast , other facets of the immune system show evidence of selection .
thus both the phylogenetic evidence and the structure of immune control logic are consistent with the defensive complexity hypothesis .
to model the effect of defensive complexity on antagonistic coevolution , we introduce a new evolutionary game , the _ signal tampering game_. this game features two players , the _ controller _ and the _ adversary_. the controller aims to respond appropriately to the state of the environment .
to do this , the controller deploys a control system intermediating between sensors that receive a cue about the state of the world , and effectors that take an action .
this control system is codified as a control logic with the cue as input and the effector responses as outputs .
the controller s payoff is a function of the world state and the actions taken .
the adversary aims to interfere , and can do so by tampering with some of the signals in the control logic .
we study what happens when this game is played in an evolutionary context .
we pay particular attention to the case in which the controller must first deploy a control logic , and the adversary then has multiple periods in which to learn how to manipulate it .
such a state of affairs could come about for many reasons .
one common biological scenario is when learning occurs at the population level by the mechanism of evolution by natural selection and evolutionary rates differ , as is the case in a vertebrate host deploying an immune control system against rapidly evolving pathogen adversaries @xcite .
this scenario provides us a well defined and formally tractable learning system to study , in which an asexual adversary population evolves on a fitness landscape and the relevant phenotype space corresponds to the set of possible manipulations to the control system
. questions of which control systems are hard to learn reduce to questions about the rate of evolution given the fitness landscape induced for the adversary by the control system .
we characterize rates of evolution for crossing long fitness valleys and demonstrate that even for a large adversary population with a high mutation rate , long fitness valleys are hard to cross .
this is counter - intuitive when applied to pathogen populations , because one might think that their large population size and high mutation rate would allow them to conduct a parallel search of many paths through fitness space . in practice
this argument turns out to be insufficient because the effective population size is not large enough to populate all possible pathways to the global optimum .
we show that complex control networks can generate sign epistasis in this fitness landscape , thereby sculpting fitness valleys that must be crossed and reducing the rate at which an antagonist population evolves to subvert the network .
thus control systems afford defensive complexity against natural adversaries if they induce fitness functions for attempted manipulation that take the form of adaptive landscapes with long deep fitness valleys . where sufficient defensive complexity is in place
, antagonistic co - evolution can lead to long periods of structural conservation instead of rapid change driven by an ongoing arms race .
rapid progress in molecular and cellular biology has provided a detailed picture of the wiring diagram of interacting cells and molecules in the immune systems of mice and men .
even a cursory glance at textbooks of immunology @xcite and review papers reveals the extraordinary complexity of this wiring diagram .
quantitative measures of complexity likewise indicate that the immune system is an outlier among physiological systems , with denser and more rich network motifs than other biological signaling networks @xcite .
the immune signaling and control network in mice is remarkably similar to that in men ; hence the utility of a mouse model system for understanding human immune function .
we now discuss how advances in sequencing and the development of phylogenetic tools for the analysis of these sequences gives us the ability to look at the evolution of immune genes in unprecedented detail .
sequencing of the genomes of various mammals including mice , primates and humans reveals that the genes associated with some immune functions have evolved faster than the rest of the genome @xcite .
this has been done in a number of ways .
the mouse genome consortium measured the rates of non - synonymous and synonymous mutations in different genes in the rodent and humans genomes , and used this to identify rapidly evolving genes @xcite . a similar study scanned for positively selected genes in the genomes of humans and chimpanzees @xcite as well as other mammals @xcite . a more refined approach to identifying selection based on the mcdonald - kreitman test compared the substitution rates of non - synonymous and synonymous changes between the human and chimpanzee genome with the amount of polymorphism of each type observed in the human population @xcite .
while many rapidly evolving genes are involved in immune function , not all immune system genes are rapidly evolving .
cytokine signaling networks associated with t - helper differentiation ( such as th1 and th2 ) are conserved .
for example , the cytokine interferon-@xmath0 ( ifn-@xmath0 ) shows evidence of conservation both in structure and function among vertebrates @xcite and is subject to strong purifying selection @xcite . among murine and human lineages all but one of the type i cytokines ( interleukin 2,3,4,5,6,7,9,11,12,13,15 , etc . ) and cytokine receptors have homologs and are thus strongly conserved @xcite .
most mammalian lineages , including murine and primate lineages , show phylogenetic evidence for purifying selection for interleukin 2 ( il-2 ) @xcite .
the cytokine interleukin 12 ( il-12 ) shows evidence of functional conservation between chickens and higher vertebrates @xcite
. the cytokine interleukin 16 ( il-16 ) shows evidence of both structural and functional conservation between human and murine lineages @xcite .
the cytokines interleukin 1 and 2 ( il-1 and il-2 ) also show evidence of conservation @xcite . these cytokines are all implicated in the th1/th2 control logic of the adaptive immune response .
in addition , mucosal - associated invariant t - cell subsets ( mait ) and natural killer t - cell ( nkt ) subsets are conserved between murine and human lineages @xcite .
finally , none of the above cytokines or cytokine receptors were highlighted by recent screens of the mouse , primate , and mammalian genomes as having fast rates of evolution or evidence of positive selection @xcite .
specifically , among the panther ontology categories highlighted as showing positive selection , cytokine and cytokine receptor categories were not included @xcite .
in contrast a number of genes including those at the major histocompatibility ( mhc ) loci , the antibody complex , and nk cell recognition via the ly49 receptor belong to the set of rapidly evolving genes .
the more rapid evolution of these immune function genes is likely to be due to host - pathogen interactions .
these interactions could give rise to positive selection as a consequence of rapid host - pathogen coevolution @xcite or as a consequence of diversifying selection at these loci .
what can we learn from the conservation of immune cytokine control logic and divergence of other immune structures ?
the rapid evolution of immune genes such as those associated the mhc and antibodies has been explained by pathogen - host co - evolution or selection for diversification , but this does not explain why the genes associated with immune control logic are relatively conserved .
one can envisage a number of hypotheses for the relative conservation of immune control logic .
one possibility is that the complexity of the immune control logic results in fragility ( in a rube goldberg sense ) and this hinders further adaptive evolution of that system @xcite .
another way to describe this is that the immune control logic is at a local maximum on the fitness landscape .
however , we would not expect a rube goldberg mechanism to be broadly robust against pathogens that evolve to disrupt the immune response .
the effectiveness of the immune system to deal with most pathogens allows us to reject this hypothesis .
another possibility is that there simply hasnt been enough time for the vertebrate lineages to diverge in their control logics .
the apparent phylogenetic conservation could be merely an artifact of the timescale at which the pathogens and immune system co - evolve .
however , the too - short - timescale hypothesis is only consistent with phylogenetic conservation of immune control logic , not with phylogenetic divergence of other immune structures such as the mhc . the relatively rapid pace of change at loci such as the mhc suggests that this is not the solution .
finally , we consider the hypothesis that the control logic of the adaptive immune system has the property of defensive complexity .
the defensive complexity hypothesis is consistent with both the conservation of the immune control logic and more rapid evolution in other immune system - related genes .
the conservation of immune control logic follows from the slowing down of the coevolutionary arms race .
divergence of other immune structures across lineages is consistent with selection for diversification or the presence of an arms race that is nt impeded by defensive complexity . in order to model defensive complexity in signaling networks like the cytokine signaling network
, we now turn to game theory and population genetics . to exploit a signaling system ,
an adversary must ( 1 ) construct or disrupt signals used in the system , and ( 2 ) do so in a way that increases its own fitness . in host - pathogen interactions ,
step 1 is often simple .
for example , viruses readily perturb the cytokine signaling network used by the host , by altering gene expression or by producing cytokine mimics and antagonists @xcite .
the latter problem manipulating the signals in advantageous ways may be much harder .
this is the challenge we focus on here .
to do so , we make the _ universal construction assumption _ : the adversary can construct any signal , but does not know what the signals do .
making the universal construction assumption allows us to reformulate the problem of evolving to manipulate the host s control network as a learning problem .
defensive complexity then reduces to non - learnability of the control system by the adversary population .
we consider the two - player game between controller and adversary illustrated in figure 1a . in each instance of the game , a cue contains information about the state of the environment .
the controller aims to transduce this cue into an appropriate response .
to do so , the controller s sensory apparatus detects the cue , and produces internal signals that will trigger the controller s response .
the response is determined by a control logic tuned to accomplish some task @xmath1 .
this control logic is selected from the set @xmath2 of minimal - cost control logics for this task , i.e. , from a set of control logics that all perform optimally on the task @xmath1 .
the adversary aims to alter the controller s response and does so by perturbing the controller s internal signals .
[ fig1 ] we explore a game in which the control logic operates on two signals @xmath3 and @xmath4
. we model the control logic as a simple branching logic on the signals as described in the methods .
a fitness landscape is induced by the example control network as follows : the adversaries attempt tamper by up- or down - regulating the signals @xmath3 or @xmath4 . up - regulation or down - regulation of the signals
cause changes to the response value @xmath5 . for modeling purposes we only need to describe the control logic and its fitness consequences on the adversary as a function of perturbations to the signals .
we can do so in a way that is equivalent to using a particular family of control logics : in the methods we show the correspondence between control logics from this family and control logics with inputs described in terms of perturbations to the signals .
we outline two simple examples of control logics and the fitness landscapes that they induce ; detailed specifications are provided in the methods .
let the fitness of the adversary be 1 when there is no perturbation to the control logic .
first consider a simple signaling network that has the control logic @xmath6 and induces a fitness landscape for the adversary as given by figure 1b .
this fitness landscape is a slope , with each step toward down - regulating @xmath3 and up - regulating @xmath4 being progressively more beneficial for the adversary .
second , contrast this with a control logic @xmath7 that is more complex in that it requires more logical operations per conditional if instruction .
the control logic @xmath7 generates a fitness landscape with fitness valleys as shown in figure 1c . on this landscape , any perturbation of a single signal away from the starting state will decrease the adversary s fitness .
we show that for a generalized version of the more complex control logic , the learning time is exponential in the number of signals .
we consider a control logic to be effectively unlearnable by natural selection if the learning time for this logic is exponential in the number of signals @xmath8 . in this case , the controller can force the learning time to blow up exponentially by adding even a modest number of signals .
the major result of this section is that one can construct an effectively unlearnable control logic , analogous to @xmath9 , that is reasonably simple ( i.e. , has a number of elementary operations linear in @xmath8 , see appendix for more details ) .
formally , there exists a reasonably simple @xmath8-signal control logic that requires a number of generations exponential in @xmath8 to learn .
we prove this theorem in the appendix ; the basic intuition is as follows .
think of the @xmath8-dimensional @xmath10 hypercube where each dimension represents perturbations ( down , none , up ) to one signal .
we establish a control logic by which all corners of the hypercube are global optima and the center is a local optimum .
all other spaces are fitness valleys . in other words ,
we construct a control logic in which global maxima occur only where each and every signal has been altered from its default value . to reach a global optimum ,
an adversary needs to tamper with @xmath8 different signals .
then we show that from the starting place where signals are left unperturbed , the expected number of generations needed to produce one of these beneficial @xmath8-mutants is exponential in @xmath8 .
this construction is just one simple example .
more complicated fitness landscapes could lead pathogen populations on detours through a sequence of local maxima , delaying convergence to the global maximum .
in his original letter to the nsa , john nash conjectured that one could construct a cryptographic system with @xmath8 complexly interacting parts would take time exponential in @xmath8 to subvert .
nash felt this conjecture was of critical importance for cryptography , because by judicial use of defensive complexity encoders will be able to slow the arms race between with decoders .
here we have argued that nash s conjecture is important for biological scenarios of antagonistic coevolution as well , because defensive complexity can slow the coevolutionary arms race between pathogens and the adaptive immune systems of vertebrates of their vertebrate hosts .
our framework suggests that the kinds of signaling networks present in the immune system induce a complex fitness landscape with valleys and local maxima for pathogens attempting to deceive the immune system .
our results are consistent with two broad observations pertaining to immunology .
first , we have cited evidence in mammalian lineages of phylogenetic conservation for immune control logic and positive selection for other aspects of immune functions .
this relative conservation is consistent with the predictions of the defensive complexity model ; a sufficiently complex immune circuitry would limit the extent to which rapidly evolving viral and bacterial pathogens provoke coevolutionary arms races that in turn would drive divergence between the immune systems of mice and men .
second , immunologists have found it considerably difficult to decipher the rules behind the functioning of the immune system .
defensive complexity might be expected to involve complex rules which are difficult for pathogens to exploit and immunologists to understand .
we implement the control logics for our example networks using kleene s three - valued logic ; for truth tables in kleene s logic , see ref . @xcite .
we denote the perturbations to signal @xmath11 with @xmath12 .
each perturbation can take on values @xmath13 , @xmath14 or @xmath15 .
when instantiating the control logic , we interpret @xmath13 as @xmath16 , @xmath14 as @xmath17 , and @xmath15 as an input being absent or unknown .
for example , in the not ( @xmath18 ) operation , when given @xmath15 as input , the output is also @xmath15
. the and operator will evaluate to true only if both its inputs are true .
the or operator will evaluate to true only if at least one of its inputs is true . after the branching program terminates which happens when a particular conditional evaluates to true
then the rest of the conditionals can not be triggered , and so the order of specification for the conditionals matters a great deal .
we use the notation @xmath19 for the and logic gate , we use @xmath20 for the or logic gate , @xmath21 for xor , and we use @xmath18 for not . if @xmath22 is the fitness of the adversary then we can implement the control logic @xmath6 as follows : 1 . if @xmath23 , @xmath24 2
if @xmath25 , @xmath26 3 .
if @xmath27 , @xmath28 4 .
if @xmath29 , @xmath30 5 .
else @xmath31 we can implement control logic @xmath7 as follows : 1 . if @xmath32 , @xmath33 2 . if @xmath34 , @xmath35 3 .
if @xmath36 , @xmath24 4 . else @xmath37 we have thus far discussed control logics in terms of perturbations to signals , but this is only a way of simplifying the full control logic for the purpose of discussion .
in fact , we can construct control logics equivalent to the perturbation - based control logics based on the following substitutions .
we say that @xmath15 corresponds to the default value @xmath38 that is set by the control logic to signal @xmath11 based on the value of the cue @xmath39 ( more explicitly , by the sensor , based on the value of @xmath39 ) , with @xmath13 and @xmath14 corresponding to any quantity larger than @xmath38 and smaller than @xmath38 respectively .
consider @xmath40 and @xmath41 , which satisfy the inequality @xmath42 for all @xmath11 .
we can translate each gate over perturbations into a gate over the signals and the original cue .
for @xmath43 we have @xmath44 for @xmath45 ( with a ternary and gate ) , we have @xmath46 where @xmath19 is the ordinary boolean and gate . for @xmath47
, we have @xmath48 . for @xmath49 , we have @xmath50 where @xmath20 is the ordinary boolean or gate .
this work was funded in part by the midas center for communicable disease dynamics nih 5u54gm088558 - 02 and by nsf grant ef-1038590 ( ctb ) , and nih r01 ai049334 and midas u01-gm070749 ( ra ) .
the authors also wish to thank daril vilhena , philip johnson and frazer meacham for helpful comments .
10 r. dawkins and j.r .
arms races between and within species . , 205(1161):489511 , 1979 .
bergstrom and r. antia .
how do adaptive immune systems control pathogens while avoiding autoimmunity ?
, 21(1):2228 , 2006 .
p. schmid - hempel .
variation in immune defence as a question of evolutionary ecology .
, 270(1513):357 , 2003 .
l. van valen . a new evolutionary law .
, 1(1):130 , 1973 .
j. nash .
letter to the national security agency . declassified recently by the nsa , 1955 .
rivest , a. shamir , and l. adleman . a method for obtaining digital signatures and public - key cryptosystems . , 21(2):120126 , 1978 .
bergstrom and m. lachmann .
the red king effect : when the slowest runner wins the coevolutionary race .
, 100(2):593 , 2003 .
j.a . damore and j.
a slowly evolving host moves first in symbiotic interactions . , 2011 .
kindt , r.a .
goldsby , b.a .
osborne , and j. kuby . .
wh freeman , 2007 .
z. frankenstein , u. alon , i.r .
cohen , et al .
the immune - body cytokine network defines a social architecture of cell interactions .
, 1(32):115 , 2006 .
chinwalla , l.l .
cook , k.d .
delehaunty , g.a .
fewell , l.a .
fulton , r.s .
fulton , t.a .
graves , l.d.w .
hillier , e.r .
mardis , j.d .
mcpherson , et al .
initial sequencing and comparative analysis of the mouse genome .
, 420(6915):520562 , 2002 .
bustamante , a. fledel - alon , s. williamson , r. nielsen , m.t .
hubisz , s. glanowski , d.m .
tanenbaum , t.j .
white , j.j .
sninsky , r.d .
hernandez , et al .
natural selection on protein - coding genes in the human genome .
, 437(7062):11531157 , 2005 .
r. nielsen , c. bustamante , a.g .
clark , s. glanowski , t.b .
sackton , m.j .
hubisz , a. fledel - alon , d.m .
tanenbaum , d. civello , t.j .
white , et al .
a scan for positively selected genes in the genomes of humans and chimpanzees .
, 3(6):e170 , 2005 . c. kosiol , t. vinar , r.r . da fonseca , m.j .
hubisz , c.d .
bustamante , r. nielsen , and a. siepel .
patterns of positive selection in six mammalian genomes . , 4(8):e1000144 , 2008 .
r. savan , s. ravichandran , j.r .
collins , m. sakai , and h.a .
young . structural conservation of interferon gamma among vertebrates .
, 20(2):115124 , 2009 .
j. manry , g. laval , e. patin , s. fornarino , m. tichit , c. bouchier , l.b .
barreiro , and l. quintana - murci .
evolutionary genetics evidence of an essential , nonredundant role of the ifn-@xmath0 pathway in protective immunity . , 2011 .
boulay , j.j .
oshea , and w.e .
molecular phylogeny within type i cytokines and their cognate receptors . , 19(2):159163 , 2003 .
d. zelus , m. robinson - rechavi , m. delacre , c. auriault , and v. laudet .
fast evolution of interleukin-2 in mammals and positive selection in ruminants .
, 51(3):234244 , 2000 .
degen , n. van daal , h.i .
van zuilekom , j. burnside , and v.e.j.c .
schijns . identification and molecular cloning of functional chicken il-12 . , 172(7):4371 , 2004 .
j. keane , j. nicoll , s. kim , d.m.h .
cruikshank , w. brazer , b. natke , y. zhang , d.m . center , and h. kornfeld .
conservation of structure and function between human and murine il-16 .
, 160(12):5945 , 1998 .
n. cohen and l. haynes .
the phylogenetic conservation of cytokines .
, pages 24268 , 1992 .
e. treiner , l. duban , i.c .
moura , t. hansen , s. gilfillan , and o. lantz .
mucosal - associated invariant t ( mait ) cells : an evolutionarily conserved t cell subset .
, 7(3):552559 , 2005 .
l. brossay , m. chioda , n. burdin , y. koezuka , g. casorati , p. dellabona , and m. kronenberg .
cd1d - mediated recognition of an @xmath51-galactosylceramide by natural killer t cells is highly conserved through mammalian evolution .
, 188(8):1521 , 1998 .
fraser , a.e .
hirsh , l.m .
steinmetz , c. scharfe , and m.w .
. evolutionary rate in the protein interaction network .
, 296(5568):750 , 2002 .
l.r . gooding .
virus proteins that counteract host immune defenses . , 71(1):5 , 1992 .
peter palese , editor . .
springer , 2005 .
kleene , ng de bruijn , j. de groot , and a.c .
introduction to metamathematics .
the synthesis of two - terminal switching circuits .
, 28(1):5998 , 1949 .
fisher et al . on the dominance ratio . , 42:321341 , 1922 .
. haldane . a mathematical theory of natural and artificial selection , part v : selection and mutation . in _ mathematical proceedings of the cambridge philosophical society _ ,
volume 23 , pages 838844 .
cambridge univ press , 1927 .
rb campbell . a logistic branching process for population genetics .
, 225(2):195203 , 2003 .
f. gao , l. han , and k. schilling .
on the rate of convergence of iterated exponentials . ,
pages 18 , 2011 .
cp farrington and ad grant .
the distribution of time to extinction in subcritical branching processes : applications to outbreaks of infectious disease . , 36(3):771779 , 1999 .
before proving the main theorem , we first specify precisely what we mean by a `` reasonably simple '' control logic . a control logic can be described as a set @xmath52 of functions
one for each conditional in the branching logic .
each of these functions @xmath53 in @xmath52 takes some input and returns either a 1 ( if the formula specifying the conditional evaluates to true ) or 0 ( otherwise , see methods ) . in the example of section 2 , the control logic would take the signals @xmath3 and @xmath4 as inputs , and determine the appropriate response by evaluating each conditional .
we can quantify circuit complexity as follows .
the circuit complexity of branch @xmath53 of the control logic is simply the minimum number of ternary logic gates ( see methods ) needed to implement @xmath53 @xcite .
the circuit complexity of the full logic @xmath52 is the maximum circuit complexity over @xmath53 in @xmath52 .
the circuit complexity of a reasonably simple control logic is @xmath54 . to cross the valley in @xmath55 dimensions ,
the individual needs to found a lineage that accumulates at least @xmath55 successive mutations and thus this lineage must survive in the fitness valley for at least @xmath55 generations .
alternatively , a lineage could pick up multiple mutations in each generation but this wo nt help . consider the case that @xmath56 mutants arise each time step , and the lineage must survive for at least @xmath57 generations .
when @xmath55 grows very large and @xmath56 is constant , asymptotic analysis tells us that the individual needs to found a lineage that survives for at least @xmath58 generations ( because @xmath59 and @xmath60 ) . in this case , generating @xmath56-mutants each time - step does nt speed up crossing the fitness valley appreciably .
but what @xmath56 is allowed to grow with @xmath55 ?
for example , let @xmath61 , where @xmath62 . in that case we would only require a lineage to survive for @xmath63 steps , and this is independent of @xmath55 .
but the time that it would take to generate the @xmath56 mutant required for each successive step is on average @xmath64 where @xmath65 is the mutation rate .
there is no escape from an average time exponential in @xmath55 .
it is therefore sufficient to determine how likely an individual who steps down into the valley is to found a lineage that survives in the valley for at least @xmath55 generations , as @xmath55 grows very large .
if this probability decreases exponentially with @xmath55 , the average time until we get the first individual who is destined to succeed will increase exponentially in @xmath55 . to model the fate of such a lineage , we approximate it as a subcritical branching process @xcite . if the relative fitness of an individual in the valley is @xmath66 , this individual will have a poisson number of offspring with mean @xmath67 .
each of these offspring will themselves have poisson numbers of offspring , again with mean @xmath68 .
first we prove a technical lemma , and then we show that the average time until we get the first individual who is destined to succeed increases exponentially in @xmath55 as @xmath55 grows large .
first we use the result ( theorem 2 in ref .
@xcite ) that since @xmath73 , @xmath74 $ ] converges to @xmath75 at a linear rate @xmath67 .
thus for any @xmath8 , we have @xmath76 - e^{\lambda}| = o ( \lambda |f[e^{\lambda e^{-\lambda}},n ] - e^{\lambda}|)$ ] . from the above , terminating with @xmath55 and solving the recurrence relation ,
we obtain : @xmath77 - e^{\lambda}| = \lambda^{o(k ) } |f[e^{\lambda e^{-\lambda}},0 ] - e^{\lambda}|,\ ] ] and so by the definition of tetration , @xmath78 - e^{\lambda}| = \lambda^{o(k ) } |1 - e^{\lambda}|$ ] .
the symmetry property of absolute value implies that @xmath79| = \lambda^{o(k ) } |1 - e^{\lambda}|$ ] multiply both sides by @xmath80 to obtain : @xmath81| = \lambda^{o(k ) } |e^{-\lambda}-1|\ ] ] since for @xmath82 , @xmath83 . because @xmath84 $ ] is the cumulative distribution function ( cdf ) of some distribution @xcite , this implies that @xmath85 .
therefore , @xmath86 , and in particular , @xmath87 .
consequently we can simplify the absolute value as follows : @xmath88 = \lambda^{o(k ) } |e^{-\lambda}-1|\ ] ] and thus this establishes the lemma .
< 1.5em - 1.5em plus0em minus0.5em height0.75em width0.5em depth0.25em [ timebound ] as @xmath55 grows large , the average time until we get the first individual that is destined to succeed assuming its lineage is modeled by a subcritical poisson branching process ( with average number of offspring @xmath89 ) is at least @xmath90 generations on average . by the subcriticality of the branching process that generates the lineage , @xmath73 .
the random variable @xmath91 is the number of generations the lineage generated by the branching process survives .
the cdf of the branching process @xmath92 gives us the probability of a lineage surviving no more than @xmath8 generations .
therefore , for our purposes , we need to characterize the probability of non - extinction for @xmath55 generations , which means we must characterize @xmath93 . by equation 2.5 in farrington and
grant @xcite , when @xmath94 , @xmath95 $ ] , where @xmath71 $ ] is the tetration function .
therefore , by lemma [ powertower ] , @xmath96 .
the number of mutants who must enter the valley before one does so successfully is on average @xmath97 , which is thus at least @xmath98 since @xmath99 .
since the rate at which mutants are produced from the wild type each generation is upper bounded by @xmath100 , the result follows .
< 1.5em - 1.5em plus0em minus0.5em height0.75em width0.5em depth0.25em by our definition above of `` reasonably simple '' , we aim to show that there exists a control logic on @xmath8 signals with circuit complexity @xmath54 that is learnable in a number of generations exponential in @xmath8 .
the adversary has three possible actions , upregulate , downregulate , or do nothing .
we will now build a control logic that downregulates @xmath5 only if the adversary upregulates or downregulates each and every one of the signals ( rather than doing nothing ) , and otherwise upregulates @xmath5 .
we will then show that for small enough adversary populations , generating an @xmath8-mutant takes time exponential in @xmath8 . where @xmath108 and @xmath109 .
here @xmath110 represents the logical equivalence operator in kleene s logic . generating an optimal type requires a beneficial @xmath8-mutant with deleterious intermediates , and by lemma [ timebound ] , this takes at least on average @xmath111 generations , for @xmath89 .
since @xmath67 is the relative fitness of an individual in the fitness valley , @xmath112 by line 2 of @xmath102
. now note that @xmath102 on the first line has @xmath8 clauses of the form @xmath113 , each of which has 1 gate , and connected by @xmath114 or gates . in total
then there are @xmath115 gates on the first line of @xmath102 ( and this is maximal ) .
the result follows .
< 1.5em - 1.5em plus0em minus0.5em height0.75em width0.5em depth0.25em | one strategy for winning a coevolutionary struggle is to evolve rapidly .
most of the literature on host - pathogen coevolution focuses on this phenomenon , and looks for consequent evidence of coevolutionary arms races .
an alternative strategy , less often considered in the literature , is to deter rapid evolutionary change by the opponent . to study how this can be done ,
we construct an evolutionary game between a controller that must process information , and an adversary that can tamper with this information processing . in this game ,
a species can foil its antagonist by processing information in a way that is hard for the antagonist to manipulate .
we show that the structure of the information processing system induces a fitness landscape on which the adversary population evolves .
complex processing logic can carve long , deep fitness valleys that slow adaptive evolution in the adversary population .
we suggest that this type of defensive complexity on the part of the vertebrate adaptive immune system may be an important element of coevolutionary dynamics between pathogens and their vertebrate hosts .
furthermore , we cite evidence that the immune control logic is phylogenetically conserved in mammalian lineages .
thus our model of defensive complexity suggests a new hypothesis for the lower rates of evolution for immune control logic compared to other immune structures . |
the results reported here derive from a numerical analysis of orbits in the triaxial dehnen potentials , where ( _ cf . _
merritt & fridman 1996 ) @xmath0 with @xmath1 assuming fixed axis ratios @xmath2 and @xmath3 but allowing for a variable cusp index @xmath4 and a variable black hole mass @xmath5 .
different segments of the same chaotic orbit can be extremely different in terms of their visual appearance and their degree of exponential sensitivity ( siopis & kandrup 2000 ) .
these differences can be quantified in terms of the sizes of short time lyapunov exponents ( _ cf .
_ kandrup & mahon 1994 ) or the _ complexity _ of their fourier spectra , _
i.e. , _ the degree to which the power in an orbit is concentrated near a few special frequencies ( _ cf .
_ kandrup , eckstein , & bradley 1997 , siopis , eckstein , & kandrup 1998 ) .
chaotic orbits typically have continuous spectra , but when they look ` nearly regular ' most of the power is concentrated near a few special frequencies . in agreement with intuition , there is a strong correlation between the complexity of an orbit segment and the value of its largest short time lyapunov exponent : chaotic orbit segments which look ` nearly regular ' and have less complex spectra tend also to exhibit comparatively small exponential sensitivity . if two chaotic orbits in the same connected phase space region are integrated for a sufficiently long time , it appears that they will eventually share the same statistical properties .
however , the time required for this can be extremely long . if , for example , a single chaotic initial condition is integrated into the future , it can take as long as @xmath6 dynamical times @xmath7 , or even longer , before the short time lyapunov exponent exhibits a reasonable convergence towards the true lyapunov exponent @xmath8 , as defined in a @xmath9 limit .
the overall rate of convergence can be quantified through an examination of distributions of short time lyapunov exponents , @xmath10 $ ] , generated for ensembles of chaotic orbit segments of varying length @xmath11 . in the absence of any significant stickiness ,
the dispersion associated with @xmath12 $ ] scales as @xmath13 with @xmath14 .
for very sticky orbits , @xmath15
. because chaotic orbits are so sticky , one might anticipate that they could be used as building blocks for the construction of self - consistent near - equilibria which , albeit not strictly time - independent , behave as nearly time - independent entities over time intervals long compared with @xmath16 , the age of the universe .
( in the language of merritt & fridman [ 1996 ] , these would be ` quasi - equilibria ' involving stochastic building blocks that are only ` partially mixed . ' ) however , this supposition relies crucially on the assumption that the statistical properties of chaotic orbit segments are relatively insensitive to the effects of weak perturbations of the form which act on real galaxies . in point of fact
, this does not appear to be the case .
orbits in the unperturbed triaxial dehnen potential were perturbed to mimic various effects to which real stars in real galaxies are typically exposed .
discreteness effects , _
i.e. , _ gravitational rutherford scattering between individual stars , were modeled as dynamical friction and white noise , _
i.e. , _ near - instantaneous kicks .
the effects of one or two companion objects or satellite galaxies were modeled as nearly periodic perturbations .
the effects of a dense cluster environment were modeled as coloured noise , _
i.e. , _ random kicks of finite duration . internal oscillations of the form that might , _ e.g. , _ be triggered by a close encounter were treated as a superposition of normal , or pseuo - normal , modes that induced a periodic driving and an incoherent combination of more irregular excitations modeled as coloured noise . the basic conclusion of this investigation ( siopis & kandrup 2000 , kandrup & siopis 2000 ) , consistent also with analyses of motions in other two- and three - dimensional potentials ( pogorelov & kandrup 1999 , kandrup , pogorelov , & siopis 2000 ) , is that low amplitude irregularities can have a surprisingly large effect both @xmath17 by accelerating diffusion within a given nearly disjoint phase space region ; and @xmath17 by accelerating diffusion along an arnold web or through cantori connecting nearly disjoint chaotic phase space regions .
some of the topological obstructions associated with the arnold web are extremely robust , so that weak perturbations have a comparatively minimal effect . however , in general such perturbations tend to accelerate dramatically the rate of phase space transport throughout the entire chaotic phase space . the perturbations act via a resonant coupling between the characteristic frequencies of the perturbations and the frequencies of the orbits . that periodic driving works in this way should be obvious .
that noise also involves a resonant coupling can be understood if one recalls ( _ cf . _ van kampen 1981 ) that a superposition of periodic forces combined with random phases is equivalent mathematically to ( in general coloured ) noise with a nonzero autocorrelation time @xmath18 . within a nearly disjoint phase space region ,
the perturbations allow microscopic motions which , in a strictly time - independent potential , are prohibited by liouville s theorem .
e.g. , _ it becomes possible for phase space trajectories to cross , which helps a collection of orbits to ` fuzz out ' on short scales .
the perturbations facilitate diffusion through cantori or along the arnold web by ` jiggling ' orbits in such a fashion as to help them find phase space holes .
the details of the perturbation appear largely immaterial : all that seems to matter is the amplitude of the perturbations and their characteristic time scales .
even the dependence on amplitude and time scale is comparatively weak .
this implies that the details associated with realistic perturbations which might be difficult to extract from observations are largely irrelevant .
the overall efficacy of the perturbations scales logarithmically in the amplitude . for time scales
@xmath19 the perturbations have almost no effect ( adiabatic limit ) . for somewhat shorter time scales ,
the dependence on @xmath18 is again logarithmic .
but what amplitude is required to have a significant effect , _
e.g. , _ by destabilising nearly time - independent building blocks ? very weak white noise corresponding to relaxation times @xmath20 can have appreciable effects within a time as short as @xmath21 , a period which , in the inner regions of a cuspy triaxial galaxy would be short compared with @xmath16 .
alternatively , coloured noise and/or periodic driving corresponding to perturbations of fractional amplitude as small as @xmath22 and a characteristic time scale @xmath18 as long as @xmath23 can prove important on a time scale @xmath24 . making @xmath18 shorter facilitates a stronger resonant coupling between the perturbation and the orbits , thus making the perturbation even more effective .
these results suggest the possibility that galaxies could settle down towards quasi - stationary states which , albeit not true collisionless equilibria , could exist as nearly time - independent entities for times @xmath25 , at least in the absence of irregularities .
this seems especially likely , given the recognition that , for triaxial systems , true equilibria will in general be substantially more complex than the equilibria associated with spherical and axisymmetric configurations . for a generic triaxial system , there is only one global integral , namely the energy @xmath26 or jacobi integral @xmath27 , but it well known that equilibria @xmath28 and @xmath29 can not be used to model triaxial systems with a strong central condensation . unless the system is assumed to be characterised by a very special potential , _
e.g. , _ an integrable staeckel potential , it can not be in a true equilibrium unless that equilibrium involves an intricate balance of ` local integrals ' ( kandrup 1998 ) .
the obvious point , then , is that even if such an intricate balance is hard to achieve , the system could evolve towards an approximate balance involving nearly time - independent building blocks .
more pragmatically , these results would also suggest that , when constructing equilibria using schwarzschild s method or any analogue thereof , it would be strongly advisable to work with an orbit library constructed from orbits that have been evolved in the presence of weak noise or some other low amplitude perturbations .
orbits evolved in the presence of such perturbations are more likely to constitute nearly time - independent building blocks and , as such , would seem less likely to be destabilised by weak irregularities associated with discreteness effects and/or a perturbing external environment .
numerical computations demonstrate that much of the behaviour associated with chaotic orbit ensembles evolved in the triaxial dehnen potentials especially those associated with orbits which , in the absence of a cusp and black hole , would correspond to regular box orbits can be reproduced by the very simple potential @xmath30 given as the sum of an anisotropic oscillator and a plummer potential .
orbits in this potential exhibit the same remarkable stickiness , yield comparable distributions of short time lyapunov exponents , and again manifest a strong susceptibility towards even very weak perturbations ( kandrup & sideris 2000 ) . that this simple toy model can reproduce the qualitative features of the more complicated potential ( 1 ) suggests strongly that _ the results derived for chaotic orbits in the triaxial dehnen potentials are generic for cuspy triaxial potentials_. that the potential is
so simple makes it comparatively easy to understand what exactly is going on . as noted , _
e.g. , _ by merritt ( 1998 ) , supermassive black holes in real galaxies are seldom if ever larger than @xmath31 the mass of the entire galaxy and the central cuspy region typically corresponds to only a small fraction of the total mass .
it follows that , for @xmath32 , @xmath33 , and @xmath34 of order unity , the physically relevant choices of @xmath35 entail @xmath36 .
however , the qualitative behaviour in this regime is easily understood by a combination of perturbation theory and common sense . for @xmath36 and @xmath37 in eq .
( 2 ) , it appears that , except for the very lowest energies ( where the potential is essentially keplerian ) , essentially all of the orbits are chaotic , but that they tend to behave in a nearly regular fashion nearly all of the time . as the orbit evolves it will usually find itself in a region where @xmath38 , so that the potential in which it is moving is very nearly integrable and the short time lyapunov exponents are extremely small
occasionally , however , the orbit will move comparatively close to the center of the galaxy , so close that @xmath39 becomes comparable to @xmath40 . when this happens the orbit feels the competing influences of two different potentials of comparable magnitude with very different symmetries and the values of the positive short time lyapunov exponents increase precipitously .
the fact that , for small @xmath41 , almost all the orbits are chaotic is not difficult to understand . in the limit
that @xmath42 , the orbits all reduce to boxes which densely fill a region in configuration space that includes the origin .
one might expect that , for small but nonzero @xmath41 , the orbits can still pass arbitrarily close to the origin but , for any nonzero @xmath41 there is a minimum radius @xmath43 inside of which @xmath39 becomes large compared with @xmath40 .
that significant chaos is only triggered when the trajectory passes relatively close to the black hole is illustrated in figure [ fig-1 ] , which exhibits a segment of a chaotic orbit evolved in the potential ( 2 ) .
curve in the top panel exhibits the phase space separation @xmath44 between the original orbit and a perturbed orbit displaced originally by a distance @xmath45 and periodically renormalised in the usual way ( _ cf . _
lichtenberg & lieberman 1992 ) .
the dashed curve shows an analogous plot of @xmath44 for a regular orbit .
the lower panel plots @xmath46 , the distance from the origin . most of the time the perturbed and unperturbed orbits remain very close together , with comparatively little systematic exponential divergence . only when @xmath47 becomes as small as @xmath48 , so that @xmath39 becomes as large as @xmath49 , do the orbits tend to diverge significantly .
this toy model is particularly simple since , in the limit @xmath50 , all the orbits are regular boxes . for generic triaxial potentials , in the absence of a cusp or black hole one would expect both centrophilic box orbits and centrophobic tubes . because the tubes are centrophobic ,
they should not in general be impacted all that much by the introduction of a cusp or a central black hole .
what , _ does _ , however , seem to be true is that many of the orbits which , in the absence of a cusp , behave as regular boxes can , in the presence of a cusp or black hole , be converted into orbits which , albeit formally chaotic , behave in a nearly regular fashion much of the time .
we are pleased to acknowledge useful collaborations with brendan bradley , barbara eckstein , salman habib , and elaine mahon .
this research was supported in part by nsf ast-0070809 and by the institute for geophysics and planetary physics at los alamos national laboratory .
kandrup , h. e. 1998 , , 299 , 1139 kandrup , h. e. , eckstein , b. l. , bradley , b. o. 1997 , a&a , 320 , 65 kandrup , h. e. , mahon , m. e. 1994 , a&a , 290 , 762 kandrup , h. e. , pogorelov , i. v. , sideris , i. v. 2000 , , 311 , 719 kandrup , h. e. , sideris , i. v. 2000 , , submitted kandrup , h. e. , siopis , c. 2000 , in preparation lichtenberg , a. j. , lieberman , m. a. 1992 , regular and chaotic dynamics .
springer : berlin merritt , d. 1998 , comments astrophys . , 19 , 1 merritt , d. , fridman , t. 1996 , , 460 , 136 pogorelov , i. v. , kandrup , h. e. 1999 , phys . rev .
e , 60 , 1567 siopis , c. , kandrup , h. e. 2000 , , in press siopis , c. , eckstein , b. l. , kandrup , h. e. 1998 , ann . n. y. acad .
, 867 , 41 van kampen , n. g. 1981 , stochastic processes in physics and chemistry .
north holland : amsterdam | this talk provides a progress report on an extended collaboration which has aimed to address two basic questions , namely : should one expect to see cuspy , triaxial galaxies in nature ? and can one construct realistic cuspy , triaxial equilibrium models that are robust ?
three technical results are described : ( 1 ) unperturbed chaotic orbits in cuspy triaxial potentials can be extraordinarily sticky , much more so than orbits in many other three - dimensional potentials .
( 2 ) even very weak perturbations can be important by drastically reducing , albeit not completely eliminating , this stickiness .
( 3 ) a simple toy model facilitates a simple understanding of why black holes and cusps can serve as an effective source of chaos .
these results suggest that , when constructing models of galaxies using schwarzschild s method or any analogue thereof , astronomers would be well advised to use orbital building blocks that have been perturbed by ` noise ' or other weak irregularities , since such building blocks are likely to be more nearly time - independent than orbits evolved in the absence of all perturbations .
a contributed talk at:_the international conference on stellar dynamics : from _ .2 in _ from classical to modern _ , sobolev astronomical institute , august 2000 .2 in |
recent and ongoing advances in technology have led to the discovery of extrasolar planets ( e.g. , mayor & queloz 1995 , marcy _ et el .
_ 1997a , b , butler & marcy 1996 , marcy & butler 1996 , cochran _ et al .
_ 1997 , noyes _
_ 1997 ; see also the references listed in the encyclopedia of extrasolar planets , www.obspm.fr/darc/planets/encycl.html ) , and promise the discovery and even the imaging of additional planets ( angel & woolf 1997 ; fraclas & shelton 1997 ; labeyrie 1996 ; brown 1996 ) .
these developments excite the imagination because they seem to bring us closer to the possible discovery of extraterrestrial life .
it is therefore interesting to ask whether microlensing searches are likely to find planets on which life could thrive .
we do not yet have a clear enough understanding of the nature of life to definitively answer this question , because the range of physical conditions compatible with life may well be wider than our limited experience would at first suggest
. it has been proposed , for example , that life may exist on the outer planets of our own solar system and/or on their moons .
( see , e.g. , reynolds et al .
1983 ; raulin _ et al .
_ 1992 ; sagan , thompson , & khare 1992 ; williams , kasting , & wade 1997 , mccord _ et al .
it has even been postulated that life may exist in non - planetary environments , including the interiors of stars and molecular clouds ( see , e.g. , feinberg & shapiro 1980 ) .
it therefore makes sense to consider any planet , however close to or far from its star , and whatever the nature of the star , as a possible harbor for life . nevertheless , in the absence of real information on the existence of life away from our own planet , one question is clearly interesting : will microlensing find evidence of planets similar to earth ?
we must of course define what we mean by similar to earth " .
if we mean that there is a chance that chemical processes necessary for earth - like life could occur , then we want to consider planets which can have similar surface and atmospheric make - up , and similar amounts of energy available to fuel the necessary chemical processes .
the range of planetary masses and distances from a solar - type star compatible with earth - like conditions , particularly the presence of liquid water , has been dubbed the goldilocks problem " , and has been studied by many researchers .
( see , e.g. , rampino & caldeira 1994 , for an overview . ) in linking the goldilocks problem to microlensing observations , we must focus on the properties of ps .
s that determine their detectability when searched for by microlensing monitoring programs .
2 is therefore devoted to a brief overview of the detection of planets via microlensing . the question of whether planets discovered through their action as microlenses may be earth - like , is addressed in 3 .
one of the primary findings of this study is that when a planet discovered via microlensingreceives from its central star a flux of radiation comparable to that received by the earth from the sun , the central star will generally be more luminous than our sun .
this means that light from the lensed star will be blended with light from the ps . .
4 is devoted to studying how blending ( a ) influences the detection of , and ( b ) helps us develop strategies to study ps .
s discovered through microlensing . in 5 ,
i summarize the conclusions .
the detection of a planet via microlensing is possible if the separation between the star and planet is close to or larger than the einstein radius , @xmath1 of the star : @xmath2 where @xmath3 is the stellar mass , @xmath4 is the distance to the lensed source , and @xmath5 is the ratio of the distance to the lens to @xmath4 . consider lensing by a mass , @xmath6 .
define @xmath7 to be the time taken for the track of a lensed star to cross a distance in the lens plane equal to the einstein diameter , @xmath8 .
note that the duration of the observable event generally differs from @xmath7 for several reasons .
first , if the photometric sensitivity is good , the event may be detectable well before the track of the source comes within @xmath9 of the lens position .
for example , the deviation from baseline is at the @xmath10 level , when the projected separation between the source and lens is @xmath11 second , the track of the source will not generally intersect the lens position . finally , blending may decrease and/or finite - source size effects may increase the time during which the event is detectable .
( see distefano 1998 ; distefano & scalzo 1997 or 1998a for discussions relevant to planet lenses . )
express the separation between the planet and star as @xmath12 .
the minimum possible value of @xmath13 for which planets can be discovered via microlensing is @xmath14 ( gould & loeb 1992 , but see wambsganss 1997 for a detailed discussion ) . for planets located between roughly @xmath15 and @xmath16 ( the _ zone for resonant lensing _ ) , the signature of a lensing planet is a short - lived perturbation superposed on a light curve whose underlying structure is due to lensing by the ps .
the time duration of the underlying event may be weeks or months .
all of the literature published to date on planet discovery via microlensing has been limited to the possible discovery of planets in the zone for resonant lensing , and observing programs designed to increase the detection efficiencies for these types of events are already underway . for a more detailed description of these short - duration , resonant " perturbations , and a discussion of the boundaries of zone for resonant lensing see , e.g. , mao & paczyski 1991 , gould & loeb 1992 , bennett & rhie 1996 , paczyski 1996 , wambsganss 1997 , peale 1997 .
the benefits of systematically extending the search , to look for planets located farther from the central star , have recently begun to be explored ( distefano & scalzo 1997 , 1998a , b , distefano 1998 ) . for larger values of @xmath17 ,
the planet generally acts as an independent lens .
when the planet is the only lens , then the event will be an isolated event of short - duration .
( for a jupiter - mass planet , e.g. , the time duration of the perturbation from baseline lasts @xmath18 of the time the deviation due to a solar - mass star would . )
when the track of the source passes through the lensing region of one planet and also that of another system mass ( typically the central star ) , the event will appear to repeat ( distefano & mao 1996 , distefano & scalzo 1997 , 1998a , b ) .
there is no maximum value of @xmath13 , other than that dictated by the dynamics of the ps .
itself i.e . ,
planets located too far from the central starmay be lost from the system . because the probability of a repeating event falls off as @xmath19 , where @xmath17 is the orbital separation , while the probability of isolated short - duration events is nearly independent of position ( and actually increases at the expense of repeating events as the orbital separation increases ) , isolated short - duration events become the dominant mode of detection for planets in wider orbits particularly since ps .
s may have several planets in s ( distefano & scalzo 1997 , 1998a ) .
calculations indicate that it is likely that microlensing by planets in s will provide an important channel and , for low - mass ( e.g. , earth - mass ) planets , possibly the dominant channel for planet detection via microlensing ( distefano & scalzo 1997 , 1998a , b distefano 1998 ) .
the phrase earth - like conditions " does not have a unique meaning .
two requirements seem natural , however .
( 1 ) the radiation flux received from the central star should be neither too large nor too small ; ( 2 ) the planet should have a rocky surface , water , and a gaseous atmosphere . for earth - like conditions to exist , the primary requirement on @xmath20 the orbital separation between the planet and the central star , is that the incident flux of radiation from the star should be comparable to the flux received by the earth from the sun .
that is , @xmath21 , should not be too different from unity .
@xmath22 the mass and luminosity of the central star are @xmath3 and @xmath23 , respectively . as usual , @xmath4 ( @xmath24 )
is the distance to the lensed source ( lens ) , and @xmath25 . given that ( 1 ) the conditions that lead to life may be flexible , ( 2 ) the effects of radiation incident from the star are likely to be strongly influenced ( either enhanced or diminished ) by the planet s atmosphere , and ( 3 ) internal heating from geological processes or radioactive materials may be important , it is not clear how large a range of values of @xmath21 may be compatible with the development of life .
we will therefore simply use @xmath26 as a guideline , and emphasize that this should not be viewed as an absolute requirement .
figure 1 shows the relationship between @xmath3 and @xmath5 for those systems that satisfy the relationship @xmath27 we have set @xmath28 and have assumed that @xmath29 in the upper plot , and @xmath30 in the bottom plot ; the former would be appropriate for main - sequence stars , while the latter would be appropriate for slightly evolved stars . as eq
. 1 and figure 1 make clear , if the planets we will discover via microlensing are to have incident flux comparable to the flux incident on earth , their stars will generally ( although not necessarily ) be more luminous than our sun .
this has two obvious implications .
the first is that the length of time during which the planet would have this flux incident will be shorter than the time to date that the earth has had roughly this flux incident .
this is because the system s star may need to be more massive than the sun , or even slightly evolved .
the time elapsed from the formation of the star until the present time could range from less than @xmath31 the present age of the sun to times comparable to the sun s main - sequence lifetime .
we do not know how long it takes for complex life forms to develop , but it may be that the process is fast enough that intelligent life can develop and thrive during a time significantly shorter than the present age of the sun . indeed , it is likely that a long sequence of independent processes must occur in order for intelligent life to develop ; thus , the probability distribution may well be log normal , and the likelihood of intelligent life developing in times much shorter than the time apparently taken on earth may be significant .
the second implication is that , since the central star must be fairly luminous , it may contribute a non - negligible fraction of the light incident along the line of sight to the lensed source ; i.e. , the light we receive may be strongly blended .
we may also argue , as follows , that planets likely to be deemed earth - like have masses within a factor of @xmath0 of the mass of the earth . assuming that we would like a rocky surface
, we also assume that the planet s average density should be similar to that of earth .
this means that the acceleration due to gravity , @xmath32 , scales as the cubed root of the planet s mass , @xmath33 .
thus , @xmath32 will be within a factor of 2.5 of @xmath34
(= 9.8 m / s@xmath35 ) if @xmath33 is within a factor of 15 of @xmath36 . increasing or decreasing the value of @xmath32 will lead to different atmospheric contents ; for any given atmospheric temperature , there is a lower limit to @xmath33 ( hence @xmath32 ) , below which an atmosphere will not be retained .
additional ( and more sophisticated ) considerations can influence these limits .
for example , the level of geothermal ( planetary - thermal ) activity may be less for planets of smaller mass , influencing atmospheric chemistry .
in fact , it has been conjectured that part of the difference between earth and mars , which is roughly @xmath37 times less massive than earth , could be related to differences in planetary - thermal activity related to their mass difference .
but it is not clear that this is the crucial difference with regard to liquid water , for example , and other possibilities exist .
( see , e.g. , rampino & caldeira 1994 . ) in this paper , however , i will not highlight such complementary restrictions .
this is because the main relevance of such considerations to microlensing observations is simply that there is likely to be a range of planet masses , possibly spanning two orders of magnitude , consistent with earth - like conditions .
this means that event durations for earth - like planets are likely to also have a range .
the factor of @xmath38 derived above corresponds to a range of event durations from less than an hour to more than a day .
in addition , finite - source - size effects could increase the time duration of events by a factor of a few ( distefano & scalzo 1997 , 1998a , distefano 1998 ) .
if the central star of a ps .
that serves as a lens is fairly luminous , then its light will blend with that from the lensed star .
this blending can have two consequences .
first , if is is measurable and can be quantified , blending makes it possible to learn more about the lens system : the spectral type of the central star can be determined and , in some cases , even the mass of the planetmay be inferred ( distefano 1998 ) . unfortunately , however , the second consequence of blending is that the peak magnification , or in some cases , the time interval during which the event is observable , may be decreased to the point that the planet - lens event is not detected at all
. we must distinguish between the detection of planets in the zone for resonant lensing , and the detection of planets in wider orbits .
the discussion below applies to cases in which the orbital separation is outside the zone for resonant lensing , when the planet acts as a more - or - less independent lens . at the end of 4.1 we return to the case of planets in the zone for resonant lensing .
let @xmath39 represent the fraction of the baseline flux contributed by the lensed star . when the central star of the ps
. is luminous , particularly if its flux comes close to satisfying the criterion studied in the last section ( @xmath40 ) , then @xmath39 can indeed be small enough for the effects of blending to be measurable .
if , for example , the apparent @xmath41 magnitude of the combined light coming along the line of sight from a lensing event is [ @xmath42 , then @xmath43 if the lens is located in the bulge and is a main - sequence star of mass roughly equal to [ @xmath44 @xmath45 . a small value of @xmath39 can allow us to reliably determine the effects of blending and to thereby learn more about the lens .
the question we address below is whether values of @xmath39 small enough to be useful may prevent the event from being detected .
( woniak & paczyski 1997 ) . in the latter case , however , evidence of blending can be obtained through comparisons of spectra taken at peak with spectra taken at baseline ( unless the lens and source are of similar spectral type ) .
astrometry can also be useful ( see , e.g. , goldberg 1988 , goldberg & wozniak 1998 ) . ] when light from the lensed source is blended with light from other sources , the observed magnification , @xmath46 is smaller than the true magnification , @xmath47 @xmath48 thus , in order for a light curve perturbation to be brought above the detection limit , @xmath49 must be larger than it would otherwise have to be , the projected distance between the source and lens must be smaller , and the event will consequently appear to have a shorter . to describe this effect systematically ,
distefano & esin ( 1995 ) introduced the blended einstein radius " , @xmath50 .
let @xmath51 be the minimum peak magnification needed for event detection .
define the of each event to be time during which the magnificationwas greater than @xmath52 the expression for the blended einstein radius is then @xmath53 figure 2 illustrates the influence of blending on event as a function of both @xmath51 and @xmath54 for a given value of @xmath55 the event is longer if @xmath51 is smaller .
thus , increasing the photometric sensitivity to accomodate smaller values of @xmath51 should increase the detection rate . note that if @xmath51 is @xmath56 then even if @xmath57 the event ( which would have been @xmath58 ) , is reduced by only a factor of @xmath59 to @xmath60 thus , while blending does tend to decrease the event , making more frequent monitoring desirable , we can expect to be able to detect a large majority of events by using sensitive photometry .
note , however , that , even with @xmath61 the time of an event with @xmath62 would be decreased by only a factor of @xmath63 . if @xmath51 could be reduced to @xmath64a formidable task for a large - scale monitoring program
then for @xmath65 an event that would have lasted for @xmath66 will have an observed of @xmath60 there has not been a detailed study of the effects of blending on the detection of planets in the zone for resonant lensing .
we can make some general observations , however .
blending does decrease the duration of stellar - lens events . since we expect most such events to last for times on the order of weeks or months
, the arguments above show that it is unlikely that blending will cause us to miss the stellar - lens event altogether . what blending can do , though , is to shorten the event in such a way that we become aware of it only after the planet - lens perturbation has occurred , if the perturbation takes place early in the event .
perturbations can take place early in stellar - lens events ( paczyski 1996 , wambsganss 1997 to @xmath67 . ] ) .
thus , one effect of blending is to decrease the rate of observable resonant planet - lens events .
the effect should not be large , however , because ( 1 ) the events that occur early are not generally the most distinctive and readily observed planet - lens events , and ( 2 ) careful monitoring , once we know the stellar - lens event is underway , can help us to catch even perturbations that take place as the measured flux declines toward baseline , so that we suffer the worst losses only on the upswing of the magnification .
the second effect of blendingis to alter the light curve shape .
when , however , finite - source - size effectsare unimportant , the spike in magnification associated with resonant events is so distinctive that frequent monitoring of the light curve during the event should allow us to detect it .
the remaining challenge is therefore to estimate the combined effects of finite - source size and blending on the overall efficiency of detecting planets in the zone for resonant lensing .
the up - side of blending is that the quantity and color of light from the central star are themselves valuable pieces of information about the ps . .
thus , if the blending is significant enough to make these features measurable , it allows us to learn something interesting about the ps . .
we would like to do the following .
\(1 ) definitively establish that there is blending .
\(2 ) establish that the light not emanating from the lensed source is most likely emitted by the central star of the ps . .
this establishes that , even if the event was an isolated event of short - duration , the lens was a ps . .
\(3 ) determine the spectral type , and possibly the mass of the central star . in cases
in which the mass ratio can be extracted from the light curve , this will establish the mass of the planet lens .
if , in addition to blending , there are measurable finite - source - size effects , then the mass of any planets that served as lenses may be determined directly .
these possibilities , together with the combined effects of blending and finite - source - size effects on event detection , are discussed in more detail elsewhere ( distefano 1998 ) , since they apply to all ps . s discovered via microlensing , not just those most likely to include an earth - like planet . here
i simply note that a combination of observations would generally be needed to accomplish these goals .
these include : light curve studies to begin to assess the role of blending and finite - source - size effects ; possible astrometric studies ( e.g. , goldberg 1988 , goldberg & wozniak 1998 , boden , shao , & van buren 1998 , dominik & sahu 1998 , mao & witt 1998 ) ; spectra taken near peak magnification , compared with spectra taken at baseline , to determine the spectral type and radius of the lensed star ; and high - spatial - resolution follow - up observations to determine whether any light not emanating from the lensed star can be explained by a chance superposition of light from other stars in the field .
microlensing searches for planets complement other types of searches .
they have the advantage of probing vast volumes of space , and of providing information about the existence of planets in diverse stellar environments .
an additional advantage is that microlensingcan discover low - velocity planets .
because they are far away , however , planets discovered via microlensing are less amenable to detailed follow - up observations . in this paper
i have shown that microlensing is a tool that can discover distant planets with earth - like conditions , should such planets exist .
in addition , i find that , among planets discovered by microlensing surveys , those likely to be experiencing earth - like conditions are particularly good targets for some follow - up studies .
this is because the central star of these ps .
s may be luminous enough that the blending of its light with light from the lensed source may be detectable and quantifiable . through a combination of spectra taken during the event , high resolution images and spectra taken after the event , and light curve fitting
, we may be able to determine the spectral classification of the central star , the distance of the ps . from us , and even , in some cases , the mass of the planet lens .
( see also distefano 1998 . )
although there may be a subset of such events that are rendered undetectable by the very blending that helps to make them potentially so interesting , the results presented here show that in most cases , the planet - lens events should be detectable . until now ,
microlensing discussions of earth - like planets have intrinsically focused on earth - mass plants .
in addition , because ( 1 ) low - mass stellar lenses are almost certainly significantly more numerous than higher - mass stars , and ( 2 ) a larger value of the mass ratio between the planet and star is preferred for detection , it has become common for calculations to focus on a low - mass ( @xmath68 ) star orbited by an earth - mass planet .
figure 1 indicates , however , that it is unlikely that a low mass star will harbor a planet with earth - like conditions that can be discovered via microlensing .
furthermore , the considerations in 3.2 indicate that there may be room for more flexibility in the mass of planets with earth - like conditions than has been assumed so far .
simple arguments indicate that the range could extend as much as a factor of ten above , and a factor of 10 below the mass of the earth .
finally , it is important to note that we should consider the possibility that microlensing ( whether by resonant or wide planets ) is most likely to discover the outer planets in a system that may contain closer planets experiencing earth - like conditions . in this case , the microlensing events serve as beacons directing us to the ps . .
in the near - term , such discoveries can contribute to developing the statistics of distant ps .
s , such as the frequency of planets as a function of spectral type and the spatial distribution of planets around stars . in the far ( and so far unforeseeable ) future
, astronomers with instruments capable of measuring the spectra of distant stars and detecting small - amplitude doppler effects can perhaps check these ps .
s for the presence of planets in orbits smaller than those occupied by the planets discovered via microlensing .
it is a pleasure to thank the referee , scott gaudi , for suggested changes to an earlier manuscript ( distefano & scalzo 1998b ) that contained much of this work ; his suggestions have helped to improve the presentation .
i would like to thank arlin crotts , andrew gould , jean kaplan , christopher kochanek , david w. latham , avi loeb , shude mao , robert w. noyes , bodhan paczyski , bill press , penny sackett , kailash sahu , michael m. shara , edwin l. turner , michael s. turner , and the participants in the 1997 aspen workshops , the formation and evolution of planets " and microlensing " for interesting discussions .
it is also a pleasure to thank the aspen center for physics and the institute for theoretical physics at santa barbarba for their hospitality while the first version of this and related papers was being written , and the inter - university center for astronomy and astrophysics in pune , india for its hospitality while the paper was revised .
this work was supported in part by nsf under ger-9450087 and ast-9619516 , as well as by funding from axaf . | are microlensing searches likely to discover planets that harbor life ? given our present state of knowledge , this is a difficult question to answer .
we therefore begin by asking a more narrowly focused question : are conditions on planets discovered via microlensing likely to be similar to those we experience on earth ? in this paper
i link the microlensing observations to the well - known goldilocks problem " ( conditions on the earth - like planets need to be _ just right " _ ) , to find that earth - like planets discovered via microlensing are likely to be orbiting stars more luminous than the sun .
this means that light from the ps .
s central star may contribute a significant fraction of the baseline flux relative to the star that is lensed .
such blending of light from the lens with light from the lensed source can , in principle , limit our ability to detect these events .
this turns out not to be a significant problem , however .
a second consequence of blending is the opportunity to determine the spectral type of the lensed star .
this circumstance , plus the possibility that finite - source - size effects are important , implies that some meaningful follow - up observations are likely to be possible for a subset earth - like planets discovered via microlensing .
in addition , calculations indicate that reasonable requirements on the planet s density and surface gravity imply that the mass of earth - like planets is likely to be within a factor of @xmath0 of an earth mass
. _ earth - mass _ # 1#2#3#4 # 1#2 -.4 true in -.2 true in |
the x - ray emission from galactic x - ray binaries is known to exhibit aperiodic variability on a vast variety of time scales , from months down to milliseconds ( for reviews see , e.g. @xcite ) .
the pca on board the _ rossi x - ray timing explorer _ ( rxte ) , has accumulated a large amount of data on the rapid aperiodic variability ( rav ) and its photon - energy dependence for a large number of x - ray binaries .
this information contains valuable hints about the source of high - energy emission in these objects .
since it is generally believed that the x - ray emission above @xmath0 10 kev is due to comptonization of soft photons in a hot , tenuous coronal gas , it is natural to assume that the rav of this emission component also reveals detailed information about the size scales , dominant physical ( heating and cooling ) mechanisms , and geometry of the comptonizing region .
theoretical calculations of the expected time - dependent signatures of comptonization in a hot , tenuous , static corona had been done for the case of a homogeneous corona ( @xcite ) and for inhomogeneous temperature and density distributions in the corona ( kazanas , hua & titarchuk 1997 , hua , kazanas & titarchuk 1997 ) for the case of central injection of soft photons into a spherical region .
this had been generalized to other geometries , including the case of a cool , outer accretion disc adjacent to an inner , hot , quasi - spherical two - temperature inflow by bttcher & liang ( 1998 ) .
while such models , in particular in the case of central injection , representative of a slab - coronal geometry , could successfully reproduce the dominant spectral and timing features of some objects , such as cyg x-1 , they generally required that the corona on top of the geometrically thin , optically thick accretion disk should extend out to radii of @xmath1 , where @xmath2 is the schwarzschild radius of the accreting compact object , and that out to those radii , the rate of energy dissipation per unit volume would have to decrease only @xmath3 , implying that most of the energy is dissipated at large distance from the central object .
this seems to be very hard to reconcile with realistic calculations of the vertical structure of accretion - disk coronae ( @xcite ) and the radial structure of optically thin , adaf - type flows ( @xcite ) , which indicate that the energy dissipation is concentrated much more strongly towards the disk surface and the central object , respectively .
the required large coronal size scales were implied by the fact that , in these comptonization models , the maximum time lag achievable between hard and soft x - rays corresponds to the difference in photon diffusion time in the process of compton upscattering and is thus of the order of the light travel time through the corona . in order to avoid this size - scale problem , two alternative models for the hard time
lags in galactic x - ray binaries have been developed . in the first one ( @xcite ) , an inner ,
hot coronal flow is assumed to be surrounded by a cool , outer disc . in irregular intervals , blobs of cool material
are breaking lose from the cool disc and spiral inward through the hot coronal flow .
( during the inspiraling process they may most certainly be tidally disrupted into ring - like density perturbations , but for simplicity , we will continue to call these inward - drifting perturbations `` blobs '' .
it has been shown in bttcher & liang [ 1999 ] that they maintain their integrity as isolated , dense regions , and do not evaporate during the inspiraling process ) . in this model ,
hard lags are due to the spectral hardening resulting from the inward - movement of the soft photon source into hotter regions of the coronal flow , combined with the gradual heating of the blob .
thus , the maximum hard time lag achievable corresponds to the drift time scale of the blobs , which may be several orders of magnitude longer than the light travel time through the corona , thus allowing for a correspondingly smaller corona . in the second alternative model ( @xcite ) ,
hard x - ray flares are assumed to be produced by magnetic flares in a hot corona on top of a cool accretion disc .
thus , in this model , the flare is produced by additional heating of the comptonizing material rather than a flare of the soft photon input .
the required spectral hardening occurs naturally during the period of efficient energy release in the corona , and hard lags could be produced in a model consistent with the observed weakness of the compton reflection component if the active regions of the corona are driven away from the disc due to radiation pressure ( @xcite ) . except for the latter model , in which the coronal
heating and cooling are the essential ingredients for modeling the hard x - ray spectral evolution , all the comptonization - based models mentioned above had so far been calculated under the assumption of a static corona with fixed electron temperature , which does not change in response to the increased soft photon input during flares . since it is generally believed that the coronal temperature at least in spectral states other than the very - low or off state of transient sources
is determined by the energy balance between various heating mechanisms and cooling dominated by compton cooling , this may be an unrealistic assumption , as has been pointed out by several authors ( e.g. , @xcite ) . the coronal response to accretion - disc flares in the case of a slab - coronal geometry
has recently been investigated for two special test cases by malzac & jourdain ( 2000 ) .
they found that coronal cooling due to the increased soft photon input leads to rapid cooling of the coronal electron population and a pivoting of the x - ray spectrum around @xmath4 kev for typical parameters , with rather small amplitude variability in the _ rxte _ pca energy range .
they conclude that the resulting variability patterns are inconsistent with the ones observed in most x - ray binaries . in this paper ,
i am presenting a detailed re - evaluation of variability models based on flaring activity in the soft photon input , taking into account self - consistently the coronal heating and cooling in response to the time - varying soft photon input .
the models considered here are the slab - coronal geometry , the hot inner two - temperature flow with outer , cool accretion disc , and the inward - drifting blob model .
the model of dynamical active coronal regions has been investigated in detail in poutanen & fabian ( 1999 ) , and has also been considered in malzac & jourdain ( 2000 ) . in section [ code ] , a brief description of the coupled monte - carlo / fokker - planck code used for this study is given . in sections
[ slab ] , [ adaf ] , and [ blob ] the results concerning the slab - coronal geometry , the hot inner two - temperature flow with outer , cool accretion disc , and the inward - drifting blob case , respectively , are presented .
section [ summary ] contains a brief summary and conclusions .
for our time - dependent simulations of the radiation transfer and electron heating / cooling , the coupled monte - carlo compton scattering and fokker - planck electron dynamics code described in detail by bttcher & liang ( 2000 ) is used .
this code employs a monte - carlo method similar to the large - particle method ( @xcite ) for the photon transport , including all relevant radiation mechanisms , such as compton scattering , cyclotron / synchrotron emission and absorption , bremsstrahlung emission and absorption , and @xmath5 absorption and pair production and annihilation .
the evolution of the electron population is treated simultaneously through an implicit fokker - planck scheme which allows for arbitrary thermal and/or nonthermal electron distributions . at each time step , the radiative energy - loss rates are normalized self - consistently to the energy transfer between electrons and photons evaluated during the photon - transport calculation in the respective time step .
electron acceleration / heating is possible through coulomb interactions with a pool of background , thermal protons or through resonant wave - particle interaction ( 2@xmath6 order fermi acceleration ) with alfvn and whistler wave turbulence , as well as through self - absorption processes .
this means that unlike in most other simulations of related problems ( e.g. , @xcite ; li , kusunose , & liang 1996 ; dove , wilms , & begelman 1996 ; @xcite ) the coronal heating compactness is the self - consistent result of specific acceleration processes , and is not a pre - specified parameter . for the different geometrical and physical scenarios treated below , the corona is generally split up into 10 or 15 vertical or radial zones within which the physical conditions ( electron and proton densities , magnetic fields , etc . ) are assumed to be homogeneous . unless otherwise specified , a globally unordered magnetic field in equipartition with the background thermal proton plasma is assumed .
the properties of all photons escaping the corona toward the observer are written into an event file , from which photon - energy dependent light curves and snapshot photon spectra at arbitrary time intervals can be extracted .
let us first investigate the coronal response to an accretion - disc flare in a slab - coronal geometry ( e.g. , @xcite ) .
such flares could be produced , e.g. , by an unsteady accretion flow through the inner portion of the disk , or by small - scale magnetic flares in the ionized surface layer of the disk .
for this case , the soft photon input from the disc is represented by a thermal blackbody spectrum of @xmath7 kev , typical of the soft excess in the x - ray spectra of black - hole x - ray binaries in the low / hard state .
an accretion disc flare is simulated by increasing the blackbody temperature to 0.5 kev ( i.e. increasing the soft photon compactness by a factor of 39 ) over a limited time interval @xmath8 .
compton reflection of coronal radiation impinging onto the disc is taken into account using the green s functions of white , lightman , & zdziarski ( 1988 ) and lightman & white ( 1988 ) , and at any given time the accretion disc flux is enhanced by the amount of coronal flux absorbed within the disc .
the corona is split up into 10 vertical zones of equal height @xmath9 .
the total height of the corona is fixed to @xmath10 cm , its thomson depth is @xmath11 , and the temperature of the background proton plasma is @xmath12 mev .
coulomb heating of the electrons is assumed to be the dominant electron acceleration mechanism .
this yields an equilibrium electron temperature of @xmath13 kev .
[ adc1 ] shows the energy - dependent light curves , the evolution of the average coronal temperature , and some snapshot energy spectra of the observable x - ray emission for a typical simulation , where @xmath14 .
the accretion - disc flare leads to strong cooling of the coronal electrons .
consequently , in agreement with the results of malzac & jourdain ( 2000 ) , at hard x - ray energies ( @xmath15 kev ) a broad dip rather than a flare results .
the simulated hard x - ray light curves can be reasonably well fitted with a function @xmath16 where @xmath17 + e^{t - t_0 \over \tau_{\rm r } } \ , \theta [ t - t_0 ] \right).\ ] ] here , @xmath18 is the heaviside function . in a series of simulations , different input parameters , such as the coronal thomson depth , proton temperature , etc
. have been varied .
the most critical parameter in these simulations is the duration of the accretion disc flare , @xmath8 , compared to the light crossing time through the corona , @xmath19 .
adc_table ] lists the relevant light curve fitting parameters according to eq .
( [ adc_lightcurves ] ) for some of those cases with the standard coronal parameters listed above .
the fourier transformations of the light curves defined in eq .
( [ adc_lightcurves ] ) are straightforward , but somewhat lengthy . if one neglects terms of order @xmath20 , the complex phase of the fourier transform is given by @xmath21 where @xmath22 the resulting phase and time lags between the 3 10 kev and the 10 50 kev bands from the three representative simulations with @xmath23 s , @xmath24 s , and @xmath25 s , respectively , are shown in fig .
[ adc_phaselags ] .
the figure illustrates that a rather rich variety of phase and time lag phenomena can result from this accretion - disc flare scenario with a slab - coronal geometry .
it appears to be consistent with a fourier - frequency independent time lag at low frequencies , breaking into a power - law with @xmath26 , where generally @xmath27 ( e.g. , @xcite for cyg x-1 ) . in the limit @xmath28 ,
the resulting time lags are dominated by the difference in the fit parameter @xmath29 for the different energy bands , which exhibits a photon energy dependence consistent with the logarithmic time lag dependence measured , e.g. , for cyg x-1 , grs 1915 + 105 , ( e.g. , @xcite ) and xte j1550 - 564 ( wijnands , homan , & van der klis 1999 ; cui , zhang , & chen 2000 ) .
the photon - energy dependence of the parameter @xmath29 is illustrated in fig .
[ t0_energy ] .
note that this logarithmic energy dependence of the maximum time lag is related to the time scales for compton cooling and relaxation back to thermal equilibrium due to coulomb heating rather than due to the difference in photon diffusion time between hard and soft x - rays as in the case of a static corona .
this naturally avoids the problem of the size - scale constraint for static - corona models for the phase and time lags in the rav of x - ray binaries and , instead , allows to place constraints on the proton temperature and density within the corona , which determine the coulomb heating time scale . assuming that both the electron and proton temperatures are non - relativistic , the coulomb heating time scale ( which is expected to be comparable to the maximum time lag between soft and hard x - rays ) can be estimated as @xmath30 ( @xcite ) . here , @xmath31 is the coronal proton density in units of @xmath32 @xmath33 , and @xmath34 is the dimensionless electron and proton temperature , respectively .
( [ tau_coulomb ] ) indicates that maximum time lags of order @xmath35
@xmath36 s can naturally occur in such a scenario .
interestingly , fig .
[ adc_phaselags ] also indicates that over a limited frequency range also negative phase lags ( i.e. soft lags ) may occur as a result of the coronal response to accretion - disc flares .
whether this is a potential explanation for the negative and alternating phase lags observed in grs 1915 + 105 ( @xcite ) and xte j1550 - 564 ( @xcite ) remains to be investigated in more detail in future work .
an important point is that in this scenario , the onset of an accretion - disc flare marks the onset of the episode of enhanced coronal cooling and thus the onset of the decay of the high - energy light curves .
consequently , considering that in a realistic scenario there will be a rapid succession of such flares occurring throughout the disc , the hard x - ray light curves will exhibit maxima around the onsets of the accretion disc flares , with no appreciable offset between the maxima at different photon energies .
this is consistent with the recent result of maccarone et al .
( 2000 ) that the peaks of the cross - correlation function between light curves at different energy bands observed in cyg x-1 are consistent with 0 time lag .
this would be inconsistent with accretion - disc flaring scenarios in slab - coronal geometry if the observed phase lags were due to the time - dependent comptonization response in a static corona ( @xcite ) .
an alternative accretion - flow geometry consists of a cool shakura - sunyaev ( 1973 ) type , thin disc in the outer regions of the flow , which exhibits a transition to an inner , quasi - spherical two - temperature flow , which produces hard x - rays via compton upscattering of self - generated synchrotron and bremsstrahlung photons as well as external soft photons from the surrounding cool accretion disc .
the inner coronal flow could plausibly be a shapiro - lightman - eardley ( 1976 ) type two - temperature flow , an advection - dominated accretion flow ( adaf ; @xcite ) or the recently discovered solution of a convection - dominated flow ( @xcite ) .
the timing properties which are subject of this paper , are rather insensitive to the details of the proton density and temperature profile within the inner coronal region .
thus , for simplicity , the density and temperature profiles of the inner corona are fixed corresponding to an adaf , choosing @xmath37 , and the proton temperature , @xmath38 , equal to the virial temperature .
such a two - phase accretion flow model has been suggested to explain the various luminosity states of x - ray binaries ( narayan , mcclintock , & yi 1996 ; @xcite ) , and has been very successful , in particular , to represent the x - ray spectrum of grs 1915 + 105 over a wide range of source luminosities ( e.g. , muno , morgan , & remillard 1999 ) .
thus , for a parameter study , i start out from parameters in the range typically appropriate for spectral fits to grs 1915 + 105 .
the inner disc temperature found in grs 1915 + 105 is typically 0.75 kev @xmath39 2 kev , and the inner disc radius is 10 km @xmath40 100 km ( @xcite ) . as default settings , for simplicity , the soft radiation input from the cool , outer disc is represented by a thermal blackbody with @xmath41 kev , and the transition radius between cool , outer disc and inner , hot ,
quasi - spherical flow is chosen as @xmath42 km .
the thomson depth of the inner corona is fixed to @xmath43 .
the event horizon is modelled as an absorbing inner boundary of the corona at @xmath44 km .
independent of the value of the transition radius , one finds that a fraction @xmath45 of the accretion disc radiation intercepts the inner corona , if any point on the disc surface is assumed to emit isotropically , the corona is assumed to be spherical , and the disc has a @xmath46 temperature profile .
a short accretion - disc flare is represented by a temporary increase of the disc blackbody temperature to twice its value during non - flaring episodes ( i.e. increasing the soft photon compactness by a factor of 16 )
. figs .
[ adaf1 ] and [ adaf2 ] shows the energy - dependent light curves , evolution of the average coronal temperature , and some snapshot energy spectra from two simulations with standard parameters as quoted above , and @xmath47 s ( @xmath48 ) and @xmath49 s ( @xmath50 ) , respectively .
the figures illustrate that in this case , the coronal response to the accretion - disc flares occurs on time scales at least an order of magnitude shorter than in the case of the slab geometry , due to the substantially higher proton densities in the hottest , central regions of the adaf .
this , combined with the short light crossing and photon diffusion times through the corona , @xmath51 s , makes it very unlikely that substantial phase and time lags ( of order @xmath52 s ) can be produced by the thermal and comptonization response in such a geometry . in particular , comparing the medium and hard x - ray light curves in figs .
[ adaf1 ] and [ adaf2 ] one finds that if ( e.g. in the case of a larger corona ) substantial time lags were produced in this geometry due to radiation transfer effects in the corona , they would be accompanied by substantial peak misalignments of the light curves in different energy bands , which is generally not observed ( e.g. @xcite for grs 1915 + 105 ; @xcite for cyg x-1 ) .
the third flaring scenario investigated here is the model of inward drifting density perturbations ( `` blobs '' ) in a hot , quasi - spherical inner accretion flow , proposed by bttcher & liang ( 1999 ) .
this model is based on the accretion - flow geometry of an outer , cool , optically thick accretion disc and an inner , hot , quasi - spherical two - temperature flow , as in the previous section .
however , in this case , flaring activity is not dominated by flares of the outer accretion disc , but by the inward - drifting of individual blobs ( or ring - like density perturbations ) from the inner boundary of the cool disc through the hot corona , which will eventually disappear within the event horizon of the black hole .
the spectral hardening required to produce hard phase and time lags is a consequence of the combined effects of the soft photon source ( the blob ) moving further inward toward the hotter and denser portions of the corona , and of the heating of the blob as it drifts inward . by default
, the system of the outer cool disc and the central corona will be described by the same set of standard parameters , appropriate to reproduce typical x - ray spectra of grs 1915 + 105 in its high state , as in the previous section : @xmath43 , @xmath37 , virial proton temperature , @xmath53 km , @xmath54 kev .
it is assumed that the blob surface starts out with the same temperature as the disc material , @xmath55 , and heats up as @xmath56 , where i investigate the special cases @xmath57 , @xmath58 , @xmath59 , and @xmath60 . the blob drifts inward with a radial drift velocity @xmath61 . in this general study ,
i choose @xmath62 .
the real value of @xmath63 may in fact be somewhat lower .
however , i choose a rather high radial drift velocity in order to keep the required simulation time at @xmath64 hr ( on a 667 mhz alpha workstation ) , since the code automatically adjusts the intrinsic time step of the simulation to a step size smaller than the time scale of the net physical heating / cooling rate . for lower values of @xmath65 the resulting timing behaviour
is simply stretched over a correspondingly longer time interval , and the respective features in the fourier power and phase lag spectra below @xmath66 khz ( see below ) are shifted to lower frequencies by the same factor , thus increasing the resulting time lags accordingly ( since @xmath67 $ ] ) .
a typical simulation result is illustrated in fig .
[ blob1 ] for a value of @xmath68 ( i. e. @xmath69 ) . except for the highest energy band ( which lies beyond the rxte pca energy range ) ,
the light curves generally show a gradual rise up to a peak , corresponding to the disappearence of the blob , and a steep decline .
since this decline happens on sub - ms time scales , the rapid variability at frequencies @xmath70 khz will be dominated by the gradual - rise phase , which can be parametrized by a functional form @xmath71 the critical fit parameters @xmath72 and @xmath73 for various values of @xmath74 are listed in table [ blob_table ] . generally , the value of @xmath73 increases with increasing photon energy .
this increase becomes more pronounced for decreasing values of @xmath74 .
@xmath72 decreases with increasing photon energy . for the fourier transform of eq .
( [ blob_lc ] ) , i did not find an analytical solution .
[ blob_fourier ] shows the numerically calculated values of the complex fourier phase and the power spectra from light curves parametrized by eq .
( [ blob_lc ] ) , for various values of @xmath73 .
@xmath75 ms has been chosen .
the effect of decreasing @xmath72 by amounts typical of the variations between different energy channels ( see table [ blob_table ] ) is similar to , but much less pronounced than the effect of increasing @xmath73 .
thus , the photon - energy dependent power spectral and phase lag features are dominated by variations in @xmath73 between different energy channels .
the figure demonstrates that the inward - drifting blob model predicts ( a ) a very significant hardening of the power spectra at high frequencies with increasing photon energy and ( b ) a photon - energy and fourier - frequency dependent phase lag .
[ blob_phaselags ] shows the phase and time lags calculated from the analytical parametrizations to the simulated light curves shown in fig .
[ blob1 ] .
one can see that at least in the parameter range investigated here the fourier - frequency dependent phase lag does generally not show a simple broken - power - law or double - broken - power - law behaviour as seen , e.g. , in cyg x-1 ( e.g. , @xcite ) .
rather , this scenario leads to approximately constant time lags at low fourier frequencies and pronounced positive - lag features around the inverse of the blob drift time scale , @xmath76 , beyond which the lags turn negative in a small frequency interval , and begin to oscillate between negative and positive values . beyond @xmath66 khz , the inadequacy of our analytical description of the light curves ( eq . [ blob_lc ] ) on very short time scales ( @xmath77 ms ) might begin to influence the results , so that no predictions can be made at this point .
it might be worth to point out again that all these typical frequencies of the features described above scale @xmath78 and will thus be shifted towards lower frequencies for lower values of the drift velocity .
it is conceivable that the oscillating phase lag features seen in fig .
[ blob_phaselags ] could be related to the peculiar , sometimes alternating phase lags apparently associated with the 0.2 10 hz qpos in grs 1915 + 105 ( @xcite ) and in xte j1550 - 564 ( @xcite ) .
assuming that the size scale of the corona deduced from spectral fitting to grs 1915 + 105 ( @xmath79 @xmath80 km ) is correct , this would require that the inward - drift velocity of density perturbations is @xmath81 .
it will be the subject of future work to determine whether such a choice of @xmath63 is realistic .
a detailed re - analysis of the time - dependent radiation transfer through hot , comptonizing coronae in various models for small - scale high - energy flares in x - ray binaries has been presented .
in contrast to previous work , the effect of coronal cooling due to the increased soft photon input during the flare has been taken into account self - consistently .
in particular , the case of a slab corona with a flare of the underlying , optically thick accretion disc , the case of an inner , hot , quasi - spherical accretion flow with a transition to an outer , cool , optically thick accretion disc , and the case of the inward - drifting blob model have been investigated . in the slab - corona case ,
the resulting light curves look drastically different from the ones calculated under the assumption of a static corona .
this had been pointed out previously by malzac & jourdain ( 2000 ) , who had used a similar model setup as used in the slab - corona - case section of this paper for two specific test cases , but did not investigate the consequences of their results for the phase and time lag spectra . in this paper ,
i have presented a systematic study of the temporal features resulting from such a scenario .
the resulting phase and time lag spectra seem still quite well suited to explain many of the features observed , e.g. , in cyg x-1 .
in particular , the overall shape of the time lag spectra , the photon - energy dependence of the time lags , and the small peak misalignment , consistent with 0 , between the light curves at different photon energy bands , can be reproduced in such a geometry . the model of a flaring disc surrounding a central hot , quasi - spherical two - temperature flow ,
can be ruled out for most objects in which significant phase and time lags have been observed since it would predict very small time lags ( of the order of the light - crossing time through the corona , i.e. @xmath82 ms ) , and light - curve peak misalignments of the order of the maximum time lag , which is generally not observed .
the light curves predicted by the inward - drifting blob model look quite similar to the ones calculated under the assumption of a static corona .
however , there are marked differences with respect to the details of the predicted phase and time lag spectra .
most remarkably , the model ( including coronal cooling ) predicts oscillating phase and time lag features with a typical frequency corresponding to the inverse of the inward - drifting time scale .
if these features are to be associated with the alternating phase lags observed during some observations of grs 1915 + 105 and xte j1550 - 564 and apparently associated with the 0.5 10 hz qpos , this would require an inward drift velocity of the density perturbations of @xmath83 .
the alternative model by poutanen & fabian ( 1999 ) where the spectral variability is produced by a flare of the coronal heat input has been very successful in explaining the time - averaged photon spectrum and power , phase lag and coherence spectra of cyg x-1 , and was in agreement with the peak of the cross - correlation functions between light curves at different energy bands being consistent with 0 . the fundamentally different flaring scenarios variation of the soft photon input vs. variation of the coronal energy input should be distinguishable by virtue of their markedly different spectral evolution and the photon - energy dependence of the rms variability . in the case of a varying soft photon input , there is considerable variability at soft x - ray energies , and the spectral evolution resembles a pivoting of the hard x - ray spectrum around a typical energy of @xmath79 20 kev .
thus , up to an energy of @xmath84 kev , the fractional rms variability should decline as a function of photon energy .
in contrast , if the variability is powered by a varying energy input into the corona , the soft x - ray band is expected to show little variability , and the rms variability should increase with increasing photon energy .
most interestingly , cyg x-1 shows both types of behavior in different observations ( @xcite ) . a consistently declining rms variability over the _ rxte _
pca energy range has been observed in gx 339 - 4 , while 1e 1740.7 - 2942 almost always shows an increasing rms variability with increasing photon energy ( @xcite ) .
this may indicate that fundamentally different variability processes are at work in different objects and even within the same object at different times . in its original version ,
the poutanen & fabian ( 1999 ) model exclusively predicts positive phase lags , i.e. hard lags .
the negative and even alternating phase lags seen in grs 1915 + 105 and xte j1550 - 564 seem to indicate that those lags are not produced by the variability of active coronal regions .
ccccc @xmath25 & 3 10 & 0.56 & 0.38 & 0.50 @xmath25 & 10 50 & 6.00 & 4.99 & 1.28 @xmath25 & 50 500 & 10.6 & 1.37 & 1.30 @xmath24 & 3 10 & 0.95 & 0.38 & 0.59 @xmath24 & 10 50 & 4.80 & 1.31 & 1.12 @xmath24 & 50 500 & 2.89 & 1.16 & 2.18 @xmath85 & 3 10 & 2.05 & 7.48 & 1.38 @xmath85 & 10 50 & 1.77 & 1.04 & 2.04 @xmath85 & 50 500 & 1.25 & 0.87 & 3.07 [ adc_table ] cccc 0 & 1 3 & & 0.00 0 & 3 10 & 37.7 & 0.81 0 & 10 50 & 5.69 & 1.66 0.25 & 1 3 & 9.90 & 0.20 0.25 & 3 10 & 4.33 & 1.28 0.25 & 10 50 & 3.16 & 2.09 0.5 & 3 10 & 4.41 & 1.11 0.5 & 10 50 & 2.62 & 2.05 0.5 & 50 500 & 2.48 & 2.80 0.75 & 1 3 & 3.26 & 1.59 0.75 & 3 10 & 2.10 & 2.02 0.75 & 10 50 & 1.91 & 2.59 [ blob_table ] | the most popular models for the complex phase and time lags in the rapid aperiodic variability of galactic x - ray binaries are based comptonization of soft seed photons in a hot corona , where small - scale flares are induced by flares of the soft seed photon input ( presumably from a cold accretion disc ) .
however , in their original version , these models have neglected the additional cooling of the coronal plasma due to the increased soft seed photon input , and assumed a static coronal temperature structure . in this paper ,
our monte - carlo / fokker - planck code for time - dependent radiation transfer and electron energetics is used to simulate the self - consistent coronal response to the various flaring scenarios that have been suggested to explain phase and time lags observed in some galactic x - ray binaries .
it is found that the predictions of models involving slab - coronal geometries are drastically different from those deduced under the assumption of a static corona . however , with the inclusion of coronal cooling they may even be more successful than in their original version in explaining some of the observed phase and time lag features .
the predictions of the model of inward - drifting density perturbations in an adaf - like , two - temperature flow also differ from the static - corona case previously investigated , but may be consistent with the alternating phase lags seen in grs 1915 + 105 and xte j1550 - 564 .
models based on flares of a cool disc around a hot , inner two - temperature flow may be ruled out for most objects where significant fourier - frequency - dependent phase and time lags have been observed .
accepted for publication in _ the astrophysical journal _ , vol .
553 ( june 1 , 2001 ) |
one of the motivations of this paper is the asymptotic analysis of the generalized laguerre polynomials , denoted by @xmath7 , with complex varying parameters , whose definition and properties can be found for instance in chapter v of szeg s classic memoir @xcite .
they can be given explicitly by @xmath8 or , equivalently , by the well - known rodrigues formula @xmath9\,.\ ] ] expressions and make sense for complex values of the parameter @xmath1 , showing that @xmath7 depend analytically on @xmath1 .
when indeterminacy occurs in evaluating the coefficients in , we understand them in the sense of their analytic continuation with respect to @xmath1 . with this convention
we see that @xmath10 so that @xmath11 for all @xmath12 .
moreover , for any @xmath12 , @xmath7 is the unique ( up to a multiplicative constant ) polynomial solution of the differential equation @xmath13 which shows that every zero of @xmath7 different from @xmath14 must be simple .
in fact , multiple zeros ( at the origin ) can appear if and only if @xmath15 . in this case
the reduction formula @xmath16 shows that @xmath14 is a zero of @xmath17 of multiplicity @xmath18 , see again @xcite for details .
orthogonality conditions satisfied by the laguerre polynomials can be easily derived from iterating integration by parts , see e.g. @xcite ; we reproduce the arguments in section [ sec : orthogonality ] for the sake of completeness .
the weight of orthogonality is the algebraic function @xmath19 and the integration goes along a contour in the complex plane .
the classical situation is @xmath20 , in which case the orthogonality of @xmath21 reduces to @xmath22 as a consequence , for @xmath23 zeros of @xmath21 are positive and simple . in this paper
we study sequences of laguerre polynomials with , in general , complex parameters @xmath1 depending on the degree @xmath2 .
more precisely , we assume that @xmath24 with these hypotheses , but only for _
real _ parameters @xmath25 , sequences @xmath26 were studied , in particular , in @xcite ( weak asymptotics ) and in @xcite ( strong asymptotics ) .
this paper is a natural continuation of this study , although complex values of @xmath6 introduce an asymmetry that makes the problem more difficult .
the main ingredient of the asymptotic analysis is the right choice of the contour of orthogonality , which is related to the trajectories of an associated quadratic differential .
precisely the description of the structure of these trajectories ( section [ sec : trajectories ] ) constitutes the core of our contribution .
it is known that under assumptions we need to perform a linear scaling in the variable in order to fix the geometry of the problem .
thus , we will study the sequence @xmath27 the zeros of @xmath28 cluster along certain curves in the complex plane , corresponding to trajectories ( known also as stokes lines ) of a quadratic differential , depending on a parameter @xmath6 , see figure [ fig : cerostray ] .
this is not surprising : as it follows from the pioneering works of stahl @xcite , and later of gonchar and rakhmanov @xcite , the support of the limiting zero - counting measure of such polynomials is a set of analytic curves exhibiting the so - called @xmath29-property . they can also be characterized as trajectories of a certain quadratic differential on the riemann surface .
the explicit expression of the quadratic differential associated to polynomials @xmath28 can be easily derived from the differential equation , see @xcite for details . for @xmath30 and @xmath31 .
] it turns out that for @xmath32 , the support of the limiting zero - counting measure for @xmath28 s is a simple analytic and real - symmetric arc , which as @xmath6 approaches @xmath33 , closes itself to form for @xmath34 the well - known szeg curve @xcite .
when @xmath35 , the support becomes an interval of the positive semi - axis .
case @xmath36 is special : generically , the support is connected , and consists of a closed loop surrounding the origin together with an interval of the positive semi - axis .
however , when @xmath25 s are exponentially close to integers , the support can split into two disjoint components , the closed contour and the interval , see @xcite for details . in the case @xmath37 the trajectories of these quadratic differentials , and subsequently , the support of the limiting zero - counting measure for @xmath28 s , have not been described ; this is done in section [ sec : trajectories ] , which as we pointed out , is probably one of the central contributions of this paper . being the generalized laguerre polynomials such a classical object , it is not surprising that their asymptotics has been well studied , using different and complementary approaches based on many characterizations of these polynomials .
many of these results correspond to the case @xmath38 , when all their zeros belong to the positive semi - axis , see e.g. @xcite .
others study the asymptotics with a _ fixed _
parameter @xmath1 , such as in @xcite , where additionally many interdisciplinary applications of laguerre asymptotics are explained .
the gonchar - rakhmanov theory was used to find the weak asymptotics ( or asymptotic zero distribution ) of @xmath28 s for general @xmath5 satisfying , see @xcite ; we extend it to complex @xmath25 s in this work , see section [ sec : results ] .
the critical case @xmath39 in this context was analyzed in @xcite , using the extremality of the family of polynomials .
the non - linear steepest descent method of deift zhou introduced in @xcite , and further developed in @xcite and @xcite ( see also @xcite ) , based on the riemann
hilbert characterization of orthogonality by fokas , its , and kitaev @xcite , is an extremely powerful technique , rendering exhaustive answers in cases previously intractable . following this approach , kuijlaars and
mclaughlin @xcite found the strong asymptotics in the whole complex plane for the family @xmath40 and arbitrary values of @xmath41 .
the crucial ingredient of this asymptotic analysis is the choice of an appropriate path of integration on the complex plane , based on the structure of the trajectories of the associated quadratic differential .
taking advantage of the results of section [ sec : trajectories ] we prove the existence and describe such paths of integration , which allows us to carry out the steepest descent analysis in the spirit of kuijlaars and mclaughlin . as a result
, we obtain in section [ sec : results ] the detailed strong asymptotics for the rescaled laguerre polynomials @xmath28 .
this also sheds light on one of the open questions mentioned in @xcite .
the generalized bessel polynomials @xmath42 can be defined as @xmath43 the asymptotic distribution of their zeros and their strong asymptotics is a straightforward consequence of theorems [ teolag1 ] and [ thm : stronga ] below , by replacing @xmath44 and @xmath45 .
let @xmath46 be a complex parameter , for which we define the monic polynomials @xmath47 with @xmath48 since @xmath49 is real - symmetric with respect to the parameter @xmath6 , without loss of generality we can assume in what follows that @xmath50 , and that the square root in stands for its main branch in the closed upper half plane . in what follows , we use the notation @xmath51 while as usual , @xmath52 and @xmath53 stand for the open positive and negative real semi - axes , respectively . on the riemann sphere
@xmath54 we define the quadratic differential @xmath55 written in the natural parametrization of the complex plane
. its _ horizontal trajectories _
( or just trajectories in the future ) are the loci of the equation @xmath56 the _ vertical _ or _ orthogonal _ trajectories are obtained by replacing @xmath57 by @xmath58 in the equation above .
the trajectories and the orthogonal trajectories of @xmath59 produce a transversal foliation of the riemann sphere @xmath54 . in order to study the global structure of these trajectories on the plane
we start by observing that @xmath59 has two zeros , @xmath60 , that are distinct and simple if and only if @xmath61 , and a double pole at the origin if @xmath62 , with @xmath63 another pole of @xmath59 is located at infinity and is of order 4 ; with the parametrization @xmath64 , @xmath65 points from @xmath66 are regular .
the local structure of the trajectories is well known ( see e.g. @xcite , @xcite , @xcite , @xcite ) . at any regular point trajectories
look locally as simple analytic arcs passing through this point , and through every regular point of @xmath59 passes a uniquely determined horizontal and uniquely determined vertical trajectory of @xmath59 , that are locally orthogonal at this point ( * ? ? ?
* theorem 5.5 ) . for @xmath67
, there are @xmath68 trajectories emanating from @xmath60 under equal angles @xmath69 . in the case of the origin
, the trajectories have either the radial , the circular or the log - spiral form , depending on the vanishing of the real or imaginary part of @xmath6 , see figure [ fig : localstructure2 ] . [ cols= " < , < , < " , ] the boundary of @xmath70 consists of critical trajectories only , so that using the definition of @xmath71 it is straightforward to see that @xmath72 is the left half plane .
moreover , by , @xmath73 and the @xmath74 side of the short trajectory @xmath75 is mapped onto the vertical segment @xmath76 $ ] . on the other hand , the `` @xmath77 '' side of @xmath78 and the `` @xmath74 '' side of @xmath79 ( oriented from the origin to @xmath80 ) are mapped onto the imaginary axis .
there are two asymptotic directions to @xmath80 from @xmath81 ; for one of them ( from the `` @xmath74 '' side of @xmath79 ) by definition @xmath82 for the other direction , from the `` @xmath77 '' side of @xmath79 , @xmath83 where the contour of integration encircles the origin in the anti - clockwise direction , and where we have used .
observe that @xmath84 for @xmath85 .
thus , the boundary @xmath86 is mapped onto the vertical line in the right half plane , passing through @xmath87 . by ,
@xmath88 so that @xmath89 corresponds by @xmath90 to the vertical segment @xmath91 $ ] .
this concludes the proof of .
the corresponding results for function @xmath92 are established using the following connection formula : @xmath93 \widetilde{\phi}(z ) + \pi i(1+a ) , & \quad \text{for } z \in \omega_- , \end{array } \right.\ ] ] which is a direct consequence of formulas , and the definition of @xmath92 : @xmath94 observe also that with proposition [ prop : confmapping ] we conclude the proof of theorem [ thm:1 ] .
our main result of the first part of this section is the following [ prop : progressive ] there exist two analytic curves , @xmath95 and @xmath96 , in @xmath97 , emanating from @xmath98 and @xmath80 , respectively , and diverging to infinity in the asymptotic direction @xmath99 , and such that @xmath100 from the considerations at the end of section [ sec : trajectories ] it follows that one instance of @xmath95 could be the orthogonal trajectory of @xmath59 emanating from @xmath98 and diverging in the @xmath99 direction , see figure [ fig : globalstructure ] or figure [ fig : trajectorieswithzeros ] below .
however , this fact is not so relevant ; what really matters for our further analysis is the existence of both curves stated in this corollary . in what follows
we will make use of the contour @xmath101 oriented clockwise , in such a way that @xmath102 is not homotopic to a point in @xmath103 , and the origin remains on the right of the curve .
observe that @xmath102 is not uniquely determined due to the freedom in the choice of @xmath104 , although @xmath75 is .
we turn now to the second goal of this section .
the main tools for the study of the weak asymptotic behavior of polynomials satisfying a non - hermitian orthogonality were developed in the seminal works of stahl @xcite and gonchar and rakhmanov @xcite .
they showed that when the complex analytic weight function depends on the degree of the polynomial , the limit zero distribution is characterized by an equilibrium problem on a compact set in the presence of an external field ; this compact set must satisfy a symmetry property with respect to the external field . for any positive borel measure @xmath105 on @xmath106 , such that @xmath107 given @xmath6 , @xmath108 , consider the harmonic _ external field _
@xmath109 in @xmath103 , where we take the main branches of @xmath110 and @xmath111 .
let also the contour @xmath102 be as defined in .
it is known that the ( unique ) probability _ equilibrium measure _
@xmath105 on @xmath102 in the external field @xmath112 can be characterized by the property @xmath113 where @xmath114 is the equilibrium constant ; for details see e.g. @xcite or @xcite .
both the measure @xmath105 and its support @xmath115 have the _ @xmath29-property _ in the external field @xmath112 if at any interior point @xmath116 of @xmath115 , @xmath117 where @xmath118 are the normals to @xmath115 .
[ prop : eq_measure ] the absolutely continuous measure @xmath119 is the equilibrium measure on @xmath102 in the external field @xmath112 .
function @xmath120 was defined in .
recall that as usual , @xmath75 is oriented from @xmath80 to @xmath98 , and @xmath121 is the boundary value on its left side . by the definition of @xmath75 , the right hand side in
is real - valued and non - vanishing along @xmath75 .
taking into account we conclude that @xmath105 is a probability measure on @xmath75 . furthermore , as in the proof of lemma [ lemma : integrals ] and using , for @xmath122 , @xmath123 thus , @xmath124 thus , for @xmath125 , @xmath126 where @xmath92 was defined in , and @xmath127 is an appropriately chosen real constant .
it remains to use that the last term in the right hand side vanishes on @xmath75 and is strictly positive on @xmath104 ( see corollary [ prop : progressive ] ) to conclude that for @xmath105 in characterization holds .
finally , since @xmath128 , harmonic in @xmath129 , is identically @xmath130 on @xmath131 and is the real part of the analytic function @xmath132 satisfying @xmath133 on @xmath75 , the symmetry property easily follows from the cauchy riemann equations . in the next section
a relevant role will be played by the so - called @xmath134-function , i.e. the complex potential of the equilibrium measure @xmath135 on @xmath75 : @xmath136 where for each @xmath137 we view @xmath138 as an analytic function of the variable @xmath139 , with branch cut emanating from @xmath140 ; the cut is taken along @xmath141 .
now we have all ingredients to formulate and prove the asymptotic results for the rescaled generalized laguerre polynomials @xmath28 , defined in in section [ sec : intro ] , under the assumption .
the analysis is based on the non - hermitian orthogonality conditions satisfied by these polynomials ( section [ sec : orthogonality ] ) . for the weak asymptotics ( or limiting zero distribution ) we can use an analogue of the gonchar
stahl s results , while the strong asymptotics is derived from the corresponding riemann - hilbert characterization of these polynomials ( section [ sec : orthogonality ] ) using the non - linear steepest descent method of deift and zhou . throughout this section
we assume that @xmath142 .
then , the generalized laguerre polynomials @xmath7 satisfy a non - hermitian orthogonality in the complex plane , see @xcite .
namely , [ thort ] let @xmath143 in @xmath144 be an unbounded jordan arc diverging in both directions toward @xmath145 , @xmath146 , and @xmath142 . then @xmath147 in addition , @xmath148 although the selection of the branch of @xmath149 is not relevant , for the sake of definiteness we take here @xmath150 , with @xmath151 . in the proof
we use @xmath152 to denote the @xmath18-th derivative of @xmath153 and @xmath154 . by the rodrigues formula ( [ rodrlag ] ) ,
@xmath155 integrating in the left hand side of @xmath2 times by parts and using ( [ rodrlag2 ] ) , we get @xmath156^{(j ) } w^{(n - j-1)}(z ; n+\alpha ) \bigg|_{\sigma } \\ + \frac{1}{n ! } \ , \int_{\sigma } \big[z^k \big]^{(n ) } w(z ; n+\alpha)\ , dz\ , .
\end{split}\ ] ] since @xmath157 is single - valued on @xmath143 , @xmath158^{(j ) } w^{(n - j-1)}(z ; n+\alpha ) \bigg|_{\sigma}=0\ , , \qquad \mbox{for } 0 \leq j \leq n-1\,.\ ] ] thus , if @xmath159 , all the terms in the right - hand side of ( [ partes ] ) vanish , and ( [ eq21 ] ) follows . furthermore , for @xmath160
, we also get @xmath161 we deform @xmath143 to the positive real axis to obtain @xmath162 where @xmath163 denotes the gamma function . by analytic continuation
the integral in ( [ eq22 ] ) is equal to ( [ eq23 ] ) for every @xmath1 , and ( [ eq22 ] ) follows .
.,width=944 ] consider now the monic rescaled polynomials @xmath164 by theorem [ thort ] , @xmath165 verifies the orthogonality @xmath166 \neq 0 , & \quad \mbox { for } k = n , \end{array } \right.\ ] ] for a contour @xmath143 specified in theorem [ thort ] ( see figure [ fig : curvaortog ] ) . by the classical work of fokas , its , and kitaev @xcite , this yields a characterization of the polynomial @xmath167 in terms of a riemann
hilbert problem : determine a @xmath168 matrix valued function @xmath169 satisfying the following conditions : 1 .
@xmath170 is analytic for @xmath171 , 2 .
@xmath170 possesses continuous boundary values for @xmath172 .
if @xmath143 is oriented clockwise , and @xmath173 , @xmath174 denote the non - tangential boundary values of @xmath170 on the left and right sides of @xmath143 , respectively , then @xmath175 3 .
@xmath170 has the following behavior as @xmath176 : @xmath177 the unique solution of this problem for @xmath178 is given by ( see @xcite ) @xmath179 { d_{n-1 } } p_{n-1}(z ) & \displaystyle{\frac{d_{n-1}}{2\pi i } \int_{\sigma } } \frac{p_{n-1}(\zeta ) \zeta^{\alpha } e^{-n\zeta}}{\zeta -z } \ , d\zeta \end{pmatrix},\ ] ] where @xmath180 is the monic generalized laguerre polynomial ( [ eq24 ] ) and the constant @xmath181 is chosen to guarantee that for the @xmath182 entry of @xmath178 , @xmath183 the orientation of @xmath143 is consistent also with the one used in section [ sec : trajectories ] .
with each @xmath28 in we can associate its normalized zero - counting measure @xmath184 , such that for any compact set @xmath185 in @xmath106 , @xmath186 ( the zeros are considered taking account of their multiplicity ) .
weak asymptotics for @xmath28 s studies convergence of the sequence @xmath187 in the weak-@xmath188 topology , that we denote by @xmath189 .
[ teolag1 ] let the sequence of ( generalized ) laguerre polynomials @xmath28 in ( [ sequencex ] ) satisfy ( [ limits ] ) with @xmath190 . then @xmath191 where @xmath105 is the probability measure given in proposition [ prop : eq_measure ] .
its support is a simple analytic arc , namely the short trajectory @xmath75 described in theorem [ thm:1 ] . regarding the strong asymptotics we have a result very similar to that in @xcite , but referring now to the short trajectory @xmath75 . because of the @xmath2-dependence of @xmath25 , all of the notions and results introduced previously are @xmath2-dependent .
for example , we have that the curves @xmath192 are all varying with @xmath2 , and so we denote them by @xmath193 ; clearly , they tend to the limiting curve @xmath75 .
likewise , we have that the functions @xmath194 , @xmath134 , @xmath92 , and @xmath195 , as well as all matrix - valued functions are @xmath2-dependent , and we also use a subscript @xmath2 to denote their dependence on @xmath2 , using the notation without subscript @xmath2 when referring to the limiting case . [
thm : stronga ] for the rescaled generalized laguerre polynomials @xmath28 as @xmath196 , 1 . *
( asymptotics away from @xmath75 . ) * + uniformly for @xmath139 in compact subsets of @xmath197 , we have as @xmath196 , @xmath198 2 . * ( asymptotics on @xmath74-side of @xmath193 , away from endpoints . ) * + uniformly for @xmath139 on the @xmath74-side of @xmath193 away from @xmath60 , we have as @xmath196 , @xmath199 . \ ] ] 3 . *
( asymptotics on @xmath77-side of @xmath193 , away from endpoints . ) * + uniformly for @xmath139 on the @xmath77-side of @xmath193 away from @xmath60 , we have as @xmath196 , @xmath200 . \ ] ] 4 . *
( asymptotics near @xmath98 . )
* + uniformly for @xmath139 in a ( small ) neighborhood of @xmath98 , we have as @xmath196 , @xmath201,\end{aligned}\ ] ] where @xmath202^{2/3}.\ ] ] recall that @xmath203 was introduced in , while @xmath134 was defined in .
see section [ sec : proofs ] below for a more detailed explanation about the conformal mapping @xmath153 .
we can write a similar asymptotic formula at the other endpoint of @xmath75 ; it will be in terms of the function @xmath71 appropriately redefined in a neighborhood @xmath204 of @xmath80 in such a way that it is analytic in @xmath205 .
namely , by , we should use there @xmath206 where again @xmath79 is oriented from the origin to @xmath80 . as a consequence of theorem [ thm : stronga
] , we can make a stronger statement about the zero asymptotics of @xmath28 ( compare it with theorem [ teolag1 ] ) : [ corollary : zeros ] under assumptions of theorem [ thm : stronga ] , for every neighborhood @xmath204 of @xmath75 , there is @xmath207 such that for every @xmath208 , all zeros of @xmath28 are in @xmath204 , and no zeros at the @xmath209 side of @xmath75 .
figure [ fig : trajectorieswithzeros ] is a good illustration of the statement of this corollary .
zeros_and_trajectories ( 36,37)@xmath210 ( 28,28)@xmath211 ( 43,58)@xmath212 ( 64,71)@xmath213 ( 75,25)@xmath214 ( 24,46)@xmath75 all proofs are based on the complex non - hermitian varying orthogonality satisfied by the generalized laguerre polynomials with complex coefficients ( theorem [ thort ] ) . the cornerstone is the existence of the equilibrium measure @xmath105 having the @xmath29-property , established in proposition [ prop : eq_measure ] .
the connection of the weak asymptotics of non - hermitian orthogonal polynomials with the equilibrium with symmetry was established first by stahl @xcite ( for a fixed weight ) and extended by gonchar and rakhmanov to a varying orthogonality in @xcite .
theorem [ teolag1 ] is a straightforward consequence of the main theorem in @xcite , see also @xcite .
on the other hand , theorem [ thm : stronga ] is established using the non - linear steepest descent analysis of deif and zhou @xcite of the riemann
hilbert ( rh ) problem described in section [ sec : orthogonality ] , where we set @xmath215 , as defined in .
the proof follows the scheme of the work @xcite almost literally .
thus , instead of repeating all the calculations step by step , we describe here the main transformations , referring the interested reader to @xcite for details .
it is convenient to redefine slightly function @xmath92 in , moving the branch cuts from @xmath78 to @xmath96 , and from @xmath79 to @xmath53 as follows .
first , we continue @xmath92 analytically from @xmath70 through @xmath78 to the domain bounded by @xmath78 and @xmath96 .
moreover , let @xmath216 be the arc of @xmath79 from @xmath80 to its first intersection with @xmath53 , see figure [ fig : critical_gr_mod ] .
then @xmath217 consists of two connected components : `` upper '' ( containing the origin ) and `` lower '' domains .
hence , if we define @xmath218 it will be holomorphic in @xmath219 , see our analysis in section [ sec : auxiliary ] . as usual , the first transformation of the riemann - hilbert problem for @xmath178 is the regularization at infinity by means of the function @xmath134 introduced in . by , there is a constant @xmath127 such that @xmath220 where @xmath221 is defined with a branch cut along @xmath53 .
we define for @xmath222 , @xmath223 here , and in what follows , @xmath224 denotes the pauli matrix @xmath225 , so that for example @xmath226 . from the riemann
hilbert problem for @xmath178 it follows by a straightforward calculation that @xmath227 is the unique solution of the following rh problem : determine @xmath228 such that 1 .
@xmath229 is analytic for @xmath222 , 2 .
@xmath229 possesses continuous boundary values for @xmath230 , denoted by @xmath231 and @xmath232 , and @xmath233 for @xmath230 , 3 .
@xmath229 behaves like the identity at infinity : @xmath234 the jump relation for @xmath227 has a different form on the three parts @xmath102 . using it
is easy to obtain the following jump relations for @xmath134 across the contour @xmath102 ( see @xcite ) : @xmath235 @xmath236 and @xmath237 the second transformation of the rh problem deals with the oscillatory behavior in the jump matrix for @xmath227 on @xmath75 , see , and is based on its standard factorization : @xmath238 as part of the steepest descent method we introduce the oriented contour @xmath239 , which consists of @xmath102 plus two simple curves @xmath240 from @xmath80 to @xmath98 , contained in @xmath70 and @xmath241 , respectively , as shown in figure [ fig : lens1 ] .
we choose @xmath240 such that @xmath242 on @xmath240 , which , as it follows from , is always possible . then @xmath243 has four connected components , denoted by @xmath244 , @xmath245 , @xmath246 , and @xmath247 as indicated in figure [ fig : lens1 ] .
lens1 ( 20,19)@xmath210 ( 10,32)@xmath248 ( -4,35)@xmath249 ( 20.5,35)@xmath250 ( 60,55)@xmath251 ( 37,52)@xmath212 ( 8,12)@xmath211 ( 51,7)@xmath252 ( 90,55)@xmath253 ( 9,42)@xmath254 ( 19,42)@xmath255 ( 50,40)@xmath256 consequently , we define @xmath257 by @xmath258 @xmath259 @xmath260 then from the rh problem for @xmath227 and the factorization ( [ eq47 ] ) we obtain that @xmath261 is the unique solution of the following riemann
hilbert problem : determine a @xmath262 matrix valued function @xmath263 such that the following hold : 1 .
@xmath264 is analytic for @xmath265 , 2 .
@xmath264 possesses continuous boundary values for @xmath266 , denoted by @xmath267 and @xmath268 , and @xmath269 @xmath270 @xmath271 and @xmath272 3 .
@xmath264 behaves like the identity at infinity : @xmath273 the global ( or outer ) parametrix corresponding to this problem is a matrix @xmath274 given by ( see e.g. @xcite , ( * ? ? ?
* section 7.3 ) , or ( * ? ? ? * section 5.1 ) ) @xmath275 where we take the main branches of the square roots .
it is easy to verify that 1 .
@xmath276 is analytic for @xmath277 , 2 .
@xmath276 possesses continuous boundary values for @xmath278 , denoted by @xmath279 and @xmath280 , and @xmath281 3 .
@xmath282 for @xmath176 .
4 . near the endpoints
@xmath60 it satisfies @xmath283 since the behavior of @xmath284 does not match the desired behavior at the endpoints @xmath60 of @xmath75 , we need one more construction around these points , namely the so - called local parametrix , well described for instance in @xcite . from its definition in ( [ def : phi ] )
it is easy to see that the @xmath92-function has a convergent expansion @xmath285 in a neighborhood of @xmath98 .
the factor @xmath286 is defined with a cut along @xmath141 .
then @xmath153 , defined by @xmath287^{2/3},\ ] ] is analytic in a neighborhood of @xmath98 .
we choose the @xmath288-root with a cut along @xmath75 and such that @xmath289 for @xmath290 .
recall that @xmath291 on @xmath95 . then @xmath292 and @xmath293 .
therefore we can deform @xmath95 locally and choose @xmath294 so small that @xmath295 is a one - to - one mapping from @xmath296 onto a convex neighborhood @xmath297 of @xmath298 . under the mapping @xmath295
, we then have that @xmath299 corresponds to @xmath300 and that @xmath301 corresponds to @xmath302 \cap f(u_{\delta})$ ] .
we can also deform @xmath240 in such a way that for an arbitrary , but fixed @xmath303 , @xmath153 maps the portion of @xmath304 and @xmath305 in @xmath306 to the rays @xmath307 and @xmath308 , respectively . with this mapping
we take @xmath309 , analytic for @xmath310 , and continuous on @xmath311 , given by @xmath312 where @xmath313 and @xmath314 is an explicit matrix valued function built out of the airy function @xmath315 and its derivative @xmath316 as follows @xmath317 \left(\begin{array}{cc } \operatorname{ai}(t ) & \operatorname{ai}(\omega^2 t ) \\
\operatorname{ai}'(t ) & \omega^2 \operatorname{ai}'(\omega^2 t ) \end{array } \right ) e^{- \frac{\pi i}{6 } \sigma_3 } \left(\begin{array}{cc } 1 & 0 \\ -1 & 1 \end{array } \right ) & \mbox{for } \theta < \arg t < \pi , \\[10pt ] \left(\begin{array}{cc } \operatorname{ai}(t ) & -\omega^2 \operatorname{ai}(\omega t ) \\
\operatorname{ai}'(t ) & -\operatorname{ai}'(\omega t ) \end{array } \right ) e^{-\frac{\pi i}{6 } \sigma_3 } \left(\begin{array}{cc } 1 & 0 \\ 1 & 1 \end{array } \right ) & \mbox{for } -\pi < \arg t < -\theta , \\[10pt ] \left(\begin{array}{cc } \operatorname{ai}(t ) & -\omega^2 \operatorname{ai}(\omega t ) \\
\operatorname{ai}'(t ) & -\operatorname{ai}'(\omega t ) \end{array } \right ) e^{- \frac{\pi i}{6 } \sigma_3 } & \mbox{for } -\theta < \arg t < 0 , \end{array } \right.\ ] ] with @xmath318 . a similar construction yields a parametrix @xmath319 in a neighborhood @xmath320 , see @xcite for details . finally ,
using @xmath284 , @xmath309 , and @xmath319 , we define for every @xmath321 , @xmath322 then @xmath29 is defined and analytic on @xmath323 .
however it follows from the construction that @xmath29 has no jumps on @xmath75 and on @xmath324 .
therefore @xmath29 has an analytic continuation to @xmath325 , where @xmath326 is the contour indicated in figure [ fig : lens2 ] .
contour @xmath326 splits the complex plane in the subdomains also indicated in figure [ fig : lens2 ] .
lens2 ( 20,19)@xmath210 ( -4,35)@xmath249 ( 20.5,35)@xmath250 ( 60,55)@xmath251 ( 35,54)@xmath296 ( 8,14)@xmath327 ( 51,7)@xmath252 ( 90,55)@xmath328 ( 15,42)@xmath329 ( 50,40)@xmath330 we also have that @xmath331 with a constant @xmath216 that is independent of @xmath139 ( it can also be chosen independently of the value of @xmath6 for @xmath6 in a compact subset of @xmath106 , see @xcite ) , and standard arguments show that jump matrices for @xmath29 are close to the identity matrix if @xmath2 is large .
although in the previous analysis we have used the value @xmath6 ( assumed fixed ) , as it was explained in the introduction to theorem [ thm : stronga ] , all of the notions and results introduced before are @xmath2-dependent . however , from the asymptotic assumption we have that the curves @xmath193 tend to the limiting curve @xmath75 , etc .
we observed already that the jump matrix for @xmath332 is @xmath333 uniformly on @xmath334 as @xmath196 .
in addition , the jump matrix converges to the identity matrix as @xmath176 along the unbounded components of @xmath334 sufficiently fast , so that the jump matrix is also close to @xmath335 in the @xmath336-sense . since the contours @xmath334 are only slightly varying with @xmath2
, we may follow standard arguments to conclude that @xmath337 uniformly for @xmath338 . finally ,
unraveling the steps @xmath339 and using ( [ eq72 ] ) , we obtain strong asymptotics for @xmath340 in all regions of the complex plane .
in particular we are interested in the ( 1,1 ) entry of @xmath340 , since this is the monic generalized laguerre polynomial .
we are not describing the details of this straightforward calculation here , and refer again the interested reader to @xcite .
this completes the proof of theorem [ thm : stronga ] .
we finish by proving corollary [ corollary : zeros ] .
the assertion that all zeros of @xmath28 are in @xmath204 ( accumulate at @xmath75 ) is a direct consequence of ( a ) of theorem [ thm : stronga ] and the fact that @xmath341 in @xmath342 .
observe that @xmath343 if and only if @xmath344 $ ] , where as in lemma [ lemma : convexity ] , we denote by @xmath345 $ ] the straight segment joining @xmath80 and @xmath98 . by lemma [ lemma :
convexity ] , for @xmath346 , @xmath347 $ ] is a boundary of a simply connected domain ; let us denote it by @xmath348 .
thus , @xmath349 preserves sign both in @xmath348 and in @xmath350 , and these signs are opposite . since @xmath351
we conclude that @xmath352 for @xmath353 , and @xmath354 for @xmath355 .
in particular , @xmath356 , the inner boundary of @xmath348 corresponds to the `` @xmath77 '' side of @xmath75 . since for @xmath346 , @xmath357 $ ]
, we conclude that this property holds for every value of @xmath346 .
in particular , we have proved that holds only on the `` @xmath77 '' side of @xmath75 . by assertions ( b ) and ( c ) of theorem [ thm : stronga ] , zeros of @xmath28 in a neighborhood of @xmath193
must satisfy @xmath358since by , @xmath359 on both sides of @xmath131 , it remains to use to conclude the proof .
the second and the third authors ( amf and pmg ) have been supported in part by the research project mtm2011 - 28952-c02 - 01 from the ministry of science and innovation of spain and the european regional development fund ( erdf ) , by junta de andaluca , research group fqm-229 , and by campus de excelencia internacional del mar ( ceimar ) of the university of almera .
additionally , amf was supported by the excellence grant p09-fqm-4643 from junta de andaluca .
mja and ft have been partially supported by the research unit ur11es87 from the university of gabs and the ministry of higher education and scientific research in tunisia .
the authors acknowledge the contribution of the anonymous referee whose careful reading of the manuscript helped to improve the presentation .
10 url # 1`#1`urlprefixhref # 1#2#2 # 1#1 a. martnez - finkelshtein , p. martnez - gonzlez , r. orive , http://dx.doi.org/10.1016/s0377-0427(00)00654-3[on asymptotic zero distribution of laguerre and generalized bessel polynomials with varying parameters ] , j. comp .
( 2001 ) , 477487 .
a. a. gonchar , e. a. rakhmanov , equilibrium measure and the distribution of zeros of extremal polynomials , mat .
sbornik 125 ( 2 ) ( 1984 ) 117127 , translation from mat .
134(176 ) , no.3(11 ) , 306 - 352 ( 1987 ) . a. a. gonchar , e. a. rakhmanov , equilibrium distributions and degree of rational approximation of analytic functions , math .
ussr sbornik 62 ( 2 ) ( 1987 ) 305348 , translation from mat .
134(176 ) , no.3(11 ) , 306 - 352 ( 1987 ) .
i. e. pritsker , r. s. varga , http://dx.doi.org/10.1090/s0002-9947-97-01889-8[the szeg curve , zero distribution and weighted approximation ] , trans .
349 ( 10 ) ( 1997 ) 40854105 . http://dx.doi.org/10.1090/s0002-9947-97-01889-8 [ ] . http://dx.doi.org/10.1090/s0002-9947-97-01889-8 d. borwein , j. m. borwein , r. e. crandall , http://dx.doi.org/10.1137/07068031x[effective laguerre asymptotics ] , siam j. numer .
46 ( 6 ) ( 2008 ) 32853312 .
http://dx.doi.org/10.1137/07068031x [ ] .
http://dx.doi.org/10.1137/07068031x p. deift , t. kriecherbauer , k. t .-
mclaughlin , s. venakides , x. zhou , asymptotics for polynomials orthogonal with respect to varying exponential weights , internat .
notices ( 16 ) ( 1997 ) 759782 .
p. deift , t. kriecherbauer , k. t .-
mclaughlin , s. venakides , x. zhou , uniform asymptotics for orthogonal polynomials , in : proceedings of the international congress of mathematicians , vol .
iii ( berlin , 1998 ) , no .
extra vol .
iii , 1998 , pp .
491501 ( electronic ) . | in this paper we study the asymptotics ( as @xmath0 ) of the sequences of laguerre polynomials with varying complex parameters @xmath1 depending on the degree @xmath2 .
more precisely , we assume that @xmath3 and @xmath4 .
this study has been carried out previously only for @xmath5 , but complex values of @xmath6 introduce an asymmetry that makes the problem more difficult .
the main ingredient of the asymptotic analysis is the right choice of the contour of orthogonality , which requires the analysis of the global structure of trajectories of an associated quadratic differential on the complex plane , which may have an independent interest .
while the weak asymptotics is obtained by reduction to the theorem of gonchar rakhmanov stahl , the strong asymptotic results are derived via the non - commutative steepest descent analysis based on the riemann - hilbert characterization of the laguerre polynomials .
trajectories and orthogonal trajectories of a quadratic differential , riemann - hilbert problems , generalized laguerre polynomials , strong and weak asymptotics , logarithmic potential , equilibrium . |
the occurrence of hot jupiters " ( hjs ) gas giants with periods @xmath3 days is estimated to be around @xmath4 for fgk stars @xcite .
it is commonly believed that these planets formed at larger separations from their host stars ( presumably beyond the ice line ) and subsequently migrated " to semi - major axes of less than @xmath5 au .
the known migration mechanisms can be ( i ) disc mediated " due to planet - disc interaction ( e.g. , * ? ? ?
* ; * ? ? ?
* ) , ( ii ) planet mediated " , including strong planet - planet scatterings and various forms of secular interactions among multiple planets ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , or ( iii ) binary mediated " , i.e. , secular interactions with a distant stellar companion @xcite . excluding disc - driven migration ,
these mechanisms require the migrating planet to reach high eccentricities in order for tidal dissipation at pericenter to shrink the planet semi - major axis down to @xmath6 au . thus
, such mechanisms are sometimes referred to collectively as high-@xmath7 migration " .
planet migration following eccentricity excitation through lidov - kozai oscillations @xcite induced by a stellar companion migration scenarios . in this paper
, we focus on lk oscillations induced by a stellar - mass companion . ]
often called lidov - kozai ( lk ) migration has been investigated by different researchers via monte carlo experiments @xcite .
the latest studies have concluded that at most @xmath8 of observed hjs can be explained by lk migration in stellar binaries .
similarly , observational evidence @xcite also indicates that the population of hjs is unlikely to be explained by a single migration mechanism .
the occurrence rate of hjs produced by lk migration in stellar binaries can be computed via @xmath9 where @xmath10 is fraction of stars having a wide binary companion , @xmath11 is the fraction of solar - type stars hosting a `` regular '' giant planet at a distance of a few au , and @xmath12 is the formation fraction / efficiency of hjs ( i.e. , the fraction of systems that result in hjs for a given distributions of binaries and `` regular '' planets ) .
optimistic values of @xmath13 ( for binary separations between 100 au and 1000 au ; see @xcite ) and @xmath14 @xcite require @xmath15@xmath16 to explain the observed hj occurrence rate ( @xmath17 ) using lk migration alone . in a recent work ,
@xcite ( hereafter asl ) conducted a comprehensive population synthesis study of hj formation from lk oscillations in stellar binaries , including all relevant physical effects ( the octupole potential from the binary companion , mutual precession of the host stellar spin axis and planet orbital axis , tidal dissipation , and stellar spin - down due to magnetic braking ) . unlike previous works , asl considered a range of planet masses ( @xmath18 ) and initial semi - major axes ( @xmath19 au ) , different properties for the host star , and varying tidal dissipation strengths .
they found that the hj formation efficiency depends strongly on planet mass and stellar type , with @xmath20 ( depending on tidal dissipation strength ) for @xmath21 , and larger ( up to @xmath22 ) for more massive planets .
the production efficiency of `` hot saturns '' ( @xmath23 ) is much lower , because most migrating planets are tidally disrupted .
remarkably , asl found that the fraction of migrating systems , i.e. those that result in either hj formation or tidal disruption , @xmath24 is roughly constant , having little variation with planet mass , stellar type and tidal dissipation strength .
asl provided a qualitative explanation of how such a robust @xmath25 may come about ( see sections 3.4 and 5.4.1 of that work ) .
note that @xcite considered @xmath26 around solar - type stars and found a similar @xmath27 as asl , but a much larger disruption fraction ( @xmath28 ) , and thus obtained @xmath29 .
the difference arises from the fact that @xcite placed all planets initially at @xmath30 au and assumed that the binary eccentricity distribution extends up to 0.95 , whereas asl considered @xmath31 au and a maximum binary eccentricity of 0.8 ( see section 6.2 of asl for further discussion ) . in this paper ,
following on the work of asl , we present an analytic study that explains ( i ) how the lk migration fraction of close - in planets are largely determined by geometrical considerations , and ( ii ) how the planet survival and disruption fractions depend on the properties of the planets and host stars .
we outline the necessary steps that allow for simple calculations of @xmath32 and @xmath33 under various conditions and assumptions ( e.g. , the distributions of initial planetary semi - major axes , binary separations and eccentricities , the masses of planet and host star ) .
our analytic calculations agree with the results of population synthesis studies .
in addition , we extend our calculations to planets in the super - earth mass range .
the success of lk migration depends on the maximum eccentricity @xmath34 that can be achieved by the planet in the lk cycles .
this eccentricity needs to be high enough such that the pericenter separation between the planet and its host star is sufficiently small for tidal dissipation in the planet to be efficient ( see section [ sec : required_distance ] ) .
the maximum eccentricity depends on the initial inclination @xmath35 between the planetary and binary orbital axes , and on various system parameters : @xmath36 ( initial planetary semi - major axis ) , @xmath37 and @xmath38 ( binary semi - major axis and eccentricity ) , @xmath39 , @xmath40 , @xmath41 ( masses of the planet , host star and outer binary companion ) , and the planet s radius @xmath42 and its internal structure ( love number ) . in @xcite ( hereafter lml ) , we have studied the maximum eccentricity in lk oscillations including the octupole potential of the binary companion and the effect of short - range forces ( associated with general relativity , the tidal bulge and rotational bulge of the planet ) . here
we summarize the key results needed for this paper .
the strength of the octupole potential from the binary companion relative to the quadrupole potential is characterized by the dimensionless ratio @xmath43 when @xmath44 , the octupole effect can be neglected . in this limit ,
the planet s dimensionless angular momentum ( projected along the binary axis ) @xmath45 is conserved , resulting in regular lk oscillations in eccentricity and inclination .
together with the conservation of energy , the maximum eccentricity @xmath34 can be calculated analytically as a function of @xmath35 ( and other parameters ) @xcite .
assuming that the planet has zero initial eccentricity , @xmath34 is determined by ( lml ) @xmath46 where @xmath47 , and the dimensionless quantities [ eq : srf ] @xmath48 represent the relative strengths of apsidal precessions due to general relativity ( gr ) , tidal bulge , and rotational distortion of the planet . here
@xmath49 is the love number , @xmath50 the apsidal motion constant and @xmath51 the rotation rate of the planet . in the pure " lk limit ( @xmath52 ) , equation ( [ eq : max_ecc ] )
reduces to the standard expression @xmath53 . for typical planetary rotation rates , the contribution from the rotational bulge ( @xmath54 ) can be neglected compared to the tidal term ( @xmath55 ) ( see lml ) the maximum value of @xmath34 is achieved when @xmath56 in equation ( [ eq : max_ecc ] ) , and is referred to as the limiting eccentricity " @xmath57 . when @xmath58 , the limiting eccentricity satisfies ( eq . 53 of lml ) @xmath59 ( the dimensionless octupole strength ) and @xmath35 ( the initial inclination angle )
. in the @xmath60-@xmath35 plane , this boundary is represented by the critical initial angle @xmath61 above which planets can reach the limiting eccentricity @xmath57 ( equation [ eq : lim_ecc ] ) .
the data points are taken from the numerical integrations of @xcite and the red curve depicts the fitting function ( equation [ eq : octupole_window ] ) . for @xmath62
, @xmath61 flattens out to a constant value @xmath63 . for comparison ,
the blue curve shows the analytic result for the critical @xmath35 for orbital flip without srf , as obtained by @xcite .
[ fig : critical_angle],scaledwidth=49.0% ] in the presence of finite octupole potential ( @xmath64 ) , @xmath65 is no longer a constant , but a slowly varying quantity @xcite , with @xmath66 . for sufficiently large initial inclinations ( @xmath67 ) , the slow evolution of @xmath65 can take it to extremely small values and even make it cross zero ( i.e. @xmath68 , see e.g. , * ? ? ?
as @xmath65 becomes small , @xmath7 can reach very high values . as described by lml ,
the effect of the octupole is to increase the range of possible initial angles @xmath35 for which @xmath57 can be reached . to capture the effect described by lml , we can approximate the maximum eccentricity with ) is restricted to @xmath69
. however , note that the system is symmetric around @xmath70 , thus the limiting eccentricity is also reached for @xmath71 . ]
@xmath72 where @xmath73 is the solution to equation ( [ eq : max_ecc ] ) and @xmath57 is the solution to equation ( [ eq : lim_ecc ] ) , and @xmath74 is the critical angle above which @xmath57 is reached ( for @xmath75 , it is required that @xmath76 , such that the quadrupole approximation is recovered ) .
note that equation ( [ eq : max_ecc_oct ] ) neglects the `` transition region '' for @xmath35 just below @xmath74 in which the octupole potential makes @xmath77 larger than @xmath78 but not as large as @xmath57 .
the numerical calculations of lml showed that when tidal distortion ( @xmath79 ) dominates the srfs , the transition region is narrow .
the critical inclination angle @xmath74 , which defines the octupole - active " window , depends on @xmath60 . figure [ fig : critical_angle ] shows the numerically computed values of of @xmath74 ( lml ) .
these can be well approximated by the fitting formula @xmath80 for @xmath81 with @xmath82 , @xmath83 , @xmath84 and @xmath85 . for @xmath86 ,
the dependence on @xmath87 flattens off at @xmath88 or @xmath89 ( fig .
[ fig : critical_angle ] ) .
note that , as shown by lml , the octupole - active window is equivalent to the range of inclinations susceptible to orbital flips " in lk cycles _ without _ srfs .
@xcite has obtained an analytic expression of the critical inclination angle for orbital flip this analytic result is valid only for @xmath90 ( see fig .
[ fig : critical_angle ] ) .
we note that , in general , the octupole - active window of extreme eccentricity is reached only when @xmath91 ( the argument of pericenter ) of the planet is circulating ( as discussed by * ? ? ?
. however , since planets commonly start with @xmath92 , this distinction is irrelevant , as planets at @xmath92 lie very close to the separatrix between the librating and circulating regions of phase space .
in order to migrate , a distant planet must reach small pericenter distance , @xmath93 , such that tidal dissipation can reduce the planet s semi - major axis within a few gyr .
the effective orbital decay rate during lk oscillations can be estimated from the orbital decay rate due to tidal dissipation at @xmath94 multiplied by the fraction of the time the planet spends in the high-@xmath7 phase ( @xmath95 ) , giving ( see section 3.2 and eq .
32 of asl ) @xmath96 where @xmath97 is the lag time , the pericenter distance at maximum eccentricity is defined as @xmath98 and the expression @xmath99 is replaced by its approximate value @xmath100 . in order to have migration within @xmath101 years ( i.e. , @xmath102 years ) ,
we require @xmath103 , with @xmath104 where @xmath105 is a dimensionless tidal enhancement factor .
we see that , except for planet radius , the dependence of the critical pericenter distance @xmath106 on the physical parameters of the system is weak , and thus it represents a very robust ( albeit not an exact ) estimate of how close planets must get to their host stars in order to migrate effectively ( see asl ) . for this pericenter to be reached
, a planet initially located at @xmath36 needs to achieve a maximum eccentricity of @xmath107 via lk oscillations .
note that , since the final semi - major axis @xmath108 of the planet after tidal decay is approximately @xmath109 , equation ( [ eq : mig_distance ] ) indicates that it is difficult to form hjs with @xmath110 au via lk migration ( @xcite ; asl ) , unless the tidal dissipation parameter @xmath111 is much larger than the canonical value of 10 .
equation ( [ eq : mig_distance ] ) sets the maximum pericenter separation that allows for lk migration .
there is also a _ minimum _
pericenter separation that permits `` successfully migrated '' planets . for pericenter separations smaller than the tidal radius @xmath112 ( where @xmath113 ; * ? ? ?
* ) , planets will be disrupted and therefore unable to turn into hjs .
as before , for this specific pericenter to be reached , a planet initially located at @xmath36 needs to achieve lk maximum eccentricity of @xmath114 note that , if @xmath115 , all migration - inducing lk oscillations will result in tidal disruptions .
this condition imposes a lower limit in the planet mass that permits the formation of close - in planets : @xmath116 this implies that saturn - mass planets can rarely migrate successfully via lk oscillations without first being disrupted ( see asl ) .
having defined the eccentricities required for migration and tidal disruption , we can now compute the fractions of migration and tidal disruption for given @xmath36 , @xmath37 and @xmath38 ( plus the other fixed parameters of the systems such as @xmath39 , @xmath40 , @xmath41 , @xmath42 and @xmath111 ) . to achieve @xmath117 , the initial planetary inclination angle ( relative to the binary )
must be sufficiently large , i.e. , @xmath118 .
for an isotropic distribution of the initial planetary angular momentum vectors , those that can lead to @xmath117 or larger occupy a solid angle @xmath119 .
thus the migration fraction is @xmath120 similarly , the disruption fraction is given by @xmath121 where @xmath122 is the critical inclination that leads to @xmath123 .
the fraction of hj formation ( migration without disruption ) is simply @xmath124 for @xmath125 , no migration is possible and we have @xmath126 . for @xmath127 , to quadrupole - order , the critical angle for migration @xmath128 can be computed analytically by directly solving for @xmath129 in equation ( [ eq : max_ecc ] ) with @xmath130 .
thus @xmath131 where @xmath132 with @xmath133 and in which the rotational distortion effect in equation ( [ eq : max_ecc ] ) has been ignored for consistency .
note that the condition @xmath127 is equivalent to @xmath134 .
similarly , the critical disruption angle @xmath135 is given by equation ( [ eq : solid_angle ] ) with @xmath136 replacing @xmath137 .
in the first panel of fig . [
fig : example_fractions ] , we show @xmath138 , @xmath139 and @xmath140 for a @xmath141 planet started at different initial semi - major axis @xmath36 and a binary companion with @xmath142 au and @xmath143 . we see that , for large @xmath36 , srfs are unimportant for determining @xmath34 around @xmath137 , thus the shape of the curve is simply @xmath144 . for small @xmath36 , @xmath57 is smaller than the required @xmath137 , making the migration fraction rapidly go to zero .
we illustrate the effect of the octupole term on @xmath138 and @xmath139 in the three last panels of fig .
[ fig : example_fractions ] . for all panels , @xmath152 au , while the eccentricity is varied from 0 to 0.8 .
as @xmath38 is increased , the value of @xmath60 for a given planet semi - major axis @xmath36 also grows .
going from @xmath143 to @xmath153 illustrates the gradual transition from @xmath154 for all @xmath36 ( pure quadrupole ) to @xmath155 for all @xmath36 ( entirely dominated by the octupole effect ) .
note that the hj fraction @xmath156 is always a narrow distribution bounded by at the no - migration " limit ( @xmath157 ) at small @xmath36 and by the tidal disruption limit ( @xmath158 ) at larger @xmath36 .
the width and height of this peak is only mildly affected by the octupole contribution . on the other hand ,
the fraction of tidally disrupted planets @xmath139 is increased significantly for larger @xmath38 .
thus , the main effect the octupole is to increase the number of disrupted planets , rather than the number of successfully migrated ones ( @xcite ; asl ) .
[ fig : example_fractions2 ] shows the same examples as fig .
[ fig : example_fractions ] , except for a planet mass of @xmath159 .
although the lower limit for non - zero migration is almost unchanged ( see eq .
47 of asl ) , the boundary between successful migration and tidal disruption occurs at a larger @xmath36 ( @xmath136 is larger for more massive planets , making disruption more difficult ; see equation ( [ eq : dis_eccentricity ] ) .
thus , a larger fraction of these massive planets can survive lk oscillations and become hjs .
the mass - dependent boundary between disruption and successful migration explains the lack of hot saturns " in observations ( asl ) .
note that , as in the previous example , the most noticeable effect of the octupole terms which differentiate the last three panels from the first is to increase the rate of tidal disruptions , not to significantly increase the number of hjs .
in the previous section , we described how to calculate the migration fraction @xmath160 and disruption fraction @xmath161 for a given set of orbital parameters @xmath36 , @xmath37 and @xmath38 ( eqs .
[ eq : mig_fraction]-[eq : dis_fraction ] ) .
we now consider the migration and disruption fractions integrating over a range of values in @xmath37 and @xmath38 : @xmath162 for concreteness , we adopt the integration limits of @xmath163 $ ] for @xmath38 and @xmath164 $ ] for @xmath165 ( asl ) , although other choices are possible .
we compute this integral using the monte carlo method , uniformly sampling @xmath38 and @xmath165 . and tidal disruption fraction @xmath166 ( see eq . [ eq : integrated_fraction ] ) for a range of planet masses @xmath39 and dimensionless tidal dissipation strengths @xmath111 .
planet masses are @xmath167 , 0.5 , 1 , 3 and 5 @xmath168 and all planet radii are taken to be @xmath169 .
the three panels correspond to @xmath170 .
[ fig : integrated_fractions],title="fig:",scaledwidth=49.0% ] and tidal disruption fraction @xmath166 ( see eq .
[ eq : integrated_fraction ] ) for a range of planet masses @xmath39 and dimensionless tidal dissipation strengths @xmath111 .
planet masses are @xmath167 , 0.5 , 1 , 3 and 5 @xmath168 and all planet radii are taken to be @xmath169 .
the three panels correspond to @xmath170 .
[ fig : integrated_fractions],title="fig:",scaledwidth=49.0% ] and tidal disruption fraction @xmath166 ( see eq .
[ eq : integrated_fraction ] ) for a range of planet masses @xmath39 and dimensionless tidal dissipation strengths @xmath111 .
planet masses are @xmath167 , 0.5 , 1 , 3 and 5 @xmath168 and all planet radii are taken to be @xmath169 .
the three panels correspond to @xmath170 .
[ fig : integrated_fractions],title="fig:",scaledwidth=49.0% ] fig .
[ fig : integrated_fractions ] shows @xmath171 and @xmath172 for different planet masses and tidal dissipation strengths . for the fiducial dissipation parameter ( @xmath173 ) ,
the hj fraction of jupiter - mass objects is @xmath174 ( for @xmath31 au ) . for saturn - mass objects ,
this fraction becomes is much smaller for a wide range of @xmath36 . for smaller @xmath111 , the `` hot planet '' fraction of saturn - mass objects is @xmath175 , while only planets with masses above @xmath159 can successfully turn into hot planets .
it is only in the extreme dissipation regime ( @xmath176 ) that all gas giants across the explored mass range can turn into hot planets at rates of a few percent .
this behavior can be easily understood from the definitions of @xmath177 and @xmath178 ( eqs .
[ eq : mig_distance ] and [ eq : dis_distance ] ) . for a given @xmath111 ,
varying the planet mass by a factor of a few does not change significantly the total migration fraction ( @xmath179 ) because @xmath177 is essentially unchanged .
however , the relative importance of @xmath180 and @xmath181 is changed as @xmath182 depends on @xmath39 . similarly , changing @xmath111 while keeping
@xmath39 fixed does not change @xmath181 , but only @xmath180 , as @xmath183 grows slightly with stronger dissipation .
these trends are in agreement with the extensive monte carlo experiments of asl . in that work ,
the computed total migration fractions ( hjs plus tidal disruptions ) depend very weakly on @xmath111 . in order to compare directly with the results of asl
, we integrate @xmath184 over a range of planet semi - major axes . assuming that @xmath36 is distributed uniformly between @xmath185 and @xmath186
, we have : @xmath187 using the values @xmath188 au and @xmath189 au , we compute the integrated migration fractions @xmath190 , @xmath191 and @xmath12 for three planet masses ( @xmath1920.3 , 1.0 and 3 @xmath168 ) and three dissipation parameters ( @xmath193 , 10 and 100 ) . the results
( see table [ tab : integrated_fractions ] ) agree with the findings of asl .
in particular , the total migration fraction is confirmed to be very robust and to lie in the range between @xmath194 and @xmath195 .
= 0.07 cm lccccccccccc & & @xmath193 & & & & @xmath173 & & & & @xmath176 & + + @xmath196 & @xmath197 & @xmath198 & @xmath199 & & @xmath197 & @xmath198 & @xmath199 & & @xmath197 & @xmath198 & @xmath199 + + @xmath200 & 0 & 0.8 & 1.8 & & 1.2 & 2.1 & 2.8 & & 2.6 & 3.1 & 3.6 + @xmath201 & 12.4 & 11.6 & 10.8 & & 12.4 & 11.6 & 10.8 & & 12.4 & 11.6 & 10.8 + @xmath202 & 12.4 & 12.4 & 12.6 & & 13.6 & 13.7 & 13.6 & & 15 & 14.7 & 14.4 + despite the robustness of the hj migration fraction found by asl , this number is not in perfect agreement with the results of other recent works , such as @xcite and @xcite ( both considered only @xmath203 ) .
the main difference with @xcite is in the choice of @xmath204 in equation ( [ eq : dis_distance ] ) . on the other hand , @xcite used @xmath113 , the same as asl .
however , these two last works differ in the properties of the assumed underlying population of companions ( see section 6.2 of asl for a discussion ) . here
, we ( and asl ) have assumed the binary orbits to follow uniform distributions in @xmath38 and @xmath205 , with @xmath206 $ ] and @xmath207 $ ] .
following one of the examples of @xcite , we choose @xmath208 $ ] and @xmath209 $ ] , and find @xmath210 , @xmath211 for a total of @xmath212 , in agreement with the results of monte carlo experiments .
the procedure outlined in the previous section can be applied to rocky planets by simple modification of the planet mass - radius relation , love number @xmath49 and dimensionless tidal strength @xmath111 . for solar system
rocky planets @xmath213s @xcite , i.e. , @xmath214 , and @xmath215 @xcite .
we take the generic mass - radius relation from @xcite for an earth - like " composition and calculate @xmath216 and @xmath217 as above
. the successful migration rate of hot planets " is , analogous to hjs , @xmath218 . in fig .
[ fig : rocky_integrated_fractions ] , we show @xmath219 ( in blue ) and @xmath217 ( in red ) for different rocky planets , ranging in mass from @xmath220 to @xmath221 .
the integrated hot planet fraction @xmath222 ( for @xmath36 between 1 and 5 au ) is between @xmath4 and @xmath223 .
thus , these planets are more likely to survive lk migration than saturn - mass planets .
the reason for this is that there is enough room between @xmath183 and @xmath182 ( eqs .
[ eq : mig_distance ] and [ eq : dis_distance ] ) to accomodate the planets into successfully migrated orbits . rewriting @xmath183 and @xmath182 in units more suitable for rocky planets , we have @xmath224 and @xmath225 interestingly , @xmath183 is very close to the value obtained for a jupiter - mass planet ; however , @xmath182 is smaller , since the higher density of rocky planets pushes the roche limit closer to the star .
thus , provided @xmath111 is high enough ( in this case 6000 ) , some small but non - negligible fraction of rocky planets can survive lk oscillations and migrate into close - in orbits . note that , for @xmath220 rocky planets , eqs .
( [ eq : mig_distance_rocky ] ) and ( [ eq : dis_distance_rocky ] ) indicate that the lk migrated hot super - earths should reside within an orbital period range of 0.7 to 4 days .
note that the dimensionless tidal dissipation parameter @xmath111 is highly uncertain .
tidal dissipation in the earth is dominated by its oceans ( e.g. , * ? ? ? * and references therein ) , which does not necessarily apply to models of migrating super - earths with uncertain amounts of gas envelopes .
if we assume a smaller tidal dissipation strength ( @xmath226 , the hot planet formation rates are reduced by a factor of two ( see the caption of fig . [
fig : rocky_integrated_fractions ] ) .
such small success rates in producing close - in rocky planets are in rough agreement with the @xmath227-body experiments of @xcite , who implemented a constant - phase - lag ( or constant tidal quality factor @xmath228 ) model to study lk migration of small planets in the binary system @xmath229 cen b. in that study , only @xmath230 integrations out of @xmath231 ended up in close - in planets . if we write the occurrence rate of hot rocky planets formed via lk migration as ( see equation [ eq : occurrence_rate ] ) @xmath232 , and assume that the planet bearing fraction of rocky planets within a few au is @xmath233 @xcite , then , with @xmath234 and @xmath235 ( see fig .
[ fig : rocky_integrated_fractions ] ) , we find @xmath236 .
thus , an intriguing possibility is that of lk migration playing a role in the origin of short - period rocky planets , especially for single - planet systems for @xmath237 au and a stellar companion located at 150 au ( see also * ? ? ?
* ) . for smaller planet masses , lk oscillation may operate for each planet independently .
observationally , ultra - short - period ( @xmath238 days ) rocky planets are found to have an occurrence rate of @xmath239 in the _ kepler _ catalog ( e.g. * ? ? ?
* ) , and the occurrence rate of rocky planets with slightly longer periods ( a few days ) is appreciably higher @xcite . while many of these are expected to be members of multiple - planet systems @xcite , some may be true `` singles '' @xcite .
lk migration of rocky planets may contribute to this population of close - in planets . and tidal disruption fraction @xmath240 as in fig .
[ fig : integrated_fractions ] but for rocky planets .
planet masses are @xmath241 , @xmath242 , @xmath243 and @xmath221 , with corresponding radii of @xcite @xmath244 , @xmath245 , @xmath246 and @xmath247 .
when integrating @xmath248 and @xmath240 over @xmath36 between 1 and 5 au , the global integrated fractions are , respectively , @xmath249 @xmath250 , @xmath251 , @xmath252 and @xmath253 . if , instead of @xmath214 we use @xmath254 , the hot planet fractions are reduced by half : @xmath249 @xmath255 , @xmath256 , @xmath257 and @xmath258 .
[ fig : rocky_integrated_fractions],scaledwidth=49.0% ]
we have derived a simple analytical method for calculating migration fractions of close - in planets via lidov - kozai oscillations in stellar binaries .
an important ingredient of the method is the relationship between the maximum eccentricity @xmath34 that can be achieved in lk oscillations as a function of the initial planet - binary inclination ( equation [ eq : max_ecc_oct ] ) the effects of short - range forces and octupole potential play an essential role .
our method can be easily implemented for various planet and stellar / companion types , and consists of the following steps : * for given initial planet semi - major axis @xmath36 and binary orbital parameters ( @xmath259 and @xmath260 ) , calculate the `` migration eccentricity '' @xmath261 , with @xmath106 ( the required pericenter distance in order for tidal dissipation to induce efficient orbital decay ) given by equation ( [ eq : mig_distance ] ) . * knowing @xmath137 and the octupole parameter @xmath87 , calculate the critical planet - binary inclination angle for migration , @xmath128 , using equation ( [ eq : solid_angle2 ] ) .
* assuming isotropic distribution of the binary orientations , the migration fraction is @xmath262 .
the tidal disruption fraction is similarly calculated from @xmath263 and the hj ( hot planet ) formation fraction is @xmath264 . * for a distribution of @xmath36 , @xmath37 and @xmath38 , the integrated migration / disruption fractions can be obtained using eqs .
[ eq : integrated_fraction ] and [ eq : global_integrated_fraction ] .
the results of our calculations are in good agreement with the recent population studies of hj formation via lk migration in stellar binaries @xcite .
they can be easily modified to reflect different types of planets / binaries and distributions of the binary s orbital elements .
this agreement confirms that the total migration fraction of planets ( the sum of hj and tidal disruption fractions ) via the lk mechanism is primarily determined by the initial geometry of the hierarchical triple , and depends weakly on the properties of the planet @xcite .
the production fraction of close - in planets , on the other hand , does depend on the internal properties of the planet ( e.g. , mass - radius relation and the strength of dissipation ) , and it decreases with decreasing mass in the case of gas giants , resulting in a natural absence of hot saturns " .
in addition , we find that rocky planets , provided they are sufficiently dissipative , may migrate more effectively than saturn - mass planets .
it is possible that some fraction of the observed short - period rocky planets are produced via lk migration in stellar binaries .
we thank kassandra anderson for useful discussions and cristbal petrovich for comments on the manuscript . this work has been supported in part by nsf grant ast-1211061 , and nasa grants
nnx14ag94 g and nnx14ap31 g . | lidov - kozai oscillations of planets in stellar binaries , combined with tidal dissipation , can lead to the formation of hot jupiters ( hjs ) or tidal disruption of planets .
recent population synthesis studies have found that the fraction of systems resulting in hjs ( @xmath0 ) depends strongly on the planet mass , host stellar type and tidal dissipation strength , while the total migration fraction @xmath1 ( including both hj formation and tidal disruption ) exhibits much weaker dependence .
we present an analytical method for calculating @xmath0 and @xmath2 in the lidov - kozai migration scenario .
the key ingredient of our method is to determine the critical initial planet - binary inclination angle that drives the planet to reach sufficiently large eccentricity for efficient tidal dissipation or disruption .
this calculation includes the effects of octupole potential and short - range forces on the planet .
our analytical method reproduces the resulting planet migration / disruption fractions from population synthesis , and can be easily implemented for various planet , stellar / companion types , and for different distributions of initial planetary semi - major axes , binary separations and eccentricities .
we extend our calculations to planets in the super - earth mass range and discuss the conditions for such planets to survive lidov - kozai migration and form close - in rocky planets .
[ firstpage ] binaries : general planets : dynamical evolution and stability star : planetary system |
supernova 1987a ( sn 1987a ) in the large magellanic cloud provides an unprecedented opportunity to observe the birth and development of a supernova remnant . _ international ultraviolet explorer _ ( _ iue _ ) observations @xcite found narrow line emission about 80 days after the explosion , demonstrating the presence of circumstellar gas around sn 1987a . images taken with the _ hubble space telescope _ ( _ hst _ ) showed that this gas consists of an equatorial inner ring ( radius @xmath10 light year , @xmath11 ) and two outer rings ( @xmath12 times the size of the inner ring , @xmath13 ) tilted towards the observer at @xmath14 @xcite .
the circumstellar ring system was excited by the ionizing radiation from the supernova during the shock breakout @xcite . in the interacting winds model ,
the sn 1987a ring system is part of a bipolar shell around the supernova @xcite . the first signal of ongoing interaction between the sn 1987a debris and the circumstellar gas was the rebirth of the supernova in x - ray ( beuermann , brandt , & pietsch 1994 ; gorenstein , hughes , & tucker 1994 ; hasinger , aschenbach , & trmper 1996 ) and radio @xcite wavelengths around day 1000 .
the size of the radio - emitting region indicated that the supernova debris expanded unimpeded at velocity @xmath15 up to @xmath16 day 1000 before slowing down to @xmath17 by the interaction @xcite . the x - ray and radio observations
have been explained as the interaction of the supernova ejecta with a rather dense ( @xmath18 ) region that separates the shocked stellar wind of the supernova progenitor from the denser gas of the inner ring @xcite .
the `` main event '' of the birth of the supernova remnant 1987a ( snr 1987a ) the interaction between the supernova debris and the circumstellar rings has been anticipated since the discovery of the circumstellar gas .
predictions of the time of the first contact ranged from 2003 @xcite to @xmath19 ( luo , mccray , & slavin 1994 ) to @xmath20 @xcite .
there have been previous model predictions of radiation from this impact in x - rays @xcite and in the uv / optical @xcite .
the first definitive sign of impact between the supernova blast wave and the inner ring was detected in the 1997 april _
hst_/space telescope imaging spectrograph ( stis ) spectral images as a blueshifted ( @xmath21 ) @xmath9 feature at position angle ( pa ) @xmath22 of the ring @xcite .
subsequent analysis of the _ hst_/wide field and planetary camera 2 ( wfpc2 ) images taken in 1997 july showed that a point emission , located in projection at @xmath23 of the distance to the ring , was increasing in brightness over a wide range of wavelengths @xcite and could be traced back to as early as 1995 march ( day 2932 ) @xcite .
the position of the brightening spot , located just inside the edge of the inner ring , suggested that this is the result of an interaction of the supernova blast wave with an inward protrusion of the ring .
this brightening spot , named spot 1 , has increased in flux by a factor of @xmath24 between 1996 and 1999 @xcite . with a number of new spots ( @xmath25 in 2000 november )
since then detected all around the inner circumstellar ring @xcite , we are now witnessing the full birth of snr 1987a .
we present here _
hst_/stis uv and optical spectroscopy of spot 1 taken by the supernova intensive study ( sins ) collaboration up to day 4606 ( 1999 october 7 ) .
results from an earlier ( day 3869 , 1997 september 27 ) stis spectrum of spot 1 have been presented in @xcite .
we describe the new observations and data reduction in [ sec - obs ] and report the results in [ sec - data ] . in [ uv ] we describe a detailed method to determine the intrinsic fluxes and widths of a few uv lines , including the @xmath261394 , 1403 and @xmath261548 , 1551 doublets , which are strongly affected by interstellar line absorption along the line of sight to the supernova .
our working model ( figure [ fig : introhydro ] ) for spot 1 , as well as the other spots , is that it is caused by the impact of the supernova blast wave with a dense inward protrusion of the ring @xcite . when the blast wave overtakes such an obstacle , slower shocks are transmitted into it . since a range of densities is present in the ring , we expect ( [ sec - globalhydro ] ) that the transmitted shocks will have a range of velocities ( @xmath27 @xmath28 ) .
not all of these shocks are responsible for the observed uv / optical emission from spot 1 though .
while some uv and optical line emission is produced right at the shock front , a much larger amount is produced if the shocked gas undergoes thermal collapse , i.e. the shock becomes radiative ( [ sec - shockstruc ] ) . the time it takes for a shock to make this transition increases with its speed and decreases with its pre - shock density .
once the post - shock gas collapses , a shock becomes an extremely efficient radiator of uv and optical emission lines ( [ sec - emissionlines ] ) .
therefore , while the range of possible velocities present in the ring is large , we are only observing the shocks which are slow and/or dense enough to have become radiative .
@xcite confirmed this general picture and found that the observed line widths and line intensity ratios indicated the emission was formed by radiative shocks in the velocity range @xmath29 entering into a dense ( @xmath30 ) gas .
nebular analysis of the observed emission lines of spot 1 ( [ sec - nebanal ] ) suggests that the emission comes from a region of high density ( @xmath31 ) , this result confirms our picture that this emission comes from radiative shocks , which can compress the pre - shock gas by a factor @xmath32 . in [ sec - shockmodels ]
, we model the uv line fluxes from spot 1 between day 3869 and day 4587 with the one - dimensional steady state shock code of @xcite .
the observed uv fluxes are fitted well with models containing emission from two radiative shocks with @xmath6 and @xmath7 .
we propose two scenarios to interpret the observed increase of uv flux with time of spot 1 . in the first scenario
the density of the obstacle is low enough so that the cooling times of shocks entering the obstacle are comparable to the age of the spot .
the increase in uv line fluxes are then due primarily to the aging of the shocks as they develop thermally collapsed layers . in the second scenario , the obstacle is dense enough so that all the shocks cool almost immediately . the increased observed fluxes are instead attributed to an increasing surface area of shock interaction .
while the second scenario fits the observed fluxes better , we suspect that the actual light curves probably manifest a combination of both effects . in [ sec - lineprof ] we discuss the observed line profiles and present simulated @xmath9 line profiles based on simple geometric models for the shock interaction . in [ sec - disc ] we discuss what we have learned by comparing results of simplified shock models with the spectral observations , and describe how future observations may elucidate some unsolved problems .
we summarize the main results in [ sec - conclusion ] .
the stis observations of spot 1 in both optical and uv wavelengths obtained by the sins team are listed in table [ tb - obs ] .
sn 1987a is located in a densely populated region of the lmc and appears to belong to a loose , young cluster region @xcite .
target acquisition was complicated by the stars present near the supernova , especially star 3 , a be star of @xmath33 16 , at 163 away and pa = 118 , and star 2 , a b2 iii star of @xmath34 = 15.0 , at 291 away and pa = 318 from the supernova @xcite .
we decided to peak - up on the nearby star s2 @xcite and offset the telescope to center the aperture on spot 1 .
we measured the required offset from the wfpc2 images @xcite . due to the uncertainties in this measurement , spot 1
was located at 008 off the center of the slit for all observations taken before 1999 august .
we reduced these data using the standard calibration files which assumed that the object was located at the center of the slit .
we estimate that the offset from the center of the slit will cause us to underestimate the measured flux by @xmath35 for the 02 data , and by @xmath36 for the 05 data . with each grating setting
, we took multiple ( @xmath37 ) observations centered at dithered positions 05 apart along the slit .
cosmic rays ( in the case of optical data ) and hot pixels ( in optical and uv data ) were removed simultaneously when the dithered raw images were combined with the calstis software developed by the stis investigation definition team at the goddard space flight center .
previous narrow - slit stis spectra processed by the sins team with this software showed that the flux calibration of far - uv ( g140l ) and near - uv ( g230l ) data agree to @xmath38 for the overlapping region , while the agreement between near - uv ( g230l ) and optical ( g430l ) data is good to @xmath35 @xcite .
the location and orientation of the aperture positions are shown in figure [ fg - slitpos ] .
for all but one of the observations , the slit was oriented ( within @xmath39 ) along the axis connecting the center of the sn 1987a debris and spot 1 , which is located at a pa@xmath40 on the inner ring @xcite . the only exception was the 1999 october g140l observation ( figure [ fg - slitpos]b ) , where the ( @xmath41 ) slit had a pa of 55@xmath42 and did not pass through the center of the sn 1987a debris . in all observations ,
the spot 1 spectrum overlapped with that from the adjacent segment of the inner circumstellar ring that was included within the stis aperture . with an expansion velocity of @xmath43 @xcite ,
the ring was not resolved spectrally in any of the stis observations reported here . in the optical data ,
the emission from the ring is the main source of background to the spot 1 spectrum and will be discussed in [ opt - low ] and [ opt - med ] .
@xcite measured the width of spot 1 in wfpc2 images up to 1999 april 21 ( day 4440 ) and showed that spot 1 was unresolved in the data and was consistent with a point source at optical wavelengths .
we compared the full width at half - maximum ( fwhm ) of spot 1 emissions in our last stis observations at 1999 september with those of point sources that were recorded in the data .
we found that the fwhm of spot 1 was @xmath44 and @xmath45 larger than a point source in the far - uv and optical wavelengths , respectively .
the latter result is consistent with measurements by @xcite with the 2000 may 1 ( day 4816 ) stis g750 m spectroscopy , which suggests that spot 1 is now moderately resolved in the _ hst _ data at optical wavelengths .
we extracted the low resolution optical spectrum of spot 1 from the stis g430l and g750l two - dimensional spectral images .
portions of the 1999 september g430l and g750l data taken with the 02 slit are shown in figures [ fg - data](c ) and [ fg - data](d ) , respectively .
the horizontal streaks near the center in the figure are broad emission lines from the inner supernova debris , which have a fwhm velocity @xmath46 @xcite . for these data ,
the lower section of the spectral image is the combined spot 1 and inner ring emission - line spectrum ( hereinafter referred to as the spot 1+north - ring , or _
s1+nr _ , spectrum ) . at the kinematic resolution of the g430l and g750l grating settings ( @xmath47 ) , neither the emission lines from spot 1 [ @xmath48 , @xcite ] or those from the ring ( @xmath49 ) are resolved .
the upper section of the spectral image is the emission - line spectrum of the segment of the inner ring subsection that is in the slit and directly opposite that of spot 1 ( hereinafter referred to as the south - ring , or _ sr _ , spectrum ) .
we measured the _
s1+nr _ and _ sr _ spectra in the g430l and g750l grating settings by integrating the @xmath50 rows of the image in which the emission - line spectra appeared .
emissions due to the diffuse lmc background in the balmer lines , [ ] @xmath263727 , 3729 , and [ ] @xmath264959 , 5007 are observed as images of the entire slit at those wavelengths in figures [ fg - data](c ) and [ fg - data](d ) .
we subtracted the contribution of this diffuse emission from the spot 1 spectrum by linear interpolation above and below the extracted rows .
since the subtracted lmc background was only a small fraction of the emission from spot 1 ( @xmath512% for [ ] @xmath525007 ) , this subtraction did not significantly increase the uncertainty of the estimated line fluxes .
we measured the flux of each emission line in the _ s1+nr _ and _ sr _
spectra by fitting a gaussian to the line profile , allowing the net flux , width , and central wavelength to vary independently for each line .
we estimated the background level from a linear fit over a 45-pixel region of the spectrum centered on the line in question but excluding any emission - line features .
we fitted the line flux by minimizing the total @xmath53 in which the coefficients defining the line and the background were free parameters .
gaussian profiles gave satisfactory fits to all the line profiles .
the reduced-@xmath54 , or @xmath55 , ( the total @xmath54 divided by the number of degrees of freedom ) , of our line fits were within the range @xmath56 , compared to the value of 1.0 for a statistically good fit .
we computed the flux of each line and its error from the best - fit parameters and their associated uncertainties .
we adjusted the statistical errors of all the line fluxes so that @xmath57 for all fits . in cases where two or more emission lines overlapped in wavelength , such as [ ] @xmath525007 + @xmath525016 , and [ ] @xmath527136 + [ ] @xmath267155 , 7172 , we fitted the emission features with multiple gaussians with the additional constraint that their wavelength separations were the known differences of laboratory wavelengths . only a subset ( @xmath58 ) of the emission lines observed in the _ s1+nr _ spectrum also appeared in the _ sr _ spectrum .
therefore , we attributed entirely to spot 1 the measured fluxes of emission lines that were seen in the _ s1+nr _ spectrum but not in the _ sr _ one . to subtract the _
_ spectrum from the _ s1+nr _ spectrum , we assumed that the fluxes of emission lines in the _ nr _ spectrum are equal to those in _ sr _
spectrum scaled by factors that are independent of time .
this assumption is reasonable because the rate of flux decrease around the ring has been shown to be relatively constant around the ring @xcite .
we estimated the scale factors by examining archival wfpc2 emission - line images in @xmath9 , [ ] @xmath525007 , and [ ] @xmath526583 obtained in 1994 february and 1994 september , before spot 1 appeared .
we measured flux ratios @xmath59 of 1.2 , 1.1 , and 1.3 , respectively from the @xmath9 , [ ] @xmath525007 , and [ ] @xmath526583 images .
we used the @xmath9 scale factor for the balmer lines . while we used the [ ] @xmath525007 factor for the [ ] @xmath264363 , 4959 , 5007 lines , and
likewise for the [ ] lines , we recognized the increase in systematic errors in the measured spot 1 fluxes of [ ] @xmath524363 and [ ] @xmath525755 because these lines are more temperature sensitive than the other lines .
for all remaining emission lines , such as [ ] and [ ] , we assumed a scale factor of @xmath60 in order to subtract the _
nr _ spectrum .
in addition to the 02 data sets described above , we obtained one observation of spot 1 with the 05 slit and the g750l grating setting .
as before , we extracted the combined _ s1+nr
_ spectrum by integrating the 6 rows of the image where the emission lines appeared . for the emission lines produced predominantly by spot 1 , such as @xmath526678 and [ ] @xmath527135
, we measured the _
_ fluxes by fitting the line emissions with single gaussian profiles .
for the lines where emission from both _
s1+nr _ were apparent , such as [ ] @xmath525755 and [ ] @xmath526300 , we estimated the contribution of the _ nr _ flux to the _ s1+nr _ flux from a linear interpolation of the _ nr _ emission - line flux adjacent to spot 1 in the slit . after subtracting this background , we fitted the remaining _ s1 _ emission lines with gaussian profiles .
the uncertainties in the background ring flux estimated in this procedure resulted in larger systematic errors in the estimated spot 1 line fluxes for this observation .
a medium resolution optical spectrum was taken on 1999 august 30 ( ut , 4368.0 days since explosion ) with a 01 slit and the g750 m ( 6581 ) grating setting ( @xmath61 ) . with a spectral resolution of @xmath62 , emission lines from spot 1 , with @xmath48 @xcite , were resolved in the data , while the emission from the unshocked inner ring , with @xmath63 , remained unresolved .
three observations of 7800 seconds each were taken at three parallel slit positions , pointed so that the middle slit position was centered on spot 1 and the two other slit positions were immediately adjacent [ cf .
figure [ fg - slitpos](d ) ] .
therefore the observation covered a segment of the ring of length 03 . with the crowded stellar field near sn 1987a , we did not execute the acquisition - peakup exposure for these 01 slit observations as suggested by the _ hst_/stis operation manual @xcite .
instead all three adjacent slit positions were placed on spot 1 by blind offsets .
therefore we can not apply the standard pipeline data reduction procedures to process these data . to determine the fluxes of emission lines from spot 1
, we first removed the wavelength - dependent aperture throughput correction function in the pipeline data for each 01 slit spectrum .
we then summed the fluxes from the three 01 slit positions , and multiplied the total flux by a new aperture correction function for an equivalent 03 slit , calculated by interpolating the pipeline corrections for the 01 , 02 , and 05 slits . for @xmath9 ,
the aperture correction led to a @xmath64 decrease of flux over the sum of the fluxes measured in the three slit positions .
we calculated the corresponding 1@xmath65 errors by combining the individual errors in quadrature . a section of the spectral image from the middle slit position is shown in figure [ fg - data](e )
again , the central horizontal streak is emission from the sn 1987a debris .
emissions from the inner ring at the two positions where the ring intersected with the slit aperture were observed in [ ] @xmath266300 , 6364 , @xmath9 , [ ] @xmath266548 , 6583 , and [ ] @xmath266717 , 6731 . in the lower section of the spectral image , emission from the stationary inner ring overlapped with the broadened emission from spot 1 .
again , the major source of contamination the spot 1 spectrum is the emission from the inner circumstellar ring within the 01 slit . as described by @xcite , we fit all the spot 1 emission features with gaussian functions .
emission from the stationary ring dominated the spectral profile near zero velocity ( @xmath66 ) and was excluded from the fit .
the majority of the line profiles could be fitted well with gaussian profiles , such as the fit to the [ ] @xmath526583 line emission shown in figure [ fg - gaussfit ] .
the sole exception was the @xmath9 line profile , where the signal was strong enough to show noticeable departures from a gaussian profile , as we shall discuss further in [ sec - lineprof ] .
we obtained low resolution uv spectra of spot 1 with the g140l and g230l grating settings .
@xcite have already presented results from the first g140l far - uv observations in 1997 september 27 taken with the 05 slit .
we detected emission lines from spot 1 in @xmath261239 , 1243 , @xmath261394 , 1403 , ] @xmath521400 , ] @xmath261483 , 1487 , @xmath261548 , 1551 , and @xmath521640 .
we detected the same set of emission lines in 1999 february 27 , observing again with the 05 slit . in 1999
october 7 , observing with the 02 slit , we also detected the @xmath521335 multiplet , [ ] @xmath521602 , and ] @xmath261661 , 1666 .
figure [ fg - data](a ) shows a section of the spectral image from this observation .
radiation from spot 1 is evident in the lower portion of the image whereas only faint line emission from the inner circumstellar ring can be seen in the upper half of the displayed image .
the broad ( @xmath67 ) @xmath68 radiation comes from the reverse shock from the interaction between the supernova debris and the region located inside of the equatorial ring @xcite .
fluxes of uv line emission from the inner ring are much less than those from spot 1 and make a negligible contribution to the measured fluxes .
this is also the case for the near - uv emission lines measured in the 1999 september 17 g230l 02 observation , shown in figure [ fg - data](b ) .
we measured the far - uv and near - uv spectra of spot 1 from the g140l and g230l data , respectively , by procedures similar to those we described in [ opt - low ] .
we fitted emission lines with gaussian profiles and linear backgrounds , except for the broad @xmath68 emission underlying the @xmath261239 , 1243 doublet which we fitted with a quadratic function .
we fitted the two components of close doublets such as @xmath261239 , 1243 , @xmath261548 , 1551 , ] @xmath261661 , 1666 , ] @xmath262139 , 2143 , and @xmath262796 , 2803 with gaussians constrained to have fixed doublet separations , identical widths , and line ratios dictated by their oscillator strengths . at the resolution of the g140l grating setting , the @xmath521403 emission of the @xmath261394
, 1403 doublet is blended with the ] @xmath521400 multiplets .
we fit the combined and ] feature with multiple gaussians , requiring the doublet to meet the same constraints as the other close doublets .
the observed fluxes of a few uv lines , such as @xmath261394 , 1403 and @xmath261548 , 1551 , were reduced by interstellar line absorption .
we describe our procedure for correcting for this absorption in [ uv - lines ] below .
table [ tb - optflux ] lists the measured spot 1 optical emission line fluxes , including previously published results from the 1998 march 7 ( day 4030 ) data by @xcite and 1@xmath65 upper flux limits for the [ ] @xmath263726 , 3729 , and [ ] @xmath265198 , 5200 doublets .
the 1@xmath65 errors tabulated are only statistical errors .
systematic effects , such as fringing for the g750l data towards the near - ir region ( @xmath69 , leitherer et al .
2000 ) , might contribute additional uncertainties to the measured fluxes .
the tabulated fluxes have been dereddened with e(@xmath70 ) of 0.16 @xcite and the extinction law of @xcite with an assumed r@xmath71 of 3.1 . in the optical band
the differences between the lmc extinction law and the galactic law are negligible at low color excess @xcite .
the interstellar extinction correction applied is listed in the last column of table [ tb - optflux ] .
spot 1 was observed in many neutral and lowly ionized species in the optical wavelengths .
we did not detect any coronal lines such as [ ] @xmath526375 .
most of the line fluxes increased with time at a rate comparable to that measured from wfpc2 photometry @xcite . at the low spectral resolution of both the g430l and g750l observing modes ( @xmath72 )
, definitive line identification remained a problem , especially for the [ ] emission lines .
the fe line identifications in table [ tb - optflux ] are based upon the modeling of the spot 1 fe lines in pun et al .
( 2002 , in preparation ) .
moreover , several lines were blended .
table [ tb - optflux ] lists possible contributing species and , in the case of [ ] lines , different multiplets .
in contrast , line blending is not a problem in the medium resolution g750 m ( @xmath73 ) data .
fluxes in [ ] @xmath266300 , 6363 and @xmath526678 were measured with the low resolution g750l and medium resolution g750 m gratings in 1999
september within 17 days of each other .
the two measurements agreed within uncertainties for the [ ] doublet .
the two results differed by @xmath74 for the low s / n @xmath526678 data , but were also within the noise level .
this difference is probably a good indication of the detection limits of such faint lines . for the medium resolution g750 m observations , apart from the line fluxes , we were also able to measure the peak emission velocities ( @xmath75 ) and the widths ( @xmath76 ) of the emission lines from the profile fits .
table [ tb - optwidth ] lists the results .
the peak emission velocity measurements have been adjusted for the sn 1987a heliocentric velocity of @xmath77 @xcite .
the 1998 march 02 g750 m results have also been adjusted for the off - center position of spot 1 within the aperture ( cf .
[ sec - obs ] ) .
the main uncertainties in the measurements of both @xmath75 and @xmath76 are due to the contributions by emission from the stationary circumstellar ring , which dominated the emission by spot 1 near zero velocity for many species ( cf .
figure [ fg - gaussfit ] ) .
the errors due to this contribution are generally smaller for the 1999 august 01 observations than the 1998 march 02 ones
. for all emission lines , the line centroids from spot 1 were blueshifted , with peak velocity @xmath78 lying within the range @xmath79 to @xmath80 .
this result is consistent with the overall physical picture in which spot 1 is located on the near side of the equatorial ring @xcite and the shock entering spot 1 is moving towards the observer .
we found no evidence that the peak velocity @xmath78 of spot 1 varied with time .
most emission lines had widths ( fwhm ) within the range @xmath81 , except [ ] @xmath526583 and @xmath9 , which had a slightly greater width , @xmath82 .
we detected no measurable change with time of the line widths except for @xmath9 and [ ] @xmath526548 .
for @xmath9 , the emission profile from the second observation in 1999 august could no longer be fit well by a single gaussian ( to be discussed further in [ sec - lineprof ] ) .
the width of the [ ] @xmath526548 emission , measured in 1999 august was 1.5 times greater than that measured in 1998 .
however , we are inclined to attribute this increase to systematic error , since we detected no such increase in the other [ ] component at 6583 , where the line fluxes were @xmath12 times stronger . near the time of outburst ,
interstellar absorption lines of @xmath521335 multiplet , @xmath261394 , 1403 , @xmath261548 , 1551 , and @xmath262796 , 2803 were observed in the uv continuum of sn 1987a with _ iue _ operating in the high resolution ( @xmath83 ) echelle mode @xcite . for each line ,
the dominant absorption component was centered at @xmath84 , near the sn 1987a heliocentric velocity of @xmath77 @xcite , and had a fwhm of @xmath85 .
therefore , narrow emission lines ( @xmath86 ) from the inner circumstellar ring from these species are completely blocked by the interstellar absorption @xcite , as demonstrated by the absence of and emission lines from the ring in figure [ fg - data](a ) .
emission lines from spot 1 , with @xmath87 , are not totally blocked by these interstellar absorption lines , as correctly predicted by @xcite .
however , the line profiles are altered and the observed fluxes are reduced substantially . therefore , we must correct the measured , , , and emission line fluxes to account for this absorption .
the appropriate correction factors depend on the profile shapes of both the spot 1 emission and the intervening absorption . to make this correction
, we assumed that the uv emission lines from spot 1 had gaussian profiles with the same parameters as the optical lines as measured in the medium resolution optical data ( [ opt - med ] ) .
we then estimated the amount of flux reduction in the uv emission lines by multiplying the assumed gaussian profiles by the absorption profiles seen in the _ iue _ data .
the corresponding flux correction factors for the various emission lines are shown in figure [ fg - fcorr](a ) .
the correction factor is greatest for the @xmath262796 , 2803 doublet , where the interstellar absorption is almost completely saturated between @xmath88 and @xmath89
. the flux correction factor due to interstellar absorption is sensitive to the assumed widths ( @xmath76 ) and peak ( @xmath75 ) velocities of assumed line profiles from spot 1 .
figure [ fg - fcorr](b ) illustrates this dependence for the important case of @xmath261548 , 1551 .
we see that the flux correction factor increases moderately with decreasing @xmath76 for @xmath90 , but becomes very sensitive to both @xmath76 and @xmath75 for @xmath91 .
as we shall discuss below in [ sec - lineprof ] , the peak velocities and widths of the emission lines from spot 1 depend on the detailed geometry and hydrodynamics of the shocks entering the spot , which are unknown .
it is not obvious that the optical and uv emission lines should have the same peak velocities and widths .
however , in the plane - parallel shock model that we describe in 4 , we found that the peak velocities of emission from spot 1 were almost identical in both the uv and optical wavelengths .
therefore , we used the measured peak spot 1 velocities from the medium resolution optical observations , @xmath92 ( cf .
table [ tb - optwidth ] ) , to estimate the flux correction factors to account for interstellar absorption of the uv emission lines .
the far - uv emission lines from spot 1 are poorly resolved at the spectral resolution of the g140l grating ( @xmath93 at 1500 ) . for these lines
, we attempted to establish the relation between the actual and observed widths through monte carlo simulations . for the simulations
, we assumed that the intrinsic spot 1 emission lines had ( i ) a gaussian shape with fwhm as a variable parameter , ( ii ) emission peaked at @xmath94 , and ( iii ) flux ratio of the doublets dictated by their oscillator strengths .
the dotted curve in figure [ fg - absorb](a ) shows a model @xmath261548 , 1551 profile , assuming @xmath95 and line flux ratio @xmath96 2:1 .
the solid curve shows a subsection of the high resolution _
spectrum with the absorption line profile for the doublet . for each input fwhm
, we multiplied the model input emission line by the measured _ iue _ absorption spectrum .
figure [ fg - absorb](b ) shows a typical result .
we convolved this profile with the line - spread function ( lsf ) of the detector for the slit aperture of the data set .
the lsf for each emission line was interpolated from the measured lsfs at 1200 , 1500 , and 1600 .
the lsf typically has a narrow peak ( @xmath97 pixel ) and a broad wing that extends to @xmath24 pixels on either side of the peak @xcite .
figure [ fg - absorb](c ) shows a typical example of the lsf for the g140l grating at 1550 with the 02 aperture . for each emission line , we normalized the resulting convolved profile to the photon counts for each observation .
the thick solid curve in figure [ fg - absorb](d ) shows the normalized model line profile for the 1999 october observation .
we constructed a simulated observed profile ( the thin solid curve in figure [ fg - absorb]d ) by applying random noise to the model profile and sampling it at the spectral resolution of the g140l grating .
finally , we fitted the simulated profile with gaussians in the same way as the real data .
the dotted curve in figure [ fg - absorb](d ) shows such a fit . for each input model fwhm velocity , we ran 10000 simulations and generated an array of the corresponding observed line widths .
figure [ fg - mc ] shows the median and the 68.3% ( 1@xmath65 ) upper and lower limits of the array plotted against the model input widths of the various observed emission lines .
we use these results to convert the observed line widths from the stis data , shown as horizontal lines in figure [ fg - mc ] , to the intrinsic widths of the lines emitted by spot 1 .
we did not attempt to model the near - uv g230l data in this way because the spectral resolution for this grating setting ( @xmath98 at 2500 ) was too low for such a procedure to yield useful results .
for all emission lines except the doublet , we found a 1@xmath65 upper limit of @xmath99 , consistent with the measurements from the medium resolution optical data .
the lower limit to the assumed fwhm is important because the flux correction factors for all uv emission lines are sensitive to the input line widths for fwhm @xmath100 .
we took this lower limit to be the same as the measured lower limit for the optical lines , that is , @xmath101 .
table [ tb - uvflux ] lists the fluxes of uv lines from spot 1 inferred from the gaussian line fits ( corrected for extinction ) and the intrinsic line widths derived from the monte carlo simulations .
the tabulated fluxes for the , , , and doublet have been corrected for interstellar line absorption assuming an intrinsic spot 1 line width @xmath102 with peak emission at @xmath103 .
the flux correction factors applied for these lines are also listed in table [ tb - uvflux ] .
we also corrected the fluxes for interstellar extinction by assuming @xmath104 , @xmath105 , and @xmath106 . at uv wavelengths ,
the choice of extinction function is important because the correction is substantial and is known to vary from place to place ( cf .
pun et al .
we used the 30 doradus extinction function of @xcite for the lmc component , and the @xcite function for the galactic component .
the far - uv g140l data had slightly higher kinematic resolution ( @xmath107 ) than the low resolution optical g430l and g750l data .
however , the individual components of @xmath521335 and ] @xmath521400 multiplets remained unresolved in the data .
there were more uncertainties with line identifications in the lower resolution ( @xmath108 ) near - uv g230l data , such as the unidentified emission features near 2737 and 2746 .
on the other hand , we identified emission features near 2324 and 2334 as the @xmath522325 multiplet and @xmath522335 , respectively .
we ruled out the alternative identification of [ ] @xmath262322 , 2332 because the observed line ratios , @xmath109 )/@xmath110 ) @xmath111 and @xmath109 )/@xmath112 ) @xmath113 , were much different from the theoretical [ ] line ratios of @xmath114 and @xmath115 , respectively . these [ ] line ratios are determined only by the atomic transition probabilities and are independent of the excitation model .
the fluxes of all far - uv lines from spot 1 increased with time during the three observations taken from 1997 september ( day 3869 ) to 1999 october ( day 4606 ) .
the rate of increase differed for features from different ions , ranging from @xmath116(3869 d)/@xmath116(4596 d ) @xmath117 for @xmath521240 to 1.9 for @xmath521550 .
we will discuss the time dependence of the uv emission - line fluxes in [ sec - shockmodels ] .
in our working model for spot 1 , we assume that the uv and optical emission lines observed are caused by radiative shocks that develop where the supernova blast wave strikes dense gas protruding inward from the circumstellar ring .
as we shall show , the spectrum and profiles of the emission lines from such shocks are sensitive to the density distribution and geometry of this protrusion , which are probably quite complex and can not be resolved even with the _
hst_. our approach here therefore is to illustrate the salient physics of the spectrum formation with a few idealized toy models , " which we believe will guide us toward a better understanding of the properties of a more realistic model . following the previous work of @xcite and @xcite , we show in figure [ fig : globalhydro ] hydrodynamic simulations for two models of a fast shock overtaking a dense gas cloud . in each model
, we assume that a fast plane - parallel blast wave traveling through a uniform medium of relatively low density overtakes a cloud of substantially greater uniform density .
the cloud boundary is approximated as a density discontinuity . in each simulation
, the blast wave drives a transmitted shock into the cloud , while a reflected shock travels backwards towards the interior of the remnant . as the blast wave overtakes the obstacle , the surface area of the transmitted shock increases .
the transmitted shock propagates with a range of velocities depending on shape and density of the obstacle .
the simulations are calculated using the hydrodynamic code vh-1 @xcite , which is based on the piecewise parabolic method of @xcite .
radiative cooling is included in the code using an operator splitting technique .
we have used a non - equilibrium ionization cooling curve calculated with the plane - parallel shock code described in [ sec - shockstruc ] with abundances typical of the equatorial ring . at the resolution of the simulations
, we are not always able to resolve the cooling time scale of the shocks
. however , this computational limitation does not seriously affect the behavior of the overall hydrodynamics of the interaction .
moreover , we have modeled the blast wave as a single planar shock rather than using the full double - shock structure present in the remnant .
we will comment on the effects of ignoring the full double - shock structure in [ sec - disc ] .
the two scenarios depicted in figure [ fig : globalhydro ] show two different behaviors for the development of radiative shocks in the obstacle .
the important parameter distinguishing their behavior is @xmath118 , where @xmath119 is the typical cooling time for the transmitted shocks , and @xmath120 is the time it takes for the fast shock to cross the obstacle . as we will discuss below in
[ sec - twoshock ] , the interpretation of the observed light curves of lines emitted from the shocks depends on which behavior is dominant in spot 1 .
the simulation on the left of figure [ fig : globalhydro ] ( scenario 1 ) is one in which @xmath119 is comparable to @xmath120 , so that not all of the shocks have become radiative . for this simulation
we assume that the obstacle is a spherical cloud of density @xmath121 and diameter @xmath122 .
this geometry allows the the shock to be driven into the back side of the obstacle .
even after the blast wave has completely overtaken the obstacle , the shocks transmitted into the front - end of the obstacle have not yet undergone thermal collapse and are non - radiative ( nr ) because of the high impact velocity . on the other hand , the shocks transmitted into the sides and back of the obstacle have lower velocities owing to the oblique incidence of the blast wave .
the shocked gas behind these parts of the transmitted shock has a lower temperature and a shorter radiative cooling time .
a dense layer of cooled gas ( r ) is developed as a consequence ( to be discussed further in [ sec - shockmodels ] ) .
scenario 2 in figure [ fig : globalhydro ] represents an example in which the cooling time for the transmitted shock is much shorter than the time scale for the blast wave to cross the cloud @xmath123 , regardless of incidence angle of the blast wave .
we assume that the obstacle is an elongated protrusion with a higher density , @xmath124 . with the high density of the obstacle ,
all the transmitted shocks undergo thermal collapse soon after impact .
we recognize that these idealized models in figure [ fig : globalhydro ] do not represent the true shape and density distribution of the obstacle .
however , given the limited observations at hand , we believe it is more fruitful to explore how well we can fit the data with a few idealized models or combinations thereof , rather than to explore fits of more complicated hydrodynamic models ( see [ sec - disc ] ) .
the driving pressure @xmath125 behind a hypersonic blast wave propagating with velocity @xmath126 through a medium of density @xmath127 is given by @xmath128 .
immediately after the blast wave passes the surface of an obstacle , a transmitted shock propagates into the obstacle with velocity @xmath129 where @xmath130 is the pressure at the surface of the obstacle and @xmath131 is the pre - shock density of the obstacle .
the value of @xmath130 at a given location on the cloud surface rises shapely as the blast wave first strikes it , then decreases as the reflected shock propagates away from it . over a time scale comparable to that for the blast wave to overtake the obstacle , @xmath130 decreases by a factor of @xmath132 @xcite . immediately after the passage of the blast wave ,
the ratio @xmath133 at any given point on the cloud surface depends on two parameters , first , the obliquity angle @xmath134 between the direction of the blast and the inward normal to the surface of obstacle , and second , the pre - shock density ratio @xmath135 between the obstacle and the ambient region .
figure [ fig : beta ] illustrates the dependence on obliquity for a blast wave with a density contrast @xmath136 impacting on a spherical cloud . at the point of first contact , i.e. , @xmath137 , the driving pressure is at its maximum value and @xmath138 . as the obliquity of the impact increases , the driving pressure decreases . with a factor of @xmath24 decrease in pressure due to obliquity ,
the initial velocity of the transmitted shock on the edges of the protrusion is only @xmath139 of that at the tip of the protrusion .
we estimate the pressure behind the blast wave , @xmath125 , from the relation @xmath128 . by fitting the radio remnant of sn 1987a from year 1992 to 1995 with a shell model , @xcite measured the velocity of the blast wave to be @xmath140 .
using more recent radio observations , @xcite updated the value of @xmath126 to be @xmath141 . however , they also noticed that in data taken since @xmath142 , the radio remnant can no longer be fit well with the simple shell model but instead with a combined shell and multiple point - sources model . @xcite
estimated that this leads to a 30% uncertainty in the measured expansion velocity .
on the other hand , by modeling the observed x - ray emission from the blast wave before any spots appeared on the circumstellar ring , @xcite found a slightly higher velocity for the blast wave , @xmath143 , and a density of the region inside the ring , @xmath144 .
these results are consistent with the estimates of @xcite and the upper limit @xmath145 determined by @xcite .
we estimate the blast - wave pressure to be within the range @xmath146 @xmath147 .
the pre - shock density , @xmath131 , of the obstacle is even more uncertain .
the density of the ring has been determined from the rate of fading of the optical and uv emission lines @xcite . at the time of emission maximum ( @xmath16 day 350 ) , the radiation was dominated by gas of relatively high density , @xmath148 .
however , the emission from the higher density gas faded rapidly and the emission at later times was dominated by gas of lower density , @xmath149 @xcite .
the gas in spot 1 might have even higher density than the value derived from optical and uv lines .
the fact that spot 1 lies inside the inner circumstellar ring might be due to the fact that it resisted ablation from the ionizing radiation and stellar wind of the progenitor as a result of its enhanced density .
we summarize the possible range of transmitted shock velocities in figure [ fig : phasespace1 ] .
the shaded gray region shows the range of transmitted shock velocity , @xmath150 , as a function of the pre - shock density of the obstacle , @xmath131 , for our best estimate of the blast - wave pressure ( @xmath151 , @xmath144 ) .
the upper boundary is the velocity of the transmitted shock at the tip of the protrusion and the lower boundary is the velocity at the side .
we also show in figure [ fig : phasespace1 ] the corresponding shock velocity range for our high estimate of the blast - wave pressure ( @xmath152 , _ dashed _ ) , and our low pressure estimate ( @xmath153 , _ dotted _ ) .
assuming a pre - shock obstacle density of @xmath154 , it is apparent that a range of shock velocities ( @xmath155 ) can be present in the obstacle . in this section
we review the relevant physics of shock fronts that we need to interpret the emission - line spectrum of spot 1 .
we use the 1991 version of the raymond shock code @xcite for illustration .
the code calculates the non - equilibrium ionization and excitation of the post - shock flow in a one - dimensional steady state shock for any given input parameters such as shock velocity and pre - shock densities . for any specified electron ( @xmath156 ) and ion ( @xmath157 ) temperatures at the shock front , the code calculates their equilibration through coulomb collisions .
we assume a ratio of @xmath158 for our models .
the code calculates the upstream equilibrium preionization and the downstream photoionization of the gas and derives the post - shock density , temperature , and ionization structure in the gas .
the code also calculates the local emissivity throughout the shock and the integrated fluxes over the entire column behind the shock front , including continuum and line emission from both allowed and forbidden transitions of h , he , c , n , o , ne , mg , si , s , ar , ca , fe , and ni . we adopted the lmc abundances measured by @xcite in all elements except for he , c , n , and o , where the ring abundances derived by @xcite are used instead .
the other exception is si where the abundance of @xcite is used because of the large uncertainty in the measurements by @xcite .
we call this set of values the `` ring '' abundance ( h : he : c : n : o : ne : mg : si : s : ar : ca : fe : ni = 1 : 0.25 : 3.24 @xmath159 : 1.82 @xmath160 : 1.58 @xmath161 : 4.07 @xmath159 : 2.95 @xmath159 : 2.51 @xmath159 : 5.01 @xmath162 : 1.95@xmath162 : 7.76 @xmath163 : 1.70 @xmath159 : 1.10 @xmath162 ) .
we show the post - shock temperature and density structure for protons and electrons in our model shock in the upper panel of figure [ fig : shock ] .
we assume a shock velocity of @xmath164 , characteristic of the shocks producing the radiation seen from spot 1 @xcite .
the shock has a high mach number ( @xmath165 ) , and compresses the obstacle gas by a factor of @xmath16 4 at the shock front , providing that the magnetic field is negligible .
the temperature of the ions after crossing the shock front is given as @xmath166 , where @xmath167 is the mass of the ion and @xmath168 is the boltzmann s constant .
this implies that each ionic species will have a different post - shock temperature , provided that there are processes to completely thermalize the post - shock distribution functions .
for collisional shocks , the post - shock electron temperature is orders of magnitude lower than the ion temperatures @xcite .
on the other hand , in collisionless shocks , plasma turbulence ( e.g. , @xcite ) can partially equilibrate the post - shock electron and ion temperatures .
eventually , the temperatures will be equilibrated by coulomb collisions to a mean shock temperature @xmath169 , where @xmath170 is the mean atomic weight per particle and @xmath171 is the mass fraction of helium . in the region ( the `` ionization zone '' ) immediately behind the shock front with a characteristic length of a few ionization lengths ,
the atoms are collisionally ionized . since radiative processes in this region
remove a negligible fraction of the thermal energy of the hot plasma , this part of the shock is called the `` adiabatic '' or `` non - radiative '' zone . given sufficient time the radiative losses will remove the thermal energy and cause a runaway thermal collapse of the shocked gas . in order to maintain pressure equilibrium across the shock in this `` cooling region , ''
an increase in density will accompany with a temperature decrease in the region .
radiative cooling in the gas continues until its temperature ( @xmath172 ) is too low for the excitation of uv resonance lines which are responsible for the rapid cooling .
moreover , once the gas starts to recombine , it quickly becomes optically thick to the ionizing radiation ( mostly extreme uv lines ) produced upstream in the cooling zone .
roughly half of this ionizing radiation propagates downstream and is reprocessed into optical emission lines in this photoionization zone , " where the gas is maintained at a temperature @xmath173 as a result of a balance between photoionization heating and radiative cooling .
the density of the gas in the photoionization zone is given by @xmath174 .
for a @xmath164 shock , where @xmath175 , the gas in the photoionization zone can be compressed by a factor @xmath176 .
however , a significant magnetic field entrained in the gas may mitigate this compression @xcite .
shocks that have developed cooling and photoionization zones are called `` radiative shocks .
'' a shock will be radiative if it has propagated for a characteristic cooling time , @xmath119 , which is defined as the time required for the shocked gas to cool from is post - shock temperature to @xmath177 . otherwise , we call it a `` non - radiative shock . ''
we calculate the cooling time @xmath178 for shocks in our models for input shock velocities @xmath179 and plot the results in figure [ fig : cooltime ] .
we fit a power law to these results and obtain the relation @xmath180 where @xmath131 is the pre - shock density .
therefore , for the range of shock velocities and pre - shock densities expected in spot 1 , which was less than 4 years old when our observations were made , some shocks may have already become radiative while others may remain non - radiative . after entering the shock front
, atoms are collisionally ionized until they come into equilibrium with the post - shock gas .
these ions are collisionally excited and emit line radiation , some of which from lower ionization stages than the final ionization state of a species .
this ionization process in the ionization zone is not in equilibrium .
emission from this region lead to the detection of `` balmer filaments '' in other supernova remnants @xcite .
the balmer emission comes from the shock front , where the h atoms are ionized .
when ions reach equilibrium with the post - shock gas , the dominant line emission are in far uv or soft x - ray wavelengths , depending on the post - shock temperature .
most of the uv and optical line emissions are produced after the shock starts its thermal collapse .
while the observed high - ionization uv lines are produced in the cooling region , the low - ionization uv and optical lines are produced in the photoionization zone , which is a dense region that is illuminated by the harder ionizing spectrum created upstream .
we plot the integrated surface emissivities of @xmath521240 , @xmath521550 , [ ] @xmath266548 , 6584 , and @xmath9 of our shock model described above in [ sec - shockstruc ] as functions of downstream column density in the lower panel of figure [ fig : shock ] .
most of these lines ( @xmath181 ) are emitted during or after the thermal collapse of the shock , rather than in the ionization zone directly behind the shock front .
therefore , the resulting spectrum and magnitude of the emission from a shock depend greatly on whether the shock is radiative or non - radiative and the cooling time of a shock is a good indication of the time it takes for a shock to `` light up '' .
the luminosity of a line produced by a shock in @xmath182 is given by @xmath183 where @xmath131 is the pre - shock density , @xmath150 is the shock speed , @xmath184 is the age ( time since first encounter ) of the shock , and @xmath185 is the surface area that the shock covers .
the function @xmath186 represents the fractional efficiency for a shock to convert its thermal energy into emission for a given line .
as shown in figure [ fig : shock ] , @xmath186 can be represented as a step function which turns on at @xmath187 .
we show the relation between the emission efficiency @xmath186 and shock velocity @xmath150 for several emission lines in figure [ fig : eta ] .
the thick lines in figure [ fig : eta ] represent shocks that have completely cooled , that is , @xmath188 , while the thin lines are shocks with @xmath189 .
shocks with velocities @xmath190 have not yet cooled by @xmath191 yrs and therefore their structures are truncated when compared to the fully developed radiative shocks .
this shows again that radiative shocks are far more efficient radiators than non - radiative shocks .
therefore , while non - radiative shocks may be present in the protrusion , their net contribution to the observed uv and optical emission from spot 1 will be negligible . for emission lines such as @xmath521240 ( formed in the cooling region ) , and @xmath9 ( formed primarily in the photoionization zone ) ,
the line efficiency @xmath186 has little dependence on the pre - shock density @xmath131 .
in contrast , forbidden lines are subject to collisional supression and therefore their emissivities are sensitive to @xmath131 . the effect of supression
is illustrated in figure [ fig : eta ] by the [ ] @xmath266548 , 6584 lines , which have a critical electron density of @xmath192 @xcite .
effects from collisional supression increases as the shock velocity increases because faster radiative shocks generate more compression .
figure [ fig : eta ] also shows that line emissions such as @xmath521240 can be produced only if the temperature of the shocked gas is high enough for that ion to be produced , as previously discussed in @xcite . above that threshold , the total line emissivity increases linearly with the shock velocity . permitted
emission lines formed in the photoionization zone have a stronger dependence on shock speed , since they are generated by the reprocessing of the ionizing photons produced upstream in the cooling layer .
for example , the @xmath9 emissivity for fully developed radiative shocks increases approximately as @xmath193 . since the gas is relatively cool ( @xmath172 ) in the photoionization zone , the optical lines have thermal widths of order @xmath194 , where @xmath195 is the element s atomic mass .
macroscopic motion of the cooled layer will cause the observed line profiles of the optical lines to be significantly broadened . as the blast wave wraps around a protrusion
, we expect to observe velocity components traveling both toward and away from us .
the measured widths of the optical lines in the early _ hst_/stis spectrum suggests that the fastest radiative shock has a projected velocity @xmath196 @xcite .
moreover , the observed line ratios of spot 1 indicate that lower velocity ( @xmath197 ) shocks must also be present ( cf .
[ sec - oneshock ] ) . following the discussion in
[ sec - velrange ] , we are not surprised that a range of shock velocities is required to explain the observations .
figure [ fig : phasespace2 ] shows the boundaries separating radiative and non - radiative shocks with @xmath119 of 2 and 4 years overlayed on the hydrodynamically allowed regions of phase space for different values for the blast - wave pressure . in order for shocks with velocities as low as @xmath198 to be radiative , the density of the obstacle @xmath199 has to be higher than most of the gas in the equatorial ring .
a lower @xmath131 is needed if the blast - wave pressure is lower than our nominal estimate .
while faster shocks propagating into less dense gas may also be present in spot 1 , they do not contribute significantly to the observed optical and uv spectra because they are non - radiative .
the _ hst_/stis spectrum of spot 1 consists of many forbidden emission lines of various species in the uv and optical wavelengths ( cf . table [ tb - optflux ] and [ tb - uvflux ] ) .
a standard nebular analysis on these data provides a good starting point because it indicates the typical values for the basic physical quantities that can be expected in the line - emitting region .
we decide to focus our attention on the data taken around 1999 september ( @xmath16 day 4570 ) when a complete set of spectrum is obtained from @xmath0 . for each ion
we consider a five - level model and include all relevant atomic processes , such as collisional excitation and de - excitation , and spontaneous radiative transitions .
the atomic data of @xcite are used for the analysis , except for , where the results from @xcite are used .
we construct a grid of line ratios as functions of electron number density @xmath200 and temperature @xmath156 .
figure [ fig : nebanal ] shows the contours for which ratios of lines of [ ] , [ ] , [ ] , [ ] , and [ ] agree with the observed values
. the observed ratios of various forbidden lines require a relatively high electron density , @xmath2 , and temperature , @xmath1 .
the upper limits for the line ratios of [ ] ( @xmath201 ) and [ ] ( @xmath202 ) are consistent with the derived high @xmath200 , as well as the ] ( @xmath203 ) line ratio .
recalling that the typical number density of the ring is @xmath204 @xcite , our result indicates that the gas in spot 1 must have been compressed by a factor much greater than 4 .
this is consistent with our general picture ( [ sec - interp ] ) that the uv and optical line emissions from spot 1 originate from a highly compressed gas behind a radiative shock .
our results also suggest the temperature and density stratification of the region where these emission lines are formed .
this is also consistent with our general picture for radiative shocks where emission lines from different ionic species are created in different regions of the shock .
as discussed earlier ( [ sec - emissionlines ] ) , the optical emission lines from a radiative shock comes from a photoionization zone in the cooled post - shock gas .
this photoionization picture is supported by the observed balmer decrement in spot 1 , @xmath9 : @xmath205 : @xmath206 : h@xmath207 , which is generally consistent with both case a and case b values typically seen in photoionized gaseous nebulae @xcite .
a complete analysis of the h and he lines will be presented in pun et al .
( 2002 , in preparation ) .
emission lines from the higher ionization stages ( e.g. , , , , ) are formed in the cooling region of a shock , where the gas temperature and density are stratified and the ionization levels can be out of equilibrium .
however , within each ionization stage , the population of all energy levels are in equilibrium .
we model these lines with a radiative shock code that properly account for these effects .
we compare our model results with the observations of six uv emission lines detected with the g140l grating ( @xmath261238 , 1243 , @xmath521393 , 1403 , ] @xmath521400 multiplet , ] @xmath521486 , @xmath261548 , 1543 , and @xmath521640 ) .
we choose these lines because they have negligible contamination from the ring emission , and because they have been observed at three different epochs from day 3869 to day 4596 .
we decide not to model the @xmath262796 , 2803 doublet , the strongest emission line in the spot 1 spectrum , because of the large uncertainties in the observed flux ( over a factor of 2 , cf .
table [ tb - uvflux ] ) , caused by huge interstellar absorption along the lmc line of sight ( cf .
[ uv - lines ] ) . moreover , with the line flux measured only one epoch
, we are not able to monitor its time evolution as for the other fuv lines . similarly , we have not attempted models to fit both the uv and the optical emission lines because of the limited time and wavelength coverage of the optical data ( cf . table [ tb - optflux ] ) . for a given set of input shock - model parameters ( shock velocity , pre - shock density , and abundances ) ,
we follow the evolution of the downstream gas until the shocked gas has cooled to 5000 k. we store the integrated flux for various emission lines at intermediate grid points behind the shock as the basic vectors by which we compare with the observed data .
we then vary the parameters in the models to obtain the best fit . to seek the best fit
, we allow the abundances to vary from our standard ring " abundance described in
[ sec - shockstruc ] . using a grid of single - shock models with @xmath208
, we were unable to reproduce the emission - line fluxes and widths observed from spot 1 at any epoch .
one of the main challenges for the single - shock model is to explain the relatively low ratios of @xmath521240 flux to the fluxes of lines from lower ionization stages ( e.g. , ] @xmath521486 and @xmath521550 ) in light of the widths ( @xmath209 ) of the observed lines . the abundance of increases rapidly for shock velocities exceeding @xmath210 , while and are present at lower shock velocities ( cf .
figure [ fig : eta ] ) . as a result ,
the model line ratios @xmath521240/ @xmath521550 @xmath211 and @xmath521240/ ] @xmath521486 @xmath212 for shocks faster than @xmath198 exceed the observed ratios of @xmath213 and @xmath214 , respectively .
the fact that this discrepancy exists for both line ratios indicates that this result does not depend on our choice of abundances .
we therefore conclude that a single - shock model can not reproduce the observed line emissions of spot 1 , as in our earlier studies @xcite .
we are not too surprised that single - shock models fail to fit the observed emissions from spot 1 because we expect there exists a range of shock velocities @xmath150 in the interaction ( cf .
[ sec - velrange ] ) .
the exact distribution of @xmath150 depends on the geometry and density distribution in the obstacle , which are both unknown . rather than attempting to explore the vast parameter space of possible shock velocity distributions
, we fit the observed line emissions with models consisting of only two distinct shocks , each having a different velocity and surface area .
we explore parameter space with shock velocities ranging from 100 to @xmath215 and find reasonable fits to the relative line strengths with a combination of one `` slow '' shock with @xmath216 and one `` fast '' shock with @xmath217 @xmath215 . as discussed below in [ sec - lineprof ] , the observed line profiles suggest that the fastest radiative shocks have velocities @xmath218 .
we therefore assume this value for the fast shock component .
the hydrodynamic simulations shown in figure [ fig : globalhydro ] of [ sec - globalhydro ] illustrate two possible scenarios to account for the observed increase of the line fluxes with time .
the crucial parameter distinguishing these scenarios is the ratio @xmath219 , where @xmath184 is the age " of the interaction ( i.e. , the time since first impact ) , and @xmath119 is the characteristic radiative cooling time for the transmitted shocks . for the case where both the blast - wave pressure and the obstacle density are low , then @xmath220 and the shocks in the obstacle would be in the process of developing their radiative layers . in this scenario
the observed increase of the line fluxes is due primarily to development of new radiative layers as the shocks age .
alternatively , for the case where the density of the obstacle is high , then @xmath221 and the shocks quickly become radiative .
the observed increase of the line fluxes is then due to the increase in the surface area of the shocks as the blast wave overtakes more of the obstacle .
we suspect that we are observing a combination of these two behaviors . given our limited data set ( 6 lines in 3 epochs ) , we decide to fit only two limiting models which probably bracket the actual situation : evolution due solely to shock aging ( model 1 ) , and evolution due solely to increasing shock areas ( model 2 ) . for model 1
we assume that the area of each shock remains constant over time and the only fitting parameters are the surface area and age of each of the two shocks .
we assume that the faster shock is older because it represents the head - on shock that appears when the blast wave first encounters the obstacle , and that the slower shock is younger because it is driven into the sides of the obstacle by an oblique blast wave at a later time .
for this model we need to assume a low pre - shock density @xmath222 so that @xmath223 for the fast shock .
we also assume that the pre - shock density is constant throughout the obstacle for both models .
the best - fit model for model 1 has a @xmath224 and its parameters are listed in table [ tab : models ] .
the model ( _ squares _ ) and observed ( _ error bars _ )
fluxes for all the 6 emission lines are plotted in figure [ fig : models ] . for the best - fit model , the surface area of the slower shock needs to be @xmath225 times greater than that of the faster shock to fit the low observed @xmath521240/@xmath521550 flux ratio . in this model ,
the fast shock was 2.2 years old at the time of the first uv observation ( day 3869 ) , while the slower shock was only one week old .
the age of the fast shock is only slightly smaller than the results of @xcite , who suggest that spot 1 was at least 2.6 years old by the time of our first uv observation .
this pleasant result is offset by the suspiciously young age of the slower shock .
for model 2 we assume that the pre - shock density is high enough so that all the shocks are fully radiative at all epochs .
the fitting parameters more this model are the areas of the two shock surfaces at each of the three observation epochs .
we find that the quality of the fits is rather insensitive to the assumed pre - shock density ( @xmath226 is assumed ) .
this is true because the 6 uv lines we are fitting are not subject to collisional supression ( including ] @xmath521486 , for which @xmath227 ) , so their emission efficiencies , @xmath186 , do not depend on the pre - shock density ( cf .
[ sec - emissionlines ] ) .
the luminosities of such lines are actually proportional to the product , @xmath228 , of the pre - shock density and shock area .
we present the best - fit paramters of the model in table [ tab : models ] and the model line fluxes in figure [ fig : models ] ( _ circles _ ) .
since model 2 provides a better statistical fit ( @xmath229 ) to the observed line strengths than model 1 , we conclude that the increase of shock areas is the dominant cause of the increase of the fluxes .
figure [ fig : area ] shows the time dependence of the fitted shock areas ( actually the product @xmath228 ) for model 2 .
similar to model 1 , we find that @xmath225 times more surface area must be covered by the slow shocks than by the fast shocks . while our best fit model results are statistically consistent with a constant ratio of slow to fast shock area ( @xmath230 in table [ tab : models ] ) , they suggest that this ratio may be decreasing with time .
this behavior is not what we would expect from simple protrusion models shown in figure [ fig : globalhydro ] . in such models
the fast shocks are created first and only the area of slow shock interaction increases as the blast wave overtakes more of the obstacle .
one possible explanation is that there may be many protrusions or clouds distributed in spot 1 and therefore the net area of fast shocks can increase with time as more of them are encountered by the blast wave .
another possible explanation arises from the fact that model 2 by construction does not allow for any aging of shocks .
however , the real light curve of the spot 1 emission is probably a combination of increased shock areas and additional emission from older , but newly cooled , radiative shocks .
the addition of fast shock area at later times seen in model 2 can therefore be attributed to shocks that were created before our first observation but added their emission later after they had cooled and become radiative .
our finding that more area is covered by the slow shocks than fast shocks suggests that the obstacle may have an elongated shape with an aspect ratio @xmath231 , where @xmath232 and @xmath233 , are the shock areas for the slow and fast shocks , respectively . assuming a simple , hemispherical geometry for the obstacle , we estimate that it has a typical scale length of @xmath234^{1/2}$ ] .
we find a best - fit size of the obstacle @xmath235 at day 4606 , the time of our last stis observation .
this is consistent with our finding ( cf .
[ sec - obs ] ) that spot 1 remains spatially unresolved in our data .
while the two - shock models successfully fit the time evolution of the uv emission lines , these models fail to account for the observed fluxes of the optical lines ( e.g. , @xmath9 , @xmath205 , [ ] @xmath266548 , 6583 , [ ] @xmath266300 , 6364 , and [ ] @xmath264959 , 5007 ) emitted in the photoionization zone ( [ sec - nebanal ] ) .
the observed optical fluxes are typically factors of @xmath236 greater than those predicted by our models .
we exclude emission from the precursor as a possible contributor to the extra emission because its emission would be narrow and center at zero velocity .
however , we exclude that portion of the line profile when we fit the spot 1 medium resolution g750 m data ( cf .
[ fg - gaussfit ] ) .
we discuss other possible explanations for this discrepancy in
[ sec - disc ] . finally , we emphasize that the satisfactory fits of the uv line fluxes were possible only if we allow abundances to vary from our standard `` ring '' abundance .
the inferred abundances are roughly consistent with the ring abundances inferred by @xcite , but differences as large as a factor of two are derived . in this section
we discuss what can be learned from the widths and profiles of the observed emission lines .
the data are sparse .
we measure line widths of spot 1 accurately for only a few optical lines ( cf .
[ opt - width ] ) .
we also only have rough estimates of the widths of a few far uv lines ( cf .
[ uv - flux ] ) .
the line widths and profiles are dominated by the shock dynamics , and not thermal broadening .
the line - emitting gas behind a radiative shock travels at approximately the shock speed and will therefore emit line radiation with a doppler shift determined by the line - of - sight velocity of the shock . for emission lines originated from the photoionization zone ( @xmath237 k ) , e.g. , the optical forbidden lines ,
the thermal broadening is @xmath238 , which is much smaller than the observed widths of the lines ( @xmath196 ) .
the thermal broadening for emission lines formed in the cooling region , such as @xmath521240 , while larger than that for the optical forbidden lines , is still much smaller than the dominating doppler broadening effects of the macroscopic motions of the shock .
although it is poorly measured , the width of @xmath521240 line ( table [ tb - uvflux ] ) appears to be higher than that of all the other optical and uv emission lines , with the possible exception of @xmath9 .
this result can also be explained by figure [ fig : eta ] , which shows that the @xmath521240 emission can only be produced in shocks faster than @xmath210 , while all the other lines can be produced by even slower shocks .
the fact that the measured widths of the [ ] , [ ] , and [ ] lines are smaller than @xmath9 ( table [ tb - optwidth ] ) indicates that the faster shocks produce these lines with lower efficiencies ( compared to @xmath9 ) than the slower shocks .
this is caused by collisional supression of the forbidden lines , which increases with increasing compression ratio of the photoionized zone , as shown previously in figure [ fig : eta ] .
the line profile shapes depend on the geometry of the shock surface and its orientation relative to the observer .
the shocks having the highest projected velocities toward and away from the observer produce the blue and red wings of the lines , respectively .
the doppler shifts of the wings of the line profile therefore provide a lower limit to the velocity of the fastest radiative shocks present while both fast transverse shocks and slow shocks contribute to the cores of the line profiles .
we construct simulated @xmath9 line profiles based on simple geometries for the protrusion ( e.g. , figure [ fig : globalhydro ] ) in order to derive constraints on the hydrodynamics model of the interaction .
the 1999 august 30 ( day 4568 ) g750 m 01 line profile of @xmath9 is chosen because it is well - observed and is not affected by collisional supression .
figure [ fig : linegeom ] shows our model shock surface geometries . for each model
we assume a distribution of shock velocities ( from 100 to @xmath7 ) normal to the surface of the protrusion .
we normalize each velocity distribution by requiring that the relative areas of fast ( @xmath239 ) and slow ( @xmath240 ) shocks in each model is @xmath241 , as determined from the best - fit two - shock models ( [ sec - twoshock ] ) .
we also assume that the axis of symmetry of the protrusion points radially inward from the circumstellar ring at the location of spot 1 for all models .
we calculate line profiles by convolving the surface emissivity ( @xmath242 , cf .
[ sec - emissionlines ] ) of the radiative shocks with the fractional areas covered by shocks having the appropriate line - of - sight velocities .
we then convolve the model profile with the stis lsf @xcite . at the @xmath9 wavelength ,
the stis lsf for the g750 m 01 aperture can be well approximated with a gaussian of fwhm 1.5 pixels .
the simulated profiles are finally normalized for comparison with the observed @xmath9 profile .
figure [ fig : lineprof ] shows the resulting @xmath9 line profiles ( _ solid _ ) from the geometry models of figure [ fig : linegeom ] . for each case
the day 4568 @xmath9 profile ( _ squares _ ) is also plotted for comparison .
the observed line width indicates that radiative shocks are present with projected velocities @xmath209 , which is a good indication of the velocities of the fastest radiative shocks present .
the line profile formed from an ellipsoidal model surface ( model a ) has multiple peaks and steep edges that are not apparent in the observed profiles .
the peaks are due to caustics in velocity space that occur where the shock surface lies tangent to the plane of the sky .
these features are less pronounced in the line profiles produced by a spherical model surface ( model c ) , where tangential surfaces comprise of a lower fraction of the total surface areas . for each case we also show a model in which no radiation comes from the tip of the protrusion ( models b and d , for elliptical and spherical surfaces , respectively ) .
these models are motivated by the observations from the _ chandra x - ray observatory _ @xcite , which show a bright spot in the x - ray image of sn 1987a coincident with spot 1 .
one possible interpretation is that the x - rays come from a faster ( @xmath243 ) non - radiative shock at the tip of the protrusion responsible for spot 1 .
we notice that this picture is also consistent with one where there is a density gradient across the ring .
while the model line profiles have similar widths as the observed @xmath9 line , they do not provide a good fit in the high velocity portion ( @xmath244 and @xmath245 ) of the line profile . as we shall discuss in [ sec - disc ] , radiative shocks are subject to instabilities that will introduce `` turbulence '' in the radiative layer .
we simulate this effect of this turbulence qualitatively by convolving the derived model line profiles with gaussians of fwhm @xmath246 .
the resulting line profiles ( _ dashed lines _ in figure [ fig : lineprof ] ) provide substantially better matches to the observed @xmath9 line .
while the idealized radiative shock models described in [ sec - interp ] and [ sec - anal ] provide a plausible framework for interpreting the spectrum of spot 1 , they do not account for all the details of the observed line ratios and profiles . in this section , we consider what the departures of the actual spectra from the model spectra may tell us about how the actual physical situation differs from our models .
first , we discuss the `` turbulent '' broadening that we introduce in [ sec - lineprof ] to make the model line profile resemble the observed one at high velocity .
this turbulence is almost certainly due in part to the cooling instabilities that exist behind radiative shocks with @xmath247 @xcite . to illustrate the effects of these instabilities , we construct a one - dimensional vh-1 simulation of a blast wave driving a radiative shock into gas of constant upstream density , @xmath248 and
the results are shown in figure [ fig : therm1d ] . in the figure , the upstream gas flows to the left with a velocity @xmath249 while the radiative gas , shown as a dark band , propagates to the right with a mean velocity @xmath196 .
the dark gray parabolas represent the positions of the actual shock front , which moves ahead of the radiative layer and then collapses back to the radiative layer over time . the velocity of the shock front relative to the upstream gas varies from @xmath250 to @xmath251 .
this process repeats itself on a time scale comparable to the cooling time of the shocked gas . in our simulation ,
the forward shock never stabilizes while the oscillation of the velocity of the radiative layer decreases as it gains mass with time .
the instabilities that exist behind radiative shock become more dramatic in the two - dimensional simulations , as described in @xcite .
we construct a 2-d vh-1 shock simulation in figure [ fig : therm2d ] to illustrate such effects .
we seed the instability with a 5% sinusoidal density perturbation of the unshocked gas in a direction perpendicular to the shock front , with the purturbation wavelength comparable to the cooling length of the transmitted shock .
as the shock enters the dense gas , a richtmyer - meshkov instability develops rapidly at the contact discontinuity .
after several cooling times ( and forward shock oscillations ) , the radiative layer is shredded into a chaotic flow of tiny blobs by a combination of thermal and rayleigh - taylor instabilities .
if the blast wave strikes the obstacle at oblique incidence , the flow of shocked gas parallel to the contact surface will cause strong kelvin - helmholtz instabilities to shred the surface even further and will introduce a component of velocity parallel to this surface in the radiative layer .
we can see these instabilities beginning to develop in the last frames of figure [ fig : globalhydro ] .
however , we have not attempted to model their consequences in our simulations of the line profiles .
since the _ hst _ only allows us to resolve structures with dimensions @xmath252 cm at the distance of sn 1987a , we have no direct knowledge of the actual shape or the density distribution of the protrusion gas responsible for the spot 1 emission . even in the case where it has a simple shape such as that shown in figure [ fig : globalhydro ] , the instabilities discussed above will introduce turbulent structures at scale lengths @xmath253 cm , which we can not hope to resolve with current technology .
moreover , it is unclear whether the assumption that the protrusion gas has a simple shape with a smooth surface is justified .
while we know that the circumstellar ring has a range of densities @xcite , we do not know the scale lengths upon which these inhomogeneities may be distributed .
if the blast wave encounters a `` lumpy '' medium , a complex flow will develop where the radiating blobs are hammered both by the reflected shocks and by the blast wave itself .
clearly , the current observations allow for many plausible scenarios for the hydrodynamics of the radiating gas for which quantitative models are beyond the scope of this paper .
we are not surprised that the simple hydrodynamic models we have considered in [ sec - lineprof ] fail to reproduce the observed line profiles .
magnetic fields may also affect the shock dynamics .
for the range of densities and shock velocities expected in spot 1 , we estimate that a magnetic field with strength @xmath254 g entrained in the upstream gas would be sufficient to suppress compression in radiative shocks and stabilize the cooling instabilities @xcite .
this will result in a photoionization zone of lower density . in our idealized shock models of [ sec - interp ] and
[ sec - anal ] , we have only considered radiative shocks in which the pressure driving the shocks is constant . however , as @xcite and @xcite have shown , the pressure behind an actual blast wave encountering an obstacle have a complex time dependency .
as the reflected blast wave travels away from the surface of the obstacle the flow diverges and the driving pressure decreases .
later , the pressure driving the radiative shock into the obstacle may suddenly rise again owing to echo " shocks produced when the reflected shock strikes the contact discontinuity and the reverse shock .
the most troubling failure of our models is their inability to account for the observed ratio of fluxes of optical to uv lines .
as we have described , the radiative shock models consistently predict optical fluxes that are factors @xmath255 lower than expected from models that are normalized to give the observed uv line fluxes .
a simple way to account for the discrepancy between the observed and model optical / uv line ratios is to assume that the extinction of spot 1 is greater than that previously determined for sn 1987a .
if we applied a reddening of @xmath256 , the corrected uv flux would increase by a factor @xmath255 , sufficient to remove the discrepancy .
considering that spot 1 is located on the far side of the ring from the observer , a column density @xmath257 of gas with a 30 doradus dust / gas ratio in the foreground of spot 1 ( perhaps the ring material itself ) could account for such additional reddening .
it is also possible that the shock itself may destroy graphite dust grains and elevate the ratio of c / n abundances in spot 1 ( compared to the unshocked ring ) to account in part for the relatively low observed ratio of @xmath521240/ @xmath521550 .
shocks faster than @xmath258 can destroy graphite grains in a timescale @xmath259 @xcite .
the revised value of @xmath260 is 0.04 ( and @xmath261 ) greater than the measurement of @xmath262 by @xcite on the nearby star 2 , and 0.07 greater than the value we adopted for table [ tb - uvflux ] .
however , modeling of the uv and optical narrow emission lines for the inner circumstellar ring by @xcite does not support such large @xmath260 value .
therefore we do not regard it to be highly probable . a second way to account for
the optical / uv line ratio is the presence of an extra source of ionization in the photoionization zone that is roughly three times stronger than the radiation from the shocked gas itself . in the one - dimensional radiative shock models described in
[ sec - interp ] , the only source of ionizing radiation for the photoionized zone is the shocked gas immediately upstream from that layer .
one possible candidate for the extra ionization is the faster shocks which have not had time to develop their own radiative layers .
however , we are not confident that such models will work .
the problem is that the presumed extra ionizing radiation must also illuminate the cooling zone where the uv emission lines are formed , and this illumination will elevate the abundance of in the cooling layer . as a result , shocks with velocity @xmath4 will produce extra @xmath521240 emission , causing the predicted @xmath521240 flux to again exceed the observed value . a more promising candidate for the extra source of ionization
is the nonthermal radiation from particles accelerated in the shock itself .
detailed modeling is required to demonstrate its feasibility .
a third possible explanation for the failure of the models to account for the ratios of optical to uv emission lines is the inadequacy of the stationary radiative shock models themselves .
our findings are similar to that from the modeling of the spectra of herbig - haro objects with simple plane - shock and bow - shock models , where high - ionization lines , such as and , are underproduced @xcite . as we have shown in figure [ fig : therm2d ] , radiative shocks are inherently unstable and the ratio of optical to uv emission lines from such unstable shocks may be different from that in a stationary flow , even in a time - averaged or ensemble - averaged sense .
for example , the temperature gradients in the unstable flow are more pervasive than in a stationary flow , in which case thermal conduction may provide an additional energy source .
future observations with the _ hst _ may be very helpful in resolving some of the issues that we have raised here .
valuable clues to resolving the optical / uv line ratio puzzle may come from observations of the same ratio in the other hot spots .
for example , if the ratio is affected by extinction by dust in the immediate foreground of spot 1 , it might have different values in the other spots , some of which lie on the near side of the circumstellar ring .
we can also benefit from observations of the uv emission lines at a higher spectral resolution .
first , we can better estimate the amount of flux reduction caused by the lmc interstellar absorption ( cf .
[ uv - flux ] ) by obtaining the line profile shapes of emissions not affected by this absorption , such as @xmath521240 , and @xmath261483 , 1486 .
this will significantly reduce the error estimates of the measured fluxes affected by this absorption , especially for @xmath522800 , where the flux correction due to lmc absorption applied is both the highest and the most uncertain .
second , to reconcile the observed low flux ratio of @xmath521240/ @xmath521550 , we have suggested a two - shock model in which the higher - velocity radiative shocks make up a relatively small fraction of the shock area compared to the lower - velocity shocks .
this model implies that the ratio of @xmath521240/ @xmath521550 line emissivity should increase in the wings of the lines , where only the fastest shocks contribute . on the other hand ,
we have also suggested that destruction of graphite grains in the faster shocks may elevate the abundance of carbon in the gas phase and lead to a decrease in the ratio of @xmath521240/ @xmath521550 in the wings of the lines .
however , as shown in figure [ fg - absorb ] , the observed profile of @xmath521550 is severely modified by interstellar absorption , introducing large uncertainties in the n / c abundance ratio .
we may get a clear look at the red wing of @xmath521550 by observing the spots on the far side of the ring ( e.g. , spots 3 , 4 , and 5 ) , where the shock emission would be mostly redshifted to wavelengths where the interstellar absorption is not so severe . when spot 1 becomes bright enough to permit observations of the line profile of ] @xmath521486 ,
the flux ratio of ] @xmath521486/ @xmath521240 as a function of position within the line profile will give us a diagnostic of shock excitation versus velocity that is independent of relative abundances and not subject to interstellar absorption .
in conclusion we summarize the main results of this paper : * we describe the _
hst_/stis uv and optical spectrum of spot 1 up to day 4606 ( 1999 october 7 ) .
we measure and tabulate fluxes , line widths , and line centroids for all lines detected in the spectrum . *
we determine correction factors to account for the substantial interstellar line absorption of several lines ( @xmath521335 multiplet , @xmath261394 , 1403 , @xmath261548 , 1551 , and @xmath262796 , 2803 ) .
however , these factors are poorly constrained because of uncertainties in the line widths and centroids . *
the observed emission is caused by radiative shocks which develop when the supernova blast wave strikes dense gas protruding inward from the equatorial ring .
a nebular analysis of several line ratios confirms that many of the emission lines are formed in a region of very high density gas ( @xmath263 ) , which has been compressed by factors @xmath32 due to radiative cooling downstream from the shock . *
the observed line widths indicate that radiative shocks with velocities as high as @xmath7 are present , while the line ratios indicate that much of the radiation must come from slower shocks ( @xmath264 ) .
the inferred range of shock velocities is a natural consequence of a model in which a blast wave overtakes a dense obstacle . *
hydrodynamic arguments show that such slow shock velocities will be present only if the density in the unshocked obstacle is sufficiently high ( @xmath265 ) .
this density lies at the high end of the range of densities observed for the ring by @xcite .
the fact that shocks with velocity @xmath164 must have become radiative after 3 years sets a lower density limit , @xmath30 . *
faster shocks ( @xmath266 @xmath28 ) may also be present in spot 1 , as would be the case for shock interaction with the lower - density gas ( @xmath267 ) present in the ring @xcite .
these shocks would be invisible in the optical and uv emission spectra though because they would not yet have developed radiative layers .
however , they may contribute to the enhancements of x - ray emission seen near the hot spots by the _ chandra x - ray observatory _ @xcite . *
we interpret a subset of the observed emission lines ( 6 uv lines observed at 3 epochs ) with a model consisting of two one - dimensional stationary radiative shocks with different velocities ( @xmath6 and @xmath7 ) . in our favored model ,
the lines brighten because the area of the obstacle overtaken by the blast increases with time .
the area covered by the slower shocks must be @xmath241 times that covered by the faster shocks . to account for the observed line strengths , the obstacle must have a characteristic scale length of @xmath268 . *
our models underestimate the ratio of observed optical / uv line fluxes by a factor of @xmath12 .
this may indicate the presence of an additional source of ionizing flux in the photoionization zone , or that the assumption of steady state shocks is too limiting . *
the observed line profiles are in qualitative agreement with simulated profiles created from simple geometries for the obstacle .
the nearly gaussian observed profiles indicate the presence of chaotic flows in the photoionized region , caused by violent thermal and shear instabilities in the shocked gas . * continued uv and optical observations of spot 1 are necessary to further our understanding of the physics of the radiative shocks present there .
resolving the profile of the @xmath521240 and other uv lines will provide critical tests of the radiative shock model .
* observations and modeling of the emission - line spectra of the other spots now appearing in the ring will provide vital clues to resolving the uncertainties of the present model for spot 1 .
observations with _ hst_/stis are required to spatially resolve their individual spectra .
we thank dan welty for sharing with us the high resolution _ iue _ observations of sn 1987a , kailash sahu for sending us the most recent stis lsf , john raymond for providing us with his radiative shock code , and john blondin for use of his vh-1 code and his help with hydrodynamic simulations for this paper .
support for this work was provided by nasa through grant go-08243 from the space telescope science institute , which is operated by the association of universities for research in astronomy inc .
, under nasa contract nas5 - 26255 .
additional support was provided by nasa through grants nag5 - 3313 and ntg5 - 80 to the university of colorado .
c.s.j.p . acknowledges funding by the stis idt through the national optical astronomy observatories .
is grateful for a guggenheim foundation fellowship .
baron , e. , et al .
2000 , , 545 , 444 beuermann , k. , brandt , s. , & pietsch , w. 1994 , , 281 , l45 bionta , r. m. , et al .
1987 , phys .
lett . , 58 , 1494 blades , j. c. , wheatley , j. m. , panagia , n. , grewing , m. , pettini , m. , & wamsteker , w. 1988 , , 334 , 308 blinnikov , s. , lundqvist , p. , bartunov , o. , nomoto , k. , & imamoto , k. 2000 , , 532 , 1132 blondin , j. m. , & lundqvist , p. 1993
, , 405 , 337 bhm , k .- h . , & goodson , a. 1997 , in iau symp .
182 , herbig - haro flows and the birth of low mass stars , ed . b.
reipurth & c. bertout ( dordrecht : kluwer ) , 47 borkowski , k. , blondin , j. , & mccray , r. 1997a , , 476 , l31 .
1997b , , 477 , 281 burrows , c. j. , et al .
1995 , , 454 , 680 burrows , d. n. , et al .
2000 , , 543 , l149 cargill , p. j. , & papadopoulos , k. 1988 , , 329 , l29 cardelli , j. a. , clayton , g. c. , & mathis , j. s. 1989 , , 345 , 245 colella , r. , & woodward , p. r. 1984 , j. comp .
phys . , 54 , 174 cox , d. p. , & raymond , j. c. 1985 , , 298 , 651 chevalier , r. a. , & dwarkadas , v. v. 1995 , , 452 , l45 chevalier , r. a , & imamura , j. n. 1982 , , 261 , 543 chugai , n. n. , chevalier , r. a. , kirshner , r. p. , & challis , p. m. 1997 , , 483 , 925 crotts , a. & heathcote , s. 2000 , , 528 , 426 cumming , r. j. , & lundqvist , p. 1997 , in advances in stellar evolution , ed .
r. t. rood ( cambridge : cup ) , 297 ensman , l. , & burrows , a. 1992 , , 393 , 742 fitzpatrick , e. l. 1986 , , 92 , 1068 . 1999 , , 111 , 63 fitzpatrick , e. l. , & walborn , n. r. 1990 , aj , 99 , 1483 fransson , c. , cassatella , a. , gilmozzi , r. , kirshner , r. p. , panagia , n. , sonneborn , g. , & wamsteker , w. 1989 , , 336 , 429 gaensler , b. m. , manchester , r. n. , staveley - smith , l. , tzioumis , a. k. , reynolds , j. e. , & kesteven , m. j. 1997 , , 479 , 845 garnavich , p. , kirshner , r. , & challis , p. , 1997
, iau circ .
6710 garnavich , p. , jha , s. , challis , p. , & kirshner , r. 2000 , iau circ .
7360 garnavich , p. , et al .
2001 , apjl , submitted gorenstein , p. , hughes , j. p. , & tucker , w. h. 1994 , , 420 , l25 hasinger , g. , aschenbach , b. , & trmper , j. 1996 , , 312 , l9 hester , j. j. , danielson , g. e. , & raymond , j. c. 1986 , , 303 , l17 hirata , r. , et al . 1987 , phys .
lett . , 58 , 1490 imamura , j. n. , wolff , m. t. , & durisen , r. h. 1984 , , 276 , 667 innes , d. e. , giddings , j. r. , & falle , a. e. g. 1987 , mnras , 226 , 67 innes , d. e. 1992 , a&a , 256 , 660 jakobsen , p. , et al .
1991 , , 369 , l63 kirshner , r. p. , sonneborn , g. , crenshaw , d. m. , & nassiopoulos , g. e. 1987 , , 320 , 602 lawrence , s. s. , sugerman , b. e. , bouchet , p. , crotts , a. p. s. , uglesich , r. , & heathcote , s. 2000a , , 537 , l123 lawrence , s. s. , sugerman , b. e. , & crotts , a. p.
s. 2000b , iau circ .
7419 leitherer , c. , et al .
2000 , `` stis instrument handbook , '' version 4.0 , ( baltimore : stsci ) , 343 lentz , e. j. , et al .
2001 , , 547 , 406 link , r. , rosenberg , d. l. , & chevalier , r. a. 2001 , in interacting winds from massive stars , a.s.p .
conference series ( san francisco : asp ) , in press ( astro - ph/0107127 ) long , k. s. , & blair , w. p. 1990
, , 358 , l13 lundqvist , p. 1999
, , 511 , 389 lundqvist , p. , & fransson , c. 1991 , , 380 , 575 . 1996 , , 464 , 924 lundqvist , p. , & sonneborn , g. 2001 , in sn 1987a : ten years after , eds .
m. phillips & n. suntzeff , a.s.p .
conference series ( san francisco : asp ) , in press ( astro - ph/9707144 ) luo , d. , & mccray , r. 1991 , , 379 , 659 luo , d. , mccray , r. , & slavin , j. 1994 , , 430 , 264 manchester , r. n. , gaensler , b. , m. , wheaton , v. c. , staveley - smith , l. , tzioumis , a. k. , bizunok , n. s. , kesteven , m. j. , reynolds , j. e. 2001 , proc .
australia , accepted ( astro - ph/0110693 ) maran , s. p. , sonneborn , g. , pun , c. s. j. , lundqvist , p. , iping , r. , & gull , t. r. 2000 , , 545 , 390 martin , c. , & arnett , w. d. 1995 , 447 , 378 masai , k. , & nomoto , k. 1994 , , 424 , 924 michael , e. , mccray , r. , borkowski , k. j. , pun , c. s. j. , & sonneborn , g. 1998a , , 492 , l143 michael , e. , et al .
1998b , , 509 , l117 . 2000 , , 542 , l53 osterbrock , d. e. 1989 , astrophysics of gaseous nebulae and active galactic nuclei .
( mill valley : university science books ) panagia , n. , romaniello , m. , scuderi , s. , & kirshner , r. p. 2000 , , 539 , 197 park , s. , burrows , d. n. , garmire , g. p. , nousek , j. a. , mccray , r. , michael , e. , & zhekov , s. 2001 , , accepted ( astro - ph/0111116 ) plait , p. c. , lundqvist , p. , chevalier , r. a. , & kirshner , r. p. 1995
, , 439 , 730 pun , c. s. j. , et al .
1995 , , 99 , 223 ramsbottom , c.a . ,
berrington , k.a . ,
hibbert , a. , & bell , k.l .
1994 , phys .
scripta , 50 , 246 raymond , j. c. 1979 , , 39 , 1 russell , s. c. , & dopita , m. a. 1992 , , 384 , 508 scuderi , s. , panagia , n. , gilmozzi , r. , challis , p. m. , & kirshner , r. p. , 1996
, , 465 , 956 seaton , m. j. 1979 , , 187 , 73p sgro , a. g. 1975 , , 197 , 621 sonneborn , g. , et al .
1997 , , 477 , 848 sonneborn , g. , et al .
1998 , , 492 , l139 staveley - smith , l. , et al . 1992 ,
nature , 355 , 147 staveley - smith , l. , briggs , d. s. , rowe , a. c. h. , manchester , r. n. , reynolds , j. e. , tzioumis , a. k. , & kesteven , m. j. 1993 , nature , 366 , 136 strickland , r. , & blondin , j. m. 1995 , , 449 , 727 suzuki , t. , shigeyama , t. , & nomoto , k. 1993 , , 274 , 883 tielens , a. g. g. m. , mckee , c. f. , seab , c. g. & hollenbach , d. j. 1994 , , 431 , 321 walborn , n. r. , phillips , m. m. , walker , a. r. , & elias , j. h. 1993 , , 103 , 1240 walder , r. , & folini , d. 1998 , a&a , 330 , l21 walker , a. r. , suntzeff , n. b. 1990 , , 102 , 131 wampler , e. j. , richichi , a. , & baade , d. 1989 , in iau colloq .
120 , structure and dynamics of the interstellar medium , ed .
g. tenorio - tagle , m. modes , & j. melnick ( berlin : springer ) , 190 wang , l. 1991 , , 246 , l69 wang , l. , et al .
1996 , , 466 , 998 welty , d. e. , frisch , p. c. , sonneborn , g. , & york , d. g. 1999 , , 512 , 636 zeldovich , ya .
b. , & raizer , yu .
, physics of shock waves and high - temperature hydrodynamic phenomena ( new york : academic press ) lccclcr g140l ( 1425 ) & @xmath269 & @xmath270 & 1.0 & 1997 september 27 & 3869.3 & 11222 + g140l ( 1425 ) & @xmath269 & @xmath270 & 1.0 & 1999 february 27 & 4387.6 & 14350 + g140l ( 1425 ) & @xmath271 & @xmath270 & 1.0 & 1999 october 7 & 4606.1 & 10478 + & & & & & & + g230l ( 2376 ) & @xmath271 & @xmath272 & 3.0 & 1999 september 17 & 4586.5 & 10125 + & & & & & & + g430l ( 4300 ) & @xmath271 & @xmath273 & 4.0 & 1999 september 3 & 4571.9 & 7583 + & & & & & & + g750l ( 7751 ) & @xmath269 & @xmath274 & 8.0 & 1999 february 21 & 4380.9 & 10500 + g750l ( 7751 ) & @xmath271 & @xmath274 & 8.0 & 1999 september 18 & 4387.4 & 7583 + & & & & & & + g750 m ( 6581 ) & @xmath271 & @xmath61 & 1.0 & 1998 march 7 & 4030.0 & 8056 + g750 m
( 6581 ) & @xmath275 & @xmath61 & 1.0 & 1999 august 30 & 4368.0 & @xmath276 + llrrrrc ] & 3425.8 & & & @xmath277 & & 2.09 + ] & 3466.5 & & & @xmath278 & & 2.08 + ] & 3726.0 + 3728.8 & & & @xmath51 6.0 & & 2.02 + h@xmath279 & 3770.6 & & & @xmath280 & & 2.01 + h@xmath134 & 3797.9 & & & @xmath281 & & 2.00 + h@xmath186 & 3835.4 & & & @xmath282 & & 2.00
+ ] & 3868.7 & & & @xmath283 & & 1.99 + + h@xmath284 & 3888.7 + 3889.1 & & & @xmath285 & & 1.98 + + ] ( 8f ) & 3933.7 + 3931.4 + 3932.7 & & & @xmath286 & & 1.97 + + + ] & 3964.7 + 3968.5 + 3968.7 & & & @xmath287 & & 1.96 + h@xmath288 & 3970.1 & & & @xmath287 & & 1.96 + & 4026.2 & & & @xmath289 & & 1.95 + ] & 4068.6 & & & @xmath290 & & 1.93 + ] & 4076.4 & & & @xmath291 & & 1.93 + h@xmath292 & 4101.7 & & & @xmath293 & & 1.92 + ] & 4244.0 & & & @xmath294 & & 1.88 + h@xmath295 & 4340.5 & & & @xmath296 & & 1.85 + ] ( 21f ) & 4358.4 & & & @xmath297 & & 1.84 + ] & 4363.2 & & & @xmath297 & & 1.84 + & 4387.9 & & & @xmath298 & & 1.84 + ] ( 6f ) & 4416.3 & & & @xmath299 & & 1.82 + ] ( 7f + 6f ) & 4452.1 + 4458.0 & & & @xmath300 & & 1.81 + + ] ( 6f ) & 4471.5 + 4470.3 & & & @xmath301 & & 1.81
+ ] & 4571.1 & & & @xmath302 & & 1.78 + ] ( 4f + 5f ) & 4664.5,4665.7 + 4665.0 & & & @xmath303 & & 1.75 + + ] ( 5f ) & 4685.7 + 4687.6 & & & @xmath304 & & 1.75 + ] ( 20f ) & 4814.6 & & & @xmath305 & & 1.71 + h@xmath306 & 4861.3 & & & @xmath307 & & 1.70 + ] ( 3f + 4f ) & 4889.7 & & & @xmath308 & & 1.70 + ] ( 20f ) & 4905.4 & & & @xmath309 & & 1.69 + & 4921.9 & & & @xmath310 & & 1.69 + ] & 4958.9 & & & @xmath311 & & 1.68 + ] & 5006.8 & & & @xmath312 & & 1.67 + & 5015.7 & & & @xmath313 & & 1.67 + ] ( 18f + 19f ) & 5158.0 + 5158.8 & & & @xmath314 & & 1.64 + ] & 5197.9 + 5200.3 & & & @xmath51 0.8 & & 1.63 + ] ( 19f + 18f ) & 5261.6 + 5268.9,5273.4 & & & @xmath315 & & 1.62 + ] ( 19f ) & 5333.7 & & & @xmath316 & & 1.60 + ] ( 19f ) & 5376.5 & & @xmath317 & @xmath318 & @xmath319 & 1.60 + + [ ] ( 17f ) & 5411.5 + 5412.6 & & @xmath316 & @xmath320 & @xmath321 & 1.59 + ] ( 34f ) & 5527.3 & & & @xmath322 & @xmath309 & 1.57 + ] & 5577.4 & & @xmath323 & @xmath324 & @xmath325 & 1.57 + ] & 5754.6 & & @xmath326 & & @xmath327 & 1.54 + & 5875.7 & & @xmath328 & & @xmath329 & 1.53 + ] & 6300.3 & @xmath330 & @xmath331 & @xmath332 & @xmath333 & 1.48 + ] & 6312.1 & & & & @xmath310 & 1.48 + ] & 6363.8 & @xmath334 & @xmath335 & @xmath336 & @xmath337 & 1.47 + ] & 6548.1 & @xmath338 & & @xmath339 & & 1.45 + h@xmath340 & 6562.8 & @xmath341 & & @xmath342 & & 1.45 + ] & 6583.5 & @xmath343 & & @xmath344 & & 1.45 + & 6678.2 & @xmath345 & @xmath346 & @xmath347 & @xmath348 & 1.44 + ] & 6716.5 & @xmath349 & & @xmath350 & & 1.44 + ] & 6730.8 & @xmath351 & & @xmath352 & & 1.44 + & 7065.2 & & @xmath353 & & @xmath354 & 1.40 + ] & 7135.9 & & @xmath355 & & @xmath356 & 1.40 + ] ( 14f ) & 7155.2 & & @xmath357 & & @xmath358 & 1.39 + ] ( 14f ) & 7172.0 & & @xmath359 & & @xmath298 & 1.39 + ] & 7291.5 & & @xmath360 & & @xmath361 & 1.38 + ] + [ ] & 7323.9 + 7320,7330 & & @xmath362 & & @xmath363 & 1.38 + ] ( 2f ) & 7377.9 & & @xmath364 & & @xmath365 & 1.37 + ] ( 14f ) & 7388.2 & & @xmath366 & & @xmath365 & 1.37 + ] ( 2f ) & 7411.6 & & @xmath367 & & * weak * & 1.37 + ] ( 14f ) & 7452.6 & & @xmath368 & & @xmath369 & 1.36 + ] ( 13f ) & 8617.0 & & @xmath370 & & @xmath371 & 1.27 + ] & 9068.6 & & @xmath372 & & @xmath373 & 1.24 + ] & 9530.6 & & @xmath374 & & @xmath375 & 1.22 + lcccc ] @xmath526300 & 154 @xmath376 7 & -13 @xmath376 18 & 172 @xmath376 5 & -34 @xmath376 5 + ] @xmath526363 & 179 @xmath376 14 & -27 @xmath376 20 & 182 @xmath376 11 & + ] @xmath526548 & 165 @xmath376 11 & -37 @xmath376 24 & 230 @xmath376 16 & -11 @xmath376 4 + h@xmath340 @xmath526563 & 253 @xmath376 4 & -29 @xmath376 3 & 225 @xmath376 3 & -35 @xmath376 2 + ] @xmath526584 & 204 @xmath376 7 & -17 @xmath376 10 & 215 @xmath376 7 & -19 @xmath376 2 + @xmath526678 & 177 @xmath376 38 & & 183 @xmath376 27 & + ] @xmath526717 & 113 @xmath376 40 & & 84 @xmath376 24 & + ] @xmath526731 & 149 @xmath376 48 & & 134 @xmath376 20 & + llcrcrcrcc & 1238.8 + 1242.8 & & @xmath377&@xmath378&@xmath379&@xmath380 & @xmath381 & @xmath382 & 5.93 + & 1335 multiplet & 2.6@xmath383 & & & & & @xmath384 & & 4.96 + & 1393.8 + 1402.8 & 2.0@xmath385 & @xmath386 & & @xmath387&@xmath388 & @xmath389 & @xmath390 & 4.57 + ] & 1400 multiplet & & @xmath391 & & @xmath392 & & @xmath393 & & 4.53 + ] & 1483.3 & & @xmath394 & & @xmath395&@xmath396&@xmath397 & @xmath398 & 4.16 + ] & 1486.5 & & @xmath399 & & @xmath400&@xmath396 & @xmath401 & @xmath398 & 4.15 + & 1548.2 + 1550.8 & 2.1@xmath402 & @xmath403 & & @xmath404&@xmath405 & @xmath406 & @xmath407 & 3.95 + ] & 1601.5 + 1601.7 & & & & & & @xmath408 & & 3.81 + & 1640.4 & & @xmath409 & & @xmath410&@xmath411 & @xmath412 & @xmath413 & 3.73 + & 1660.8 + 1666.2 & & & & & & @xmath414 & & 3.69 + ] & 1908.7 & & & & & & @xmath415 & & 3.56 + ] & 2139.0 + 2142.8 & & & & & & @xmath416 & & 4.02 + & 2325 multiplet & & & & & & @xmath417 & & 3.43 + & 2334.6 & & & & & & @xmath418 & & 3.39 + ] & 2470.3 & & & & & & @xmath419 & & 2.93 + & 2736.7 & & & & & & @xmath420 & & 2.51 + & 2746.3 & & & & & & @xmath421 & & 2.50 + & 2795.5 + 2802.7 & 6.6@xmath422 & & & & & @xmath423 & & 2.45 + lll [ tab : models ] @xmath424 ( years ) & 0.03 @xmath376 0.01 & + @xmath425 ( years ) & 2.21 @xmath376 0.01 & + @xmath426 & 3.6 @xmath376 0.7 & + @xmath426(1 ) & & 5.9 @xmath376 3.3 + @xmath426(2 ) & & 3.3 @xmath376 1.2 + @xmath426(3 ) & & 2.8 @xmath376 0.9 + he & 1 & 1 + c & 1.85 @xmath376 0.59 & 2.06 @xmath376 0.66 + n & 0.55 @xmath376 0.08 & 0.54 @xmath376 0.08 + o & 0.83 @xmath376 0.16 & 0.75 @xmath376 0.14 + si & 1.15 @xmath376 0.37 & 0.78 @xmath376 0.24 + | we report and interpret _ hst_/space telescope imaging spectrograph ( stis ) long - slit observations of the optical and ultraviolet ( uv ) ( @xmath0 ) emission - line spectra of the rapidly brightening spot 1 on the equatorial ring of sn 1987a between 1997 september and 1999 october ( days 3869 4606 after outburst ) .
the emission is caused by radiative shocks created where the supernova blast wave strikes dense gas protruding inward from the equatorial ring .
we measure and tabulate line identifications , fluxes and , in some cases , line widths and shifts .
we compute flux correction factors to account for substantial interstellar line absorption of several emission lines .
nebular analysis shows that optical emission lines come from a region of cool ( @xmath1 ) and dense ( @xmath2 ) gas in the compressed photoionized layer behind the radiative shock .
the observed line widths indicate that only shocks with shock velocities @xmath3 have become radiative , while line ratios indicate that much of the emission must have come from yet slower ( @xmath4 ) shocks .
such slow shocks can be present only if the protrusion has atomic density @xmath5 , somewhat higher than that of the circumstellar ring .
we are able to fit the uv fluxes with an idealized radiative shock model consisting of two shocks ( @xmath6 and @xmath7 ) .
the observed uv flux increase with time can be explained by the increase in shock surface areas as the blast wave overtakes more of the protrusion
. the observed flux ratios of optical to highly - ionized uv lines are greater by a factor of @xmath8 than predictions from the radiative shock models and we discuss the possible causes
. we also present models for the observed @xmath9 line widths and profiles , which suggests that a chaotic flow exists in the photoionized regions of these shocks .
we discuss what can be learned with future observations of all the spots present on the equatorial ring . |
the planetary nebula ( pn ) ngc6369 possesses a double - shell morphology consisting of a round bright inner shell and a fainter filamentary outer shell or envelope @xcite .
its most remarkable morphological features are two disparate extensions of the inner shell along the east and west directions ; the latter appears as a large , filamentary blister or _ ansae _ , whilst the former is a bifurcated structure . using low - dispersion spectroscopy , @xcite modelled the peculiar morphology of ngc6369 as a tilted collar of gas and two symmetrically located caps of emission .
this interpretation is consistent with the spheroidal shell and faint bipolar extensions proposed by @xcite based on a single long - slit high - dispersion spectrum obtained along the nebular major axis .
these simplified models do not deal with the bifurcated structure to the eastern side of the inner shell , nor they consider the filamentary appearance of the envelope .
these have been proposed to represent fast low ionisation emission regions ( fliers , * ? ? ?
* ) , although evidence for fast expansion velocities is lacking . the h i and he ii zanstra temperatures of ngc6369 have been reported to be @xmath1(h i ) @xmath2(he ii ) @xmath3 70,000 k ( e.g. , * ? ? ?
* ) , comparable to estimates deriving from the energy balance method @xcite .
the similarity of the h i and he ii zanstra temperatures suggests that the nebula is optically thick to h ionizing radiation , and thus that significant amounts of molecular material and dust may be present outside the ionisation bound optical nebula .
this is consistent with the _ iso _ ( infrared space observatory ) indications of emission from polycyclic aromatic hydrocarbons ( pahs ) bands @xcite that would be produced at a photo - dissociation region ( pdr ) located outside of the ionized shell .
there is also evidence of dust emission from ngc6369 .
_ iso _ spectroscopic observations of ngc6369 have showed it to be one of the few evolved pne to possess the 21 and 30 @xmath4 m emission features @xcite that are typically observed in c - rich agb stars and pne ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
the origin of these spectral features is unclear , although it has been suggested that the 21 @xmath4 m emission may arise from an extraordinarily wide range of agents , including hydrogenated fullerenes , nanodiamonds , sic+sio@xmath0 grains , and tic nanoclusters ( see e.g. , * ? ? ?
* and references therein ) .
the carbon dust origin in ngc6369 is consistent with its high c / o ratio , @xmath52 @xcite . in order to investigate the kinematics and three - dimensional structure of the main nebula and envelope of ngc6369 and to confirm the occurrence of an outer envelope of molecular material , we have acquired new narrow - band near - ir h@xmath0 and optical images , and long slit high - dispersion spectroscopic observations that have been examined in conjunction with archival mid - ir _ iso _ spectra and _ spitzer _ images and spectra , as well as optical narrow - band _ hst _ images .
the observations and archival data are presented in
[ sec_obs ] , and the results on the morphology , and mid - ir spectroscopy are described in [ sec_mor ] , and [ mid_ir_sec ] , respectively .
the kinematics and the discussion are presented in
[ sec_dis ] and final conclusions are presented in [ sec_con ] .
narrow - band images in the [ n ii ] , h@xmath6 , and [ o iii ] emission lines were obtained on september 1 , 2008 using alfosc ( andalucia faint object spectrograph and camera ) at the 2.56-m nordic optical telescope ( _ not _ ) in the observatorio de roque de los muchachos ( orm , la palma , spain ) .
the detector was a 2048@xmath72048 eev ccd with plate scale 0@xmath819 arcsec pix@xmath9 , resulting in a field of view ( fov ) of 6@xmath105@xmath76@xmath105 .
the central wavelengths and bandwidths of the filters are @xmath11=6589 and @xmath12=9 for [ n ii ] , @xmath11=6568 and @xmath12=8 for h@xmath6 , and @xmath11=5007 and @xmath12=8 for [ o iii ] .
the spatial resolution achieved during the observations , as determined from the fwhm of stars in the fov , was 0@xmath895 .
either two or three individual frames with integration times of 300s were taken for each filter , leading to total exposure times of 900s for the [ n ii ] image , and 600s for the h@xmath6 and [ o iii ] images .
the data were bias - subtracted and flat - fielded with twilight flats using standard iraf routines .
narrow - band _ hst _ wfpc2 images of ngc6369 in the [ n ii ] @xmath136583 , h@xmath6 , and [ o iii ] @xmath135007 emission lines were retrieved from mast , the multimission archive at the space telescope science institute ( prop .
i d : 9582 , hubble heritage observations of ngc6369 , pi : keith noll ) .
the nebula was registered by the wf3 detector with a pixel scale of 009 .
four individual exposures were obtained using the f658n ( pivot wavelength @xmath14=6590.8 , @xmath12=28.5 ) , f656n ( @xmath14=6563.8 , @xmath12=21.5 ) , and f502n ( @xmath14=5012.4 , @xmath12=26.9 ) filters , resulting in total exposure times of 1,040s for the [ n ii ] @xmath136583 image , and 640s for the h@xmath6 and [ o iii ] @xmath135007 images ( figure [ hst.img ] ) .
narrow - band h@xmath0 , br@xmath15 and continuum @xmath16 near - ir images of ngc6369 were obtained on 2010 june 27 using liris , the long - slit intermediate resolution infrared spectrograph @xcite , at the cassegrain focus of the 4.2-m william herschel telescope ( _ wht _ ) on roque de los muchachos observatory ( orm , la palma , spain ) .
the detector was a 1k@xmath71k hawaii array sensitive in the spectral range from 0.8 to 2.5 @xmath4 m .
the plate scale is 0@xmath825 pix@xmath9 and the fov 4@xmath1027@xmath74@xmath1027 .
the narrow - band filters have @xmath17=2.1218 @xmath4 m and @xmath12=0.032 @xmath4 m for h@xmath18 ( 1@xmath190 )
s(1 ) , @xmath17=2.1658 @xmath4 m and @xmath12=0.032 @xmath4 m for br@xmath15 , and @xmath17=2.270 @xmath4 m and @xmath12=0.034 @xmath4 m for @xmath16 .
ten 60s exposures were taken for each filter , for total effective exposure times of 600s .
the jitter option was used to increase the sampling of the sky : the telescope pointing of each exposure is shifted by a few pixels , thus rastering the nebula to different locations on the detector .
each series of observations on the object was followed by similar series of observations on adjacent blank sky positions .
the reduction of the liris data was carried out using the dedicated software lirisdr ( liris data reduction package ) , a pipeline for the automatic reduction of near - ir data developed within the iraf environment .
the reduction performed by lirisdr includes standard and additional non - standard steps such as bad pixel mapping , cross - talk correction , flat - fielding , sky subtraction , removal of reset anomaly effect , field distortion correction , and final image co - addition . the continuum contribution to the h@xmath0 and br@xmath15 emission
was subsequently removed using the @xmath16 image . in this process , we did not find necessary to scale the images since the h@xmath0 , br@xmath15 , and @xmath16 filters have very similar bandwidths and transmissions .
the spatial resolution on the wht images , as determined from the fwhm of stars in the field of view , is @xmath50@xmath88 .
the narrow - band continuum - subtracted h@xmath0 and br@xmath15 images of ngc6369 are presented in figure [ h2.img ] .
_ spitzer _ infrared array camera ( irac ; * ? ? ?
* ) images of ngc6369 ( figure 3 ) were retrieved from the nasa / ipac infrared science archive ( irsa ) . the images , obtained on 2005 september 22 as part of program 20119 ( the darkest cloud , an irac / mips survey of the pipe nebula , pi : c.j .
lada ) , have been processed as described in the irac data handbook .
the observations employed filters having isophotal wavelengths ( and bandwidths @xmath12 ) of 3.550 @xmath4 m ( @xmath12=0.75 @xmath4 m ) , 4.493 @xmath4 m ( @xmath12=1.9015 @xmath4 m ) , 5.731 @xmath4 m ( @xmath12 = 1.425 @xmath4 m ) , and 7.872 @xmath4 m ( @xmath12 = 2.905 @xmath4 m ) . irac spatial resolution varies between @xmath2017 and @xmath2020 @xcite , and is reasonably similar in all irac bands , although the diffraction halo at 8 @xmath4 m is stronger than at the other bands .
long - slit high dispersion optical spectroscopy of ngc6369 was obtained on july 17 - 19 2008 using the manchester echelle spectrometer ( mes , * ? ? ?
* ) mounted on the 2.1 m ( f/7.5 ) telescope at the observatorio astronmico nacional de san pedro mrtir ( oan - spm , mexico ) .
a 1k@xmath71k site ccd was used as detector with a 2@xmath72 on - chip binning , resulting on a plate scale of 06pix@xmath9 and a dispersion of 0.1 pix@xmath9 .
since mes has no cross dispersion , a @xmath12=90 bandwidth filter was used to isolate the 87@xmath21 order covering the spectral range which includes the h@xmath6 and [ n ii ] @xmath22 lines . at this spectral order ,
the slit width of 150-@xmath4 m ( 2 ) that was set during the observations corresponds to a spectral resolution of @xmath23kms@xmath9 .
five long - slit spectra , labeled from 1 to 5 on figure [ slits.img ] , were obtained to map the kinematics at different regions of the nebula .
the position angles ( pas ) for slits 1 to 5 are @xmath24 , @xmath25 , @xmath26 , @xmath27 , and @xmath28 , respectively .
the exposure time was 1,200s for the spectrum at pa=+80 ( slit # 2 ) and 1,800s for the rest .
the seeing during the observations , as determined from the fwhm of stars in the fov , varied from 12 to 20 .
the spectra were wavelength calibrated with a th - ar arc lamp to an accuracy of @xmath29kms@xmath9 .
_ spitzer _ infrared spectrograph ( irs ; * ? ? ?
* ) mid - ir spectroscopic observations of ngc6369 obtained through the _ spitzer _
program 40115 ( dual dust chemistry in wolf - rayet planetary nebulae , pi : g. fazio ) were retrieved from irsa .
the observations consisted of short ( 5.1 - 8.5 @xmath4 m ) and long ( 7.4 - 14.2 @xmath4 m ) wavelength spectra taken with the _ spitzer _ irs short - low module 2 ( sl2 ) and 1 ( sl1 ) , respectively .
both modules have rectangular apertures @xmath536@xmath757 that , for these particular observations , were oriented along pa=177.18@xmath30 .
the observations were acquired at closely similar positions @xmath6(j2000.0)=17@xmath3129@xmath322037 and @xmath33(j2000.0)=23452455 ( positions a & c ) , and @xmath6(j2000.0)=17@xmath3129@xmath322044 and @xmath33(j2000.0)=23454355 ( positions b & d ) , resulting in overlapping slits covering the nebula along its minor axis .
the spectral resolution of the observations , @xmath13/@xmath12 , varied between 64 and 128 .
sl1 & sl2 spectra of the source are illustrated in figure [ irs ] , where we show the results for overlapping apertures across the centre of the source ( labelled a to d ) .
all positions cover the inner shell , but positions a and c map the north portion of the envelope and positions b and d the south portion .
similarly , we made use of mid - ir _ _ iso _ _ short wave spectrometer ( sws ; * ? ? ?
* ) spectroscopic observations of ngc6369 acquired on 1997 february 14 as part of program 45601901 ( pi : s. grny ) . _ iso _ sws provides spectra in the short ( 2.38 - 12.1 @xmath4 m ) , medium ( 12 - 29 @xmath4 m ) , and long ( 29 - 45.3 @xmath4 m ) wavelength ranges , acquired through apertures with sizes 14@xmath720 , 14@xmath727 , and 20@xmath722 , respectively
. the _ iso _
sws spectra have an exposure time of 1,140s , and were centred on @xmath6(j2000.0)=17@xmath3129@xmath322078 and @xmath33(j2000.0)=234532.1 .
an inspection of the optical and infrared images of ngc6369 ( figures [ hst.img ] , [ h2.img ] and [ irac_maps ] ) confirms the three morphological components previously described in the main nebula : a bright inner shell , two extensions , and an envelope .
faint blobs of diffuse emission are also detected outside the main nebular shells .
next we describe in detail the morphology of the main nebular shell and this newly detected extended emission .
the _ hst _
wf3 h@xmath6 and [ o iii ]
images of the inner shell and envelope of ngc6369 presented in figure [ hst.img ] show a thick ( @xmath555 ) annular shell with inner radius @xmath2083 .
the surface brightness is noticeably uneven , with patches of bright and diminished emission .
several dark knots and lanes , reminiscent of those seen in other pne ( e.g. , ic4406 and ngc6720 , * ? ? ?
* ) are also detectable .
the [ n ii ] image reveals a dramatically different view of the inner shell , with the h@xmath6 and [ o iii ] dark knots and lanes turning into bright [ n ii ] spots and filaments , and an intricate system of [ n ii ] filaments filling the innermost regions of the shell .
radial filaments emanate outwards from most of the [ n ii ] knots , which are thus revealed as `` tadpole '' or `` cometary '' features very similar to those seen , e.g. , in ngc2392 @xcite and ngc7354 @xcite .
interestingly , there is a bright ring of these cometary knots , at the location of the outer rim of the bright h@xmath6 and [ o iii ] shells at a distance of @xmath20124 of the central star , whose tails extend at least up to @xmath520 . image . ]
the ratio maps in fig .
[ hst.img ] reveal that the bright rim has high excitation , with the lowest [ n ii]/h@xmath6 and [ n ii]/[o iii ] values at radial distances 812 , but the regions within this shell , closer to the central star , have anomalously low ionization . at the location of the ring of knots , at @xmath20124 from the central star ,
the `` cometary '' features are highlighted in the [ n ii]/h@xmath6 and [ o iii]/h@xmath6 ratio maps ( fig .
[ hst.img ] ) , showing high ratio values in the former and low ratio values in the latter .
the tails of these knots define an annular region with outer radius @xmath518
. the peculiar extensions of the inner shell of ngc6369 @xcite along pa@xmath2063@xmath30 are best seen in the h@xmath6 and [ n ii ] images .
the western extension is reminiscent of a lobe or a large _ `` ansa '' _ , whilst the eastern one displays a complex morphology best described as a bifurcated structure .
the ratio maps indicate that these features have low excitation , with values of [ n ii]/[o iii ] up to 20 times larger than those in the higher excitation regions of the inner shell , and up to twice those of the ring of knots at the rim of the inner shell .
the envelope is barely detected in [ o iii ] , and only a little brighter in h@xmath6 . the [ n ii ] emission is brighter and reveals it to be composed of an ensemble of knots , blobs , and filaments .
the appearance of these knots and their tails is less defined than those in the inner shell .
overall , the emission is mostly distributed within a fragmentary annular region with inner and outer radii 2534 , as illustrated by the not composite - colour picture of ngc6369 presented in figure [ ext_emission]-_left_. the morphology of the bright inner shell and envelope of ngc6369 in the near - ir br@xmath15 line is similar to that in h@xmath6 ( fig . [
h2.img]-_right_ ) .
the lack of differential extinction between the optical and near - ir lines of h i suggests little dust content in the nebula along the line of sight .
by contrast , the image in the near - ir h@xmath0 line ( fig . [ h2.img]-_left_ ) offers a very different view of the main shell of ngc6369 .
the region interior to the bright inner shell , which showed anomalously low excitation emission , is revealed to be filled with h@xmath0 emission .
this region is then encompassed by two concentric annular regions , the first one where the h@xmath0 emission is very low , and the second one , with radius @xmath513 , where we find bright h@xmath0 emission .
the interior annulus is coincident with regions of high excitation in the [ n ii]/h@xmath6 and [ n ii]/[o iii ] ratio maps , whilst the exterior , h@xmath0-bright annulus is coincident with the rim of the inner shell defined by the heads of the [ n ii ] `` cometary '' knots .
this provides evidence that the [ n ii ] knots and filaments are associated with h@xmath0 emission , i.e. , they include significant amounts of neutral material .
the extensions of the inner shell are not detected in h@xmath0 , but the northwest quadrant of the envelope and some bright knots exhibit h@xmath0 emission .
surprisingly , the radius of the northwest arc detected in h@xmath0 is @xmath525 , i.e. , this h@xmath0 emission is interior to the annulus of [ n ii ] emission in the envelope . as for the h@xmath0 knots , some ( e.g. , k3 and k5 ) have easily identifiable [ n ii ] counterparts , while for some others ( e.g. , k2 and k4 ) the identification is hampered by their spatial coincidence with the [ n ii]-bright eastern and western extensions of the main shell . among the h@xmath0 knots detected in the envelope , the morphology of knots k2 and k3 is quite noticeable , with a bright head and two tails pointing outwards from the central star .
the mid - ir morphology of ngc6369 is illustrated by the `` unsharp '' _ spitzer _ irac colour - composite picture presented in figure [ ext_emission]-_right_. in this picture , the bright optical shell is the brightest feature , whereas the outer envelope and east and west extensions are fainter but still noticeable . despite the reduced spatial resolution , the mid - ir morphology of ngc6369 is consistent with that observed in the _ hst _ ( fig .
[ hst.img ] ) and not ( fig [ ext_emission]-_left _ ) optical images .
the reddening of the eastern and western extension of the main shell and outer envelope in this colour - composite irac picture implies a higher relative contribution of 8.0 @xmath4 m emission at these morphological components .
this is further evidenced by the contours in figure [ irac_maps ] ; the inner ring is intense in all of the irac images , whereas the surface brightness of the envelope becomes relatively brighter at longer wavelengths , in the 5.8 and 8.0 @xmath4 m bands .
the mid - ir surface - brightness profiles extracted from background - corrected irac images through the minor and major axes of the nebula ( figure [ prof.plot ] ) confirm this result .
the shape of the surface brightness profile interior to the bright main shell is similar in all of the channels , with flux increasing with wavelength .
by contrast , the 3.6 and 4.5 @xmath4 m surface brightness profiles fall - off sharply at @xmath525 , whereas the emission at 5.8 and 8.0 @xmath4 m show broad skirts up to @xmath570 .
this becomes in a sharp increase of the 5.8@xmath4m/4.5@xmath4 m and 8.0@xmath4m/4.5@xmath4 m ratios at radial distances greater than 20 ( fig .
[ prof.plot ] ) .
we also note that the 5.8@xmath4m/4.5@xmath4 m and 8.0@xmath4m/4.5@xmath4 m ratios take minimum values at a radial distance @xmath515 , i.e. , the position of the bright [ n ii ] and h@xmath0 ring . the _ hst
_ wf3 images ( fig .
[ hst.img ] ) hint the presence of irregular condensations outside the envelope of ngc6369 , towards the east and close to the edge of the wf3 fov .
the not [ n ii ] image ( figure [ ext_emission]-_left _ ) , which covers a greater fov than the _ hst _ images , and very especially the wht h@xmath0 image ( figure .
[ h2.img]-_left_ ) reveal patches of diffuse emission and blobs , as marked by arrows in the h@xmath0 image , well beyond the main ionised nebula , up to @xmath570 from the central star in [ n ii ] and @xmath580 in h@xmath0 .
these [ n ii ] and h@xmath0 blobs are distributed irregularly , mostly along the east - west direction , with some hints of point - symmetry .
these blobs are also detected in the _ spitzer _ irac 8 @xmath4 m image ( figure [ ext_emission]-_right _ ) , especially those that are well beyond the 8 @xmath4 m bright envelope .
the _ iso _ spectrum of ngc6369 in the 2.44.1 @xmath4 m spectral range shown in figure [ iso ] reveals the prevalence of emission lines , with br@xmath6 being the most intense and relatively weak or negligible emission of the 3.3 @xmath4 m pah band .
there is also evidence for the h@xmath0 1@xmath190 q(3 ) @xmath132.424 @xmath4 m and h@xmath0 1@xmath190 q(7 ) @xmath132.500 @xmath4 m lines , which is consistent with the detection of h@xmath0 emission in our wht liris images and at 17.036 @xmath4 m @xcite .
although the spectrum remains relatively noisy , and some possible transitions are unidentified , we see evidence for pf@xmath33 and pf@xmath15 , as well as a couple of humphrey transitions . the longer wavelength _ iso _ spectra are relatively noisy , precluding reliable analysis in the range @xmath344.110 @xmath4 m , although the enhancements in emission at 6.2 and 7.8 @xmath4 m suggests that pah band features are present .
the _ spitzer _
irs spectra ( figure [ irs ] ) show more convincing evidence for these pah components . the shorter wavelength sl2 spectrum ( fig
. [ irs]-_left _ ) shows evidence for the h@xmath0 0@xmath190 s(7 ) @xmath135.511 @xmath4 m line , but the spectrum is , otherwise , dominated by four pah bands at 6.2 , 6.9 , 7.8 , and 8.6 @xmath4 m , and an underlying continuum which raises towards longer wavelengths and that can be attributable to cool dust emission .
the same situation applies at longer wavelengths as well ( fig .
[ irs]-_right _ ) , where the flux is dominated by pah band emission yet again , with features at 7.8 , 8.6 , 11.3 , 12.7 and 14.2 @xmath4 m contributing a significant fraction of the total emission .
we also see evidence for strong transitions due to the [ ar iii ] @xmath138.991 @xmath4 m and [ s iv ] @xmath1310.51 @xmath4 m lines , and the possible contribution of the hu@xmath6 @xmath1312.372 @xmath4 m line .
one final point is related to the relative intensities of the pah bands themselves .
the 7.8 @xmath4 m feature arises primarily through c - c stretching modes , whilst the 11.3 @xmath4 m band is a consequence of c - h out - of - plane bending modes .
this would normally lead to i(7.8@xmath4m)/i(11.3@xmath4m)@xmath351.3 for neutral pahs , but radiation fields in pne are capable of appreciably ionising the pah molecules , reducing the strength of the 11.3 @xmath4 m feature and leading to i(7.8@xmath4m)/i(11.3@xmath4m)@xmath3512 for fully ionised pahs @xcite .
it is thus puzzling that the i(7.8@xmath4m)/i(11.3@xmath4 m ) ratio appears significantly less than unity ( fig .
[ irs]-_right _ ) at any location of the nebula .
the mid - ir spectroscopy of ngc6369 described above allows us to discuss the nature of the emission detected in the irac bands .
the _ iso _ data show convincingly that the emission in the irac 3.6 and 4.5 @xmath4 m bands is largely dominated by h i lines , with the br@xmath6 @xmath134.052 @xmath4 m line in the 4.5 @xmath4 m irac band being the brightest line .
the contribution to these bands from bremsstrahlung and the 3.3 @xmath4 m pah band is very small or negligible .
the contribution from other ionic lines in the 4.1 - 5.1 @xmath4 m spectral range ( e.g. , [ mg iv ] @xmath134.486 @xmath4 m , [ ar vi ] @xmath134.529 @xmath4 m , and [ k iii ] @xmath134.618 @xmath4 m , * ? ? ?
* ) is also possible , but the _ iso _ spectrum has too low s / n and the _ spitzer _ irs sl2 spectrum missed this spectral range .
meanwhile , the emission in the 5.8 and 8.0 @xmath4 m irac channels is dominated by the 6.2 , 6.9 , 7.8 , and 8.6 @xmath4 m pah bands , the [ ar iii ] @xmath138.991 @xmath4 m emission line , and an underlying continuum from cool dust which raises towards longer wavelengths .
the close resemblance between the images and surface brightness profiles in the 3.6 and 4.5 @xmath4 m irac channels is due to the prominence of ionic and h i transitions in these two bands .
the higher relative emission in the 5.8 and 8 @xmath4 m bands at large radial distances implies that molecular material and cool dust , responsible of pah bands and continuum emission , are important at the eastern and western extensions of the inner shell and at the envelope , whilst ionic lines dominates the emission from the bright inner shell .
the relative increase of 5.8 and 8.0 @xmath4 m , [ n ii ] , and h@xmath0 emission at the interior of the bright inner shell most likely reveal an opening of the inner shell at its polar caps , so that the emission detected at this region comes from regions far from the nebular center , above and below the bright inner shell . finally , we noted in [ ext_sec ] that the blobs and knots of diffuse emission outside the main nebular shells are detected in the optical [ n ii ] , near - ir h@xmath0 , and irac 8 @xmath4 m images . since these blobs present both emission from ionised ( n@xmath36 ) and molecular ( h@xmath0 ) material , and the irac 8 @xmath4 m band of the main nebula is dominated by pah bands and dust continuum , the origin of the emission in the irac 8 @xmath4 m band may be multi - fold .
the [ ar iii ] @xmath138.991 @xmath4 m line is the brightest ionic line in this band , but the ionization potential of ar@xmath37 ( ip=27.6 ev ) is higher than this of n ( ip=14.5 ev ) and thus ar@xmath38 may be not present in these regions . indeed ,
the lack of [ o iii ] emission , whose ip(o@xmath37 ) is 35.1 ev , indicates that only low - excitation species are present in these regions . since no emission from these blobs is detected in the irac 5.8 @xmath4 m image , it seems unlikely that pah bands or dust continuum produce the emission detected in the irac 8 @xmath4 m image , although we concede that the levels of dust continuum and pah emission in the irac 5.8 @xmath4 m may be lower than in the irac 8 @xmath4 m band .
given the previous difficulties for ionic , pah or continuum emission and that near - ir h@xmath0 1@xmath190 s(1 ) emission is detected , we favor the h@xmath0 0@xmath190 s(4 ) @xmath138.0251 @xmath4 m emission line as the responsible for the emission from these blobs in the irac 8 @xmath4 m image .
figure [ pv.img ] displays the individual h@xmath6 and [ n ii ] position - velocity ( pv ) maps of ngc6369 obtained from the five slits positions illustrated in fig . [ slits.img ] .
overall , the h@xmath6 and [ n ii ] lines show similar kinematical features , but these are best seen in [ n ii ] given that its thermal broadening is smaller than that of the h@xmath6 line .
the radial velocity of the source with respect to the local standard of rest was determined to have a value @xmath39kms@xmath9 , based on the mean expansion velocity at the position of the central star .
this result is in excellent agreement with the value of @xmath40kms@xmath9 reported by @xcite .
the line shapes of the echellogram at slit position # 2 , which cuts the nebula along the east and west extensions , is closely similar to that presented by @xcite : the east and west extensions are detected as red- and blue - shifted loops , respectively , and their shape is typical of the emission in bipolar outflows
. the structure of the h@xmath6 line may give the false impression that the inner shell is a closed ellipsoid , but the sharper view offered by the [ n ii ] line reveals that the loops of the east and west extensions emanate from gaps at the tips of the inner shell .
the eastern extension is mapped in more detail by the slits at positions # 3 , # 4 , and # 5 . in the echellograms at positions # 4 and # 5 ,
this extension is detected as a red - shifted loop , as expected for a blister - like structure .
however , the echellogram at position # 2 reveals a waist and two open loops that are indicative of a more complex structure .
previous three - dimensional modeling of ngc6369 @xcite concur that this nebula has a bipolar structure , although the details make them differ significantly .
@xcite obtained spatially - resolved information of the physical conditions within the nebula that , in conjunction with a photo - ionization model , allowed them to synthesize narrow - band images . in order to match the morphology observed in several emission lines , they adopted for ngc6369 a clumpy hourglass structure with a thick equatorial belt and two symmetrically located caps of emission whose orientation is tilted by 30@xmath30 from the symmetry axis of the main nebula . on the other hand , @xcite used a single high - dispersion long - slit spectrum of ngc6369 obtained along its major axis to illustrate the capabilities of _ `` shape '' _ , an interactive 3-d modeling tool . using a hubble - type ( @xmath41 ) expansion law for ngc6369 , as it is typically assumed in pne ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , they concluded that there is no equatorial waist in ngc6369 , but its three - dimensional structure can rather be described by a partial spheroidal shell with opened tips at its major axis from where two fainter blisters protrude as the east end west extensions .
m , together with the ranges corresponding to the 3.6 @xmath4 m ( solid profile ) and 4.5 @xmath4 m ( dashed profile ) _ spitzer _ irac channels . ]
our more detailed spatial coverage of the kinematics of ngc6369 allows us to refine @xcite s model and to investigate the spatio - kinematical structure of the nebular envelope . using the latest version of the code _ shape _ ( version 10 ; * ? ? ?
* ) , we have constructed a spatio - kinematical model for ngc6369 which fits its narrow - band images and pv maps .
the quality of the fit can be judged in figure [ model1 ] that presents a comparison between the synthetic morphology and kinematics from the model and those observed in ngc6369 . the synthetic image of ngc6369 ( right image of the upper panel in fig . [ model1 ] )
is well matched with its _ hst _ h@xmath6 image ( left image ) .
similarly , the kinematics of ngc6369 are fairly reproduced as shown in the lower panel of fig .
[ model1 ] which presents the pv for slits # 2 to # 5 accompanied by the corresponding pv map extracted from the model .
our best fit model , illustrated in figure [ model2 ] , is composed of : + @xmath42 an equatorial barrel with expansion velocity @xmath2055 km s@xmath9 and whose axis of symmetry is inclined by @xmath550 with respect to the line of sight ( rlos ) , + @xmath42 a southern lobe approaching to us with an inclination angle @xmath540 ( rlos ) and polar expansion @xmath2060 km s@xmath9 , and + @xmath42 two northern lobes with inclination angles @xmath5115 and @xmath5127 receding from us with polar expansion velocities @xmath2077 and @xmath2084 km s@xmath9 , respectively .
+ in this way , the physical structure of ngc6369 is reminiscent of other barrel - shaped pne with polar extensions that result in a spindle shape ( e.g. , ngc3918 , * ? ? ?
* ) .
according to this model and for the distance of 1,550 pc proposed by @xcite , the kinematical ages of the barrel and bipolar lobes are @xmath202,000 yrs and @xmath203,200 yrs , respectively .
the kinematical age of the inner shell is thus smaller than that of the bipolar lobes , most likely indicating important dynamical effects that have made the kinematical and real ages to differ .
these dynamical effects may also be revealed by the distorted eastern structure whose complex shape would arise from the interaction of an expanding , regular lobe with extended and irregular material along its direction .
alternatively , this bifurcated structure may arise from intrinsic multipolar collimated outflows as has been claimed in other pne @xcite .
the emission from the outer envelope of ngc6369 is registered by the slits at positions # 1 , # 3 , # 4 , and # 5 , although the echellograms at positions # 3 and # 4 only cover a small fraction of this shell . in the echellograms at positions # 1 and # 5 , the h@xmath6 and [ n ii ] lines are extremely narrow , with @xmath43 and @xmath44 km s@xmath9 , respectively .
we note , however , that the velocity centroid along these lines is not constant : the echellogram at position # 1 shows a weavy pattern ( best seen in the [ n ii ] line ) , whereas at position # 5 displays an arc - like shape .
we measure velocity differences up to 17 km s@xmath9 .
we also note that emission is patchy , especially in the [ n ii ] line .
clearly the envelope is not inert , but the narrow and weavy / arced line shape does not comply with the expectations for an expanding shell , even if it had the low expansion velocity @xcite presumed for the slow agb wind . the knotty and fragmented appearance of the envelope of ngc6369 indicate that the kinematics revealed by the echellograms is governed by the emission from individual knots , rather than by the emission from a continuous shell .
this situation is reminiscent of many other pne where bright low - ionization knots are embedded within a faint envelope around a bright inner shell , e.g. , ngc2392 and ngc7354 @xcite . as for these two pne ,
the relatively low velocity differences measured for the knots of the envelope of ngc6369 would suggest that they lay at different latitudes with respect to the symmetry axis of the barrel , but mostly on a weavy disk - like or flattened structure , as for ngc2392 and ngc7354 . the relative sizes of the regions dominated by ionized material and molecules and dust are illustrated in figure [ prof.han2h2i4i8 ] that displays the normalized surface brightness profiles in the _ hst _ h@xmath6 and [ n ii ] , wht liris h@xmath0 , and _ spitzer _
irac 4.5 and 8.0 @xmath4 m images .
the comparison between these profiles confirms that the surface brightness profiles of the inner shell of ngc6369 of the optical h@xmath6 emission line and mid - ir irac 4.5 and 8.0 @xmath4 m bands are very similar in shape and extent .
the mid - ir and optical profiles start to depart from each other at the edge of the inner shell : the emission from h@xmath6 and [ n ii ] drops considerably in an annular region between 20 and 25 ( i.e. , the inner regions of the envelope ) , while the emission in the irac 4.5 and 8.0 @xmath4 m bands stays higher and the h@xmath0 profile shows increased emission . on the contrary , on the 2535 outer rim of the envelope , the emission in the h@xmath6 and [ n ii ] lines is enhanced , while the emission in the irac 4.5 and 8.0 @xmath4 m bands flattens and then decays smoothly .
there is no h@xmath0 emission in this range of radial distances .
the comparison of the relative shapes and extents of these different surface brightness profiles is relevant to the understanding of the physical structure of ngc6369 .
its inner shell is mostly ionized and thus the emission in the irac 4.5 and 8.0 @xmath4 m bands is dominated by ionic emission .
the shell is enclosed by a series of `` tadpole '' or `` cometary '' knots whose prominent low - excitation [ n ii ] and molecular h@xmath0 emission at their heads result in bright [ n ii ] and h@xmath0 peaks at @xmath2013 .
the inner shell is surrounded by a region dominated by molecular material , the inner rim of the envelope between @xmath517 and @xmath525 .
this region , outside the bright inner shell and beyond the annular region defined by the tails of the `` cometary '' knots , seems to be a real pdr .
the relative brightening of the emission in the irac 4.5 and 8.0 @xmath4 m bands at these radial distances is very likely due to molecular h@xmath0 emission , although a contribution from continuum dust or pah emissions is also very likely as noted in other pne ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) . finally , the outer rim of the envelope is again dominated by ionic emission , but we note that low - excitation emission dominates .
this may explain the flattening of the emission in the irac 4.5 and 8.0 @xmath4 m emission due to an additional contribution of ionic emission .
it is finally worth noting that sources with comparable he ii and h i zanstra temperatures are likely to be ionisation bound and enveloped by a shell of neutral material .
this leads to a correlation between the zanstra ratio t@xmath45(he ii)/t@xmath45(h i ) , and mid - ir extent outside of the ionised shell ( see e.g. , * ? ? ?
ngc6369 seems to be a peculiar case of this correlation : whereas the emission from molecular material causes a larger dimension of its inner shell at mid - ir wavelenghts , the presence of ionized material in its singular envelope results in very similar final optical and mid - ir dimensions . in
[ neb_morp_sec ] we noted the large jump in the [ n ii ] to [ o iii ] ratio at the filamentary structures that enclose the eastern and western extensions of the inner shell of ngc6369 .
given that these structures are at the tips of bubble - like structures that may arise from bipolar outflows , it is very likely that the unusually high [ n ii]/[o iii ] ratios stem from shock interactions of these structures with surrounding material . to investigate the structure of these features in more detail ,
we show in figure [ fig.shocks ] a radial profile across a particularly bright filament of the eastern extension extracted from the _ hst _ h@xmath6 , [ n ii ] , and [ o iii ] images and ratio maps .
the comparison of the surface brightness profiles indicates that the peak of all these three emission lines are almost coincident , but the spatial profile of the [ o iii ] line is noticeable broader .
the inner shoulder of this profile , closer to the central star ( negative offsets in fig .
[ fig.shocks ] ) , can be explained as the result of a higher photo - excitation .
the shoulder of the line farther away from the nebula , however , can not be explained in this way and it very likely reveals the presence of a forward shock propagating onto a lower density medium .
the [ o iii]/h@xmath6 ratio profile in the lower - panel of fig . [ fig.shocks ] allows us to derive a quantitative value for the spatial extent of this post - shock region , @xmath46 . at the distance of 1,550
pc , this corresponds to @xmath53@xmath710@xmath47 cm , comparable to the post - shock cooling widths determined from modelling of shocks ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
there is a growing number of pne in the literature that show knots or blobs of emission well separated from their main nebular shells that do not form a halo . in some cases ,
the kinematics of these knots imply high - velocity collimated outflows , e.g. , fleming 1 @xcite or mycn18 @xcite , while in some others , e.g. , ic4634 @xcite , the velocity of the knots is close to the systemic velocity , may be due to motions close to the plane of the sky . as for ngc6369 , we note that one of the blobs located outside its inner shell and envelope is detected in the h@xmath6 and [ n ii ] emission lines at the slit position # 1 ( fig .
[ slits.img ] ) .
this blob , at an offset of @xmath570from ngc6369 central star , has a relatively small velocity shift , @xmath5 + 17 km s@xmath9 , with respect to the systemic velocity , and its velocity width is small , @xmath48 in h@xmath6 and @xmath49 km s@xmath9 in [ n ii ] .
the kinematical properties of this outer condensation make ngc6369 closer to the case of the lower velocity knots of ic4634 than to a case of high - velocity collimated outflows .
meanwhile , the velocity shift , although small , would preclude that these outer condensations are part of an incomplete halo ( as , e.g. , the case of ic4593 , * ? ? ?
the origin of these external condensations is uncertain , but they may probe earlier , point - symmetric ejections , or arise from shocks associated with the east - west that excite material surrounding the nebula . the spatio - kinematical model of the inner shell of ngc6369 , the kinematics of its envelope , and the intrincate distributions of ionized , molecular , and mid - ir emission can be used to obtain a coherent view of its structure . the inner shell and extensions of ngc6369
can be described as a tilted equatorial barrel with bipolar extensions .
this inner shell is encircled by a ring of [ n ii]-bright `` cometary '' knots .
the envelope is formed by an ensemble of discrete low - ionization and molecular blobs and knots probably embedded within a much fainter , fully ionized shell .
these knots seem to be distributed at slightly different latitudes with respect to the barrel symmetry axis , but mostly on a weavy disk - like or flattened structure .
near- and mid - ir h@xmath0 emission encompasses the inner shell , revealing a real pdr that seems to be inside the envelope , a shell of ionized gas .
it is unclear whether this pdr exists within the envelope , or whether this is a projection effect , so that the pdr is located at high nebular latitudes , above and below the envelope that would be a rather flat structure only present at low latitudes .
( black ) , [ n ii ] ( blue ) , wht h@xmath0 ( cyan ) , irac 4.5 @xmath4 m ( green ) , and irac 8.0 @xmath4 m ( red ) along pa 330 through the central star of ngc6369 .
the profiles have been normalized to the emission peak at the bright inner shell .
the location of the inner shell end envelope edges are marked by vertical dotted lines . ]
using narrow - band optical and near - ir images , broad - band mid - ir images , optical long - slit echelle spectra and mid - ir spectra , we have investigated the physical structure of the planetary nebula ngc6369 . the inner shell and its east - west extensions
can be modeled in terms of a tilted barrel of ionized gas and bipolar outflows at its tips , in broad agreement with previous observations of the source .
the nebula is optically thick to h ionizing radiation , thus allowing the existence of a pdr that is confirmed by its h@xmath0 emission and suggested by an additional contribution of dust continuum and pah band emission to the irac images in the 5.8 and 8 @xmath4 m bands .
the envelope does not seem to be a real shell enclosing the inner shell , but rather a flattened structure at the equatorial regions of ngc6369 .
molecular and low - excitation emission of [ n ii ] is also detected in small irregular condensations well outside the main nebular shells .
finally , we note that _ hst _ images show the presence of `` cometary '' knots at the rim of the inner shell whose appearance is similar to those of features in ngc2392 and ngc7354 , among other pne . ,
h@xmath6 , and [ o iii ] surface brightness profiles through a likely shock structure in the eastern extension of the inner shell of ngc6369 .
negative offsets are closer to the central star , while positive offsets corresponds to regions farther away from the main nebula .
( _ bottom _ ) [ n ii]/h@xmath6 and [ o iii]/h@xmath6 ratio profiles . ]
grl acknowledges support from conacyt and promep ( mexico ) .
mag and grl are partially funded by grant aya2008 - 01934 of the spanish ministerio de ciencia e innovacin ( micinn ) .
rv , mag and grl thank support by grant in109509 ( papiit - dgapa - unam ) .
we also would like to thank to the referee r. corradi who made very excellent comments for the improvement of the paper . based on observations made with the observatorio astronmico nacional at the sierra de san pedro mrtir , oan - spm , wich is operated by the instituto de astronoma of the universidad nacional autnoma de m ' exico . based on observations made with the nordic optical telescope , operated on the island of la palma jointly by denmark , finland , iceland , norway , and sweden , in the spanish observatorio del roque de los muchachos of the instituto de astrofsica de canarias .
this research has made use of the nasa/ ipac infrared science archive , which is operated by the jet propulsion laboratory , california institute of technology , under contract with the national aeronautics and space administration . | ngc6369 is a double - shell planetary nebula ( pn ) consisting of a bright annular inner shell with faint bipolar extensions and a filamentary envelope .
we have used ground- and space - based narrow - band optical and near - ir images , broad - band mid - ir images , optical long - slit echelle spectra , and mid - ir spectra to investigate its physical structure .
these observations indicate that the inner shell of ngc6369 can be described as a barrel - like structure shape with polar bubble - like protrusions , and reveal evidence for h@xmath0 and strong polycyclic aromatic hydrocarbons ( pahs ) emission from a photo - dissociative region ( pdr ) with molecular inclusions located outside the bright inner shell .
high - resolution _ hst _ narrow - band images reveal an intricate excitation structure of the inner shell and a system of `` cometary '' knots .
the knotty appearance of the envelope , the lack of kinematical evidence for shell expansion and the apparent presence of emission from ionized material outside the pdr makes us suggest that the envelope of ngc6369 is not a real shell , but a flattened structure at its equatorial regions .
we report the discovery of irregular knots and blobs of diffuse emission in low - excitation and molecular line emission that are located up to 80 from the central star , well outside the main nebular shells .
we also show that the filaments associated to the polar protrusions have spatial extents consistent with post - shock cooling regimes , and likely represent regions of interaction of these structures with surrounding material .
[ firstpage ] ( ism : ) planetary nebulae : individual : ngc 6369 ism : jets and outflows infrared : ism ism : lines and bands |
@xcite proposed an unorthodox explanation for luminous blue variable stars ( lbvs / s dor variables ) . in their scenario
all lbvs have gained mass from companion stars , with various consequences .
far more complex than the standard view , this idea was motivated by a statistical allegation that lbvs tend to be spatially `` isolated '' from young objects such as o - type stars . according to smith and tombleson
, each pre - lbv star has an extended lifetime due to the mass that it has gained , and an extra velocity due to the `` kick '' ( i.e. , recoil ) when its companion becomes a supernova .
these two effects are said to explain the alleged isolation .
but we noted later @xcite that four facts together invalidate the isolation argument .
* smith and tombleson s sample of magellanic `` lbvs '' included a substantial fraction of stars that show no relation to the lbv / s dor phenomena , and others that are doubtful . moreover , they employed a statistical test that responds poorly to mixed samples .
* there are good reasons to separate bona fide lbvs into two classes with different evolutionary histories .
this probable fact has been recognized for more than two decades , and separating the classes is not difficult . nevertheless , smith and tombleson mixed them indiscriminately in their statistical sample .
they offered no justification . *
it turns out that the luminous classical lbv s _ are _ statistically associated with the set of late o - type stars that smith and tombleson used in their test .
this result is expected in the standard view but not in their model . * in the traditional view , lower - luminosity lbvs should be too old to be associated with o - type stars ; and this prediction turns out to be correct .
the two classes do indeed differ in the statistical tests , although smith and tombleson assumed that they are equivalent .
these results support the standard view of lbvs , while smith and tombleson s `` kicked mass '' scenario fails the test that was cited as its motivation .
we ( humphreys et al . ) examined only the spatial - distribution question , ignoring other concerns about physical processes .
@xcite has recently objected , at great length , to these conclusions .
the best of his counter - arguments depend on doubtful assumptions , while the others are false : * text and results in @xcite are misquoted or misrepresented . * the hypothesis in question smith and tombleson s scenario is modified or supplemented with additional complications and assumptions . *
test criteria are likewise altered .
smith suggests that the standard view of lbvs not the proposed new model must pass new tests .
the most important of these tests depend on some doubtful implicit assumptions . in effect
, he attempts to transfer the burden of proof away from the smith and tombleson hypotheses .
* some fairly well known observational facts are ignored without explanation . *
smith continues to emphasize a particular type of formal statistical test , which , however , is known to perform poorly if a sample set contains some inappropriate members .
since most of these flaws are evident if one consults @xcite , normally there should be no need for us to respond .
unfortunately , though , smith s lengthy critique may discourage readers from checking the actual content in humphreys et al .
that is why the following account is necessary . to assess smith s objections accurately , _
one must carefully read the paper that he criticizes , _ with a considerable amount of background material in mind .
below `` st15 '' refers to @xcite , `` hwdg '' is @xcite , and `` smith '' is @xcite .
`` hd94 '' signifies @xcite , which remains essential for this discussion despite its age .
( smith often cites his own much later papers for concepts that were explained in hd94 . )
a parameter @xmath0 is the projected distance to the nearest o - type star ; it statistically correlates with stellar age . for other details , see st15 and hdwg and other references therein .
in his 2 , smith identifies `` factual errors and misconceptions '' in hwdg , but each example is either demonstrably false , or trivial , or requires alterations of the st15 model . hence we must review his list . since full explanations
would be extremely tedious here , see the original papers for details .
the first two points are most crucial .
\(1 ) lbvs have higher @xmath1 ratios than other stars in the same parts of the hr diagram .
smith is simply wrong in calling this fact `` an assertion that has no empirical verification '' and `` a conjecture from single - star evolutionary models , not observations , '' see @xcite .
high @xmath1 makes sense in the standard view of lbvs , for reasons outlined by hd94 and @xcite .
\(2 ) near the end of his 2 , smith asserts , as though it were well established , that `` statistically lbvs appear to receive an extra velocity spread ( or longer lifetimes ) beyond that given to the rest of massive stars . '' but _
st15 failed to prove that assertion , _ see hwdg .
then he demands formal tests `` rather than picking a few stars out of a sample . ''
this view has two obvious faults .
first , hwdg did not merely `` pick a few stars . ''
they reviewed all 19 members of st15 s magellanic lbv sample and showed that seven of them definitely should not have been included , while five of the others are unconfirmed `` candidates . ''
hwdg further reviewed the spatial locations of all known lbvs in m31 and m33 .
secondly , the ks statistical test that he advocates is _ quite vulnerable to mixed samples and false sample members . _
it produces erroneous results if either extreme of the distribution includes a separate class of objects .
that is why it failed badly for st15 s sample in the magellanic clouds , see 4.2 of hwdg .
\(3 ) `` @xcite originally defined lbvs ... ''
conti originated the acronym lbv but he did not define the class ; see @xcite ; and references therein .
criteria to make lbvs a coherent species developed in the 1980s , e.g. @xcite .
they led to the definition expressed in hd94 .
\(4 ) smith asserts that hwdg wrongfully omitted a star in the triple system hd 5980 from a list of classical lbvs .
in fact , as clearly stated , that star s outburst did not match a normal lbv eruption .
it should be considered a giant eruption candidate , possibly due to non - lbv causes .
suppose we speculate , however , that the hd 5980 event was a real lbv eruption truncated by the presence of companion stars .
then , instead of having the largest @xmath0 in a particular set as smith states , it would arguably have the smallest ! because one of its companions is an o - type star .
that would strengthen hwdg s conclusions , not weaken them .
even smith s choice of @xmath0 for hd 5980 is smaller than those of the lower - luminosity lbvs .
one can debate the reasoning , but each additional counter - argument tends to reveal more unstated assumptions in the st15 analysis .
\(5 ) he next affirms at length that mwc 112 is different from sk @xmath2 142a ; but hwdg do not say otherwise .
\(6 ) according to smith , `` another misconception expressed by hwdg '' concerns lbv velocities .
first he criticizes hwdg s omission of a quantitative velocity discussion , though st15 offered none either .
hwdg merely noted that confirmed lbvs show no sign of the enhanced velocities that one expects in the scenario that st15 originally described . in order to transfer a significant fraction of material
, their assumed type of binary has to be close , with individual orbital speeds of the order of 200 km s@xmath3 . allowing for complications
, one might reasonably expect runaway speeds well above 50 km s@xmath3 in at least a few cases contrary to observations .
but smith now explains how the velocities can always be less .
after lengthy preliminaries , he suggests that the required supernova occurs after its star has been reduced to only about 4 @xmath4 , so the other ( future lbv ) star s orbital speed is only 20 km s@xmath3 or so .
maybe ; but st15 did not say this , @xmath5 seems unusually small for a core - collapse supernova , and no clear reason is given .
smith says `` a key point is that the relatively low velocity acts together with a longer lifetime ... to make the stars appear more isolated . ''
but hwdg showed that the lower - luminosity lbvs probably have sufficient lifetimes anyway , while the classical lbvs are no more isolated than one would expect .
\(7 ) smith notes that the st15 model allows lbvs in binary or multiple systems , a possibility not clearly stated in their original paper .
( their stated topic was the alleged `` isolation '' of lbvs . ) since most of his comments about binaries apply as well to the standard view of lbvs , they provide no support for st15 .
\(8 ) in his @xmath63 , he attempts to show that `` candidate lbvs '' should be included in the statistical sample .
his test , however , is analogous to proving that grapes can be allowed in a sample of cherries because both classes have similar size distributions relative to oranges . by quoting p - values , dispersions , and other formal parameters
, one can make this fruity conclusion seem plausible unless we examine the objects . in the lbv case , a particular set of `` candidate '' stars with circumstellar ejecta have moderately advanced ages like some lbvs , hence smith deems them to be lbvs for statistical purposes .
this is a more relaxed attitude than statisticians usually prefer .
much worse , about 35% of st15 s sample members are almost certainly not lbvs ( see hwdg ) but smith still includes them in his figure 1 without explaining why .
this is unacceptable statistical procedure .
as mentioned above , the original st15 sample of `` lbvs '' included a mixed set of objects .
hwdg explained why the mixture led to false results . but smith s 4 is titled `` cherry - picking the sample , '' with an implication that subdividing it would amount to fudging the statistics .
we note , however , that this activity is traditionally called stellar classification , not cherry - picking .
the stars physical differences are real .
\(1 ) smith protests the identification of classical lbvs in the magellanic clouds . referring to hwdg s set lbv1 of three classical lbvs
, he says : `` one may question the validity of selecting the tail end of a distribution ... , '' and , in his next paragraph , a warning `` if one extracts the tail end ... ''
he means that these stars have the smallest values of the distribution parameter @xmath0 . but this criticism makes no sense , because @xmath0 played no part in selecting the set ! see below . in the st15 scenario
there was no reason why the three brightest lbvs should have the smallest @xmath0 s .
but in the standard view it s obvious : classical lbvs are younger than the less - luminous lbvs .
what he calls `` the tail end '' is a physically distinct set , recognized long ago without any reference to spatial distributions or @xmath0 .
\(2 ) he alleges an `` unquantified selection bias '' in set lbv1 . but _ those three stars were identified more than two decades ago as the classical lbvs in the magellanic clouds , _ see hd94 .
no more have been found since . above . even with the larger @xmath0 value that smith prefers , adding this object to the sample would slightly strengthen hwdg s conclusions in a formalistic sense .
but there is a substantial likelihood that it is not really an lbv . ]
every member of the less - luminous set lbv2 is definitely below the luminosity criterion .
hence the membership of set lbv1 is well established , and it is hard to see what smith means by `` selection bias '' for such a tiny set with no borderline members .
\(3 ) his next objection appears to arise from a fundamental misunderstanding of one important result .
he asserts that hwdg `` found that [ set lbv1 has ] a similar distribution ... to late o - type dwarfs . ''
he claims that this similarity reveals a contradiction in the standard model , because classical lbvs should not have the same distribution as o8o9 dwarfs which are much less massive .
in fact , however , his argument is based entirely on the term `` similar distribution '' which hwdg did not use .
the largest @xmath0 in set lbv1 is only 12 pc , but the alleged discrepancy refers to values larger than that ! according to st15 s figure 4 , the o8o9 distribution differs seriously from other o stars only for @xmath7 pc .
the following details may be clearer if one views hwdg s figure 5 , which is copied here , see next page .
first , a set of only three objects obviously can not be `` similar '' to a broad distribution in the sense that smith implies .
hwdg s actual result was that _
one can not prove that the three members of set lbv1 have a distribution different from the o - type stars .
_ this was a much weaker conclusion , but it was important because it would not be true in the st15 model .
hdwg did not dwell on the formal meaning of their term `` statistically indistinguishable , '' because it seemed obvious in the figure .
no one viewing that plot would imagine that set lbv1 gives any information for @xmath7 pc .
all three stars in lbv1 have @xmath0 s in the range 412 pc .
these values fall near the distribution medians for subtypes o8o9 and also o6o7
i.e. , for _ the entire range of o stars that can be 3 to 4 myr old .
_ in the standard model , most classical lbvs should have roughly that age ( see item 5 below ) .
\(4 ) smith continues by saying that the lowest - mass rsg s should now be deleted from st15 s original rsg comparison sample ; evidently cherry - picking is sometimes good . in principle
he is right about the sample , but he does not present a revised distribution . he _ conjectures _ that it would alter hwdg s results .
the low - mass rsgs were under - represented in simbad anyway , so the original st15 comparison set may be more suitable than smith supposes .
( a survey for lower - luminosity rsgs appeared after st15 : @xcite . ) instead of arguing about rsg comparison samples , we can simply review the four stars in hwdg s less - luminous set lbv2 .
two of them , r71 and r110 , are located more than 200 pc from their nearest known o - type stars .
does this fact imply greatly extended lifetimes ?
probably not , since the expected ages can very likely account for large @xmath0 s without mass - gaining or other exotic effects .
conventionally these two lbvs should be more than 5 myr old , making them older than almost all o - type stars ( see @xcite , @xcite , hwdg s figure 1 , and the 5-myr isochrone in fig .
12 of @xcite ) . ,
but they evolve beyond o - type in half that time .
see , e.g. , the isochrones in fig .
12 of @xcite . ]
therefore , we can reasonably surmise that all the massive stars which formed along with r71 and r110 have had enough time to evolve beyond o9 . in that case
there is no strong reason to expect nearby o stars .
moreover , this assessment can be enhanced with rotation , extra mass loss , etc . , and
in principle both of these lbvs might be extreme examples . in order to argue that some less - luminous lbvs @xmath0 values are too large for the standard model
, one would need to concretely disprove ( not merely doubt ) this possible explanation and its variants .
contrary assumptions and ks tests with altered rsg sets would not suffice .
the chief point , again , is that hwdg s analysis established that low - luminosity lbvs are older than classical lbvs , irrespective of rsg comparison details .
none of smith s objections contradicts this result .
\(5 ) next he asserts that the standard view requires classical lbvs to be associated with o2o5 stars , not merely o stars in general .
this is a serious misrepresentation , because there is no strong reason to expect stars less than 1.5 myr old about half of the o2o5 sample to associate with lbvs .
( see next paragraph below . ) in smith s figure 2 , the o2o5 distribution greatly differs from o6o9 because 40% of its members have separation parameters less than 3 pc .
such small values indicate average ages less than 1 myr .
the entire o2o5 sample includes stellar ages throughout the range 0 to 2.5 myr , while classical lbvs are thought to be roughly 3.5 myr old .
may be less than 3 myr old , but no such lbv is currently known in the magellanic clouds .
] many o6o9 stars have about the same ages as classical lbvs , but nearly all o2o5 stars are younger than lbvs .
half of them are much younger .
therefore smith s comparison of lbvs vs. o2o5 stars is unjustified .
consider the star formation timescales .
since stars with @xmath8 are seen in fairly compact groups ( @xcite and dozens of authors since ) , very likely they form in localized bursts . presumably these require special circumstances that are not needed for most stars below @xmath9 .
( this is a long - known empirical statement , not theoretical . )
suppose that each burst persists for about 1 myr , a reasonable possibility which can even explain the o2o5 distribution in smith s figure 2 .
in that case , practically all of the new stars evolve beyond o5 before any of them reaches its lbv stage .
more subtle considerations may also apply , because very - massive - star formation is so poorly understood .
hence the situation is complicated but the main point is simple : smith s figure 2 relates more to star formation history than to lbv physics .
the kolmogorov - smirnov or preferably anderson - darling test would be more appropriate for questions that are defined better .
smith also mentions wnh stars , but in his figure 2 they appear more closely related to the early - o stars , than to lbvs . but
there is no reason to suppose that quantitative ages can safely be deduced from that figure .
smith s 5 concerns observed velocities , already mentioned in item 6 of [ sec : misconcepts ] above .
as noted there , st15 s scenario can be modified so that recoil speeds are too low to be identifiable . in his figure 3
he recounts more ks tests , now referring to velocities in m31 and m33 .
they require some unstated assumptions e.g. , that the measurement errors are equal for lbvs and rsgs and they produce no significant results apart from the obvious lack of high - speed runaways .
he adjusts a model of recoil speeds to optimize its ks p - value , and notes that it differs from the zero - extra - velocity case ; but that is merely an indirect way of saying that there are statistical errors .
( abbreviated explanation : the model parameter is non - negative , @xmath10 . if its true value is @xmath11 , the average best - fit value will probably not be zero , since that is the limit of the allowed range . ) no statistically significant result is explicitly claimed in smith s long discussion .
if the st15 scenario is not required to have large recoil speeds ( [ sec : misconcepts ] above ) , then the observational data are obviously not precise enough to give an answer one way or the other .
for the magellanic clouds , smith informally hints that a velocity range of `` 10 s of km s@xmath3 ... appears at first glance to be consistent with ... the mass gainer scenario in st15 . '' but this remark could be applied equally well to various other classes of objects .
one misconception should be noted here to forestall it from propagating in the literature .
smith says that r71 `` has a radial velocity that is offset by @xmath12 km s@xmath3 relative to the systemic velocity of the lmc . ''
but that is irrelevant , because the star s velocity agrees well with the rotation curve _ at its particular location .
_ hwdg carefully and correctly stated that the lmc lbv velocities are consistent `` with their position based on the h i rotation curve available at ned . ''
this debate has some value as an illustration of traditional standards for scientific reasoning .
it features two competing scenarios or sets of hypotheses : * the conventional view of lbvs has been familiar since about 1990 .
it has not strictly been proven yet , but it is conceptually straightforward , it is consistent with existing data , it suggests a credible reason for the lbv high mass loss events , and no one has identified any major errors .
the mass - loss events are said to be caused by one or more of the instabilities that arise when @xmath1 exceeds about half the the eddington limit ( hd94 , @xcite ) .
the chief defects are that no one has yet identified exactly which instability dominates , and no one has yet calculated a realistic lbv outburst . *
@xcite propose a different scenario motivated by statistical arguments .
it is indisputably more complicated , it has multiple free parameters , and it employs a number of assumptions that the authors did not clearly acknowledge . despite the extra complexity , it does not account for some crucial facts that the standard model does explain .
it offers no physical rationale for the lbv instability . in these circumstances , according to normal standards the larger burden of proof rests on smith and tombleson .
they appear to argue that their model is true because statistical tests disprove the standard view .
this is illogical even if one discards the standard model , because other unorthodox models are also available .
for example , we can extend the lifetime of an lbv progenitor simply by assuming that it has rapid rotation .
@xcite showed that the standard view _
does not _ fail the statistical tests that st15 proposed .
nevertheless , @xcite repeatedly attempts to shift the burden of proof onto the long - standing conventional model ( [ sec : cherryp ] above ) .
this case also highlights the dangers of uncritical formalism and sampling . as we noted several times above
, the kolmogorov - smirnov test can produce false results if the sample set is a mixture of disparate classes .
hwdg showed that st15 s set of magellanic `` lbvs '' contained at least three different types of stars and so , as a result , the p - values in their table 2 are meaningless . the reasons can easily be understood by reading hwdg s 4.2 and then comparing hwdg figures 4 and 5 . again
this case involves a fundamental principle : if there is a cogent objection to a member of a statistical sample , then it should be removed .
@xcite in effect denies this , emphasizing statistical formalism instead of careful sampling .
he repeatedly invokes the ks test with flawed samples
e.g. , st15 s original `` magellanic lbv '' and rsg lists , and the o2o5 stars in point 5 of [ sec : cherryp ] above . * confirmed lbvs lie along a locus in the hr diagram , often called the lbv instability strip .
* they have abnormally high @xmath1 ratios .
* they occasionally produce high mass loss episodes with a particular set of characteristics .
indeed these are the
_ main requirements _ for a viable theory .
the standard view accounts for them as described in hd94 and @xcite .
smith and tombleson , however , did not give any reasons why the above facts should be true .
their scenario is sufficiently flexible that we can adapt it to fit , adopting the same instabilities as the standard model .
then the hypothesis becomes essentially this : a star gains mass from its companion , _ and then it evolves in the same way as the conventional view of lbvs . _ but why is the mass exchange necessary ?
if lbv progenitors need extended lifetimes , rapid rotation is arguably a simpler alternative .
anyway , as explained above and in hwdg , there is no clear reason to postulate extended lifetimes .
we suspect that the mass - gainer hypothesis will eventually refer to `` maybe some lbvs '' instead of `` certainly all of them . ''
the two lbv classes merit new work , because they have never been explored in much detail . strictly speaking
, the two - classes concept rests on just two empirical statements : ( 1 ) stars with @xmath13 can not evolve all the way across the hr diagram but those below @xmath14 can ; and ( 2 ) the less - luminous lbvs have surprisingly small masses . these two items together imply that lower - luminosity lbvs are substantially older than the classical lbvs , see hwdg s figure 1 .
perhaps all lbvs form a continuous mass - dependent set , and in every case the instability arises at the time when @xmath1 first exceeds some critical value . for the less luminous stars ,
that time occurs quite late in the evolution . but these details have little effect on hwdg s clearly distinct sets lbv1 and lbv2 , which invalidated st15 s statistical assertions .
there is nothing wrong with hoping that binary interactions play some interesting role for lbvs , but it is wrong to _ assume _ that they do . in order to replace the long - standing model with something very different
, one needs to satisfy the classic hume - laplace - truzzi - sagan rule : extraordinary claims require extraordinary proof .
bohannon , b. 1989 , in _ physics of luminous blue variables , _ iau colloq .
113 , p. 35
chen , y. , bressan , a. , girardi , l. et al .
2015 , , 452 , 1068 conti , p.s .
1984 , in _ observational tests of the stellar evolution theory , _ iau symp .
105 , p. 233
davidson , k. , moffat , a.f.j . , &
lamers , h.j.g.l.m .
1989 , _ physics of luminous blue variables , _ iau colloq . 113
davidson , k. 2016 , arxiv:1602.03925 ekstrom , s. , et al .
2012 , , 537 , a146 gonzlez - fernndez , c. et al .
2015 , , 578 , a3 hubble , e. , & sandage , a. 1953 , , 118 , 353 humphreys , r.m . , &
davidson , k. 1979 , , 232 , 409 humphreys , r.m . , & davidson , k. 1984 , science , 223 , 243 humphreys , r.m .
1989 , in _ physics of luminous blue variables , _ iau colloq .
humphreys , r.m . , & davidson , k. 1994 , , 106 , 1025 ( `` hd94 '' ) humphreys , r.m . , weis , k. , davidson , k. , & gordon , m.s .
2016 , , 825:64 ( `` hwdg '' ) kudritzki , r.p . , et al 1989 , in _ physics of luminous blue variables _ , iau colloq . 113 , p. 67
lamers , h.j.g.l.m .
1986 , in _ luminous stars and associations in galaxies _
de loore et al . ) , p. 157 lamers , h.j.g.l.m .
, & de loore , c.w.h .
1987 , _ instabilities in luminous early type stars _
martins , f. , schaerer , d. , & hillier , d.j .
2005 , , 436 , 1049 pauldrach , a.w.w . , &
puls , j. 1990 , , 237 , 409 smith , n. 2016 , , 461 , 3353 smith , n. , & tombleson , r. 2015 , , 447 , 598 ( `` st15 '' ) stahl , o. , et al 1990 , , 228 , 379 sterken , c. , et al 1991 , , 247 , 383 vink , j.s .
2012 , in _ eta carinae and the supernova impostors , _ assl 384 , p. 221
vink , j.s .
, & de koter , a. 2002 , , 393 , 543 wolf , b. 1989 , , 217 , 87 | @xcite asserted that statistical tests disprove the standard view of lbvs , and proposed a far more complex scenario to replace it . but @xcite showed that smith and tombleson s magellanic `` lbv '' sample was a mixture of physically different classes of stars , and genuine lbvs are in fact statistically consistent with the standard view .
@xcite recently objected at great length to this result . here
we note that he misrepresented some of the arguments , altered the test criteria , ignored some long - recognized observational facts , and employed inadequate statistical procedures .
this case illustrates the dangers of uncareful statistical sampling , as well as the need to be wary of unstated assumptions . |
the theory of bounded subnormal operators was initiated by halmos ( cf .
@xcite ) . the definition and the first characterization of their unbounded counterparts were given independently by bishop ( cf.@xcite ) and foia ( cf .
the foundations of the theory of unbounded ( i.e. , not necessarily bounded ) subnormal operators were developed by the fourth - named author and szafraniec ( cf .
the study of this topic turned out to be highly successful .
it led to a number of challenging problems and nontrivial results in various branches of mathematics including functional analysis and mathematical physics ( see , e.g. , @xcite for the case of bounded operators and @xcite for unbounded ones ) .
this area of interest still plays a vital role in operator theory .
the first characterization of bounded subnormal operators was given by halmos himself .
it was successively simplified by bram ( cf .
@xcite ) , embry ( cf .
@xcite ) and lambert ( see @xcite ; see also ( * ? ? ?
* theorem 7 ) where the assumption of injectivity was removed ) .
the lambert characterization states that a bounded hilbert space operator is subnormal if and only if it generates stieltjes moment sequences ( see section [ n&t ] for definitions ) .
it turns out that this characterization also works for unbounded operators which have sufficiently many analytic vectors ( cf .
* theorem 7 ) ) .
however , it is no longer true for arbitrary unbounded operators ( cf .
* section 3.2 ) ) . recall that subnormal operators with dense set of @xmath2-vectors always generate stieltjes moment sequences ( see ( * ? ? ?
* proposition 3.2.1 ) ) .
it is also worth pointing out that subnormal composition operators in @xmath0-spaces , as opposed to abstract subnormal operators , are always injective ( see ( * ? ? ?
* corollary 6.3 ) ) .
hence , there arises the question whether or not composition operators in @xmath0-spaces generating stieltjes moment sequences are injective ( see problem [ open2 ] ) . in a recent paper @xcite , we have developed a completely new , even in the bounded case , approach to studying subnormality of composition operators ( in @xmath0-spaces over @xmath3-finite measure spaces ) which involves measurable families of probability measures satisfying the so - called consistency condition .
this approach provides a criterion ( read : sufficient condition ) for subnormality of composition operators , which does not refer to the density of domains of powers .
the corresponding technique for weighted shifts on directed trees worked out in @xcite ( see also @xcite ) enabled us to construct an unexpected example of a subnormal composition operator whose square has trivial domain ( cf .
@xcite ) . as shown in (
* example 4.2.1 ) , there are unbounded injective operators generating stieltjes moment sequences which are not even hyponormal , and thus not subnormal .
in fact , it was proved there that if @xmath4 is a leafless directed tree which has exactly one branching vertex and if the branching vertex itself has _
infinite _ valency is understood as the number of outgoing edges at @xmath5 .
] , then there exists a non - hyponormal ( injective ) weighted shift on @xmath4 with nonzero weights generating stieltjes moment sequences . up to isomorphism
, there is only one rootless directed tree of this kind , denoted in @xcite by @xmath6 .
a weighted shift on @xmath6 with nonzero weights is unitarily equivalent to an injective composition operator in an @xmath0-space over a discrete measure space ( see ( * ? ? ?
* lemma 4.3.1 ) and ( * ? ? ?
* theorem 3.2.1 ) ) .
since the directed graph induced by the symbol of such a composition operator coincides with @xmath6 ( see section [ sect - gr ] for the definition ) , it is _ not locally finite_. this raises the question as to whether there exists a non - hyponormal injective composition operator over is _ over _ a directed graph @xmath7 if @xmath7 is induced by the symbol of @xmath8 . ]
a _ locally finite _ connected directed graph generating stieltjes moment sequences , and , if this is the case , how _ simple _ such a directed graph can be , where simplicity is understood with respect to local valency ( cf .
remark [ val1 ] ) .
the present paper addresses both of these questions . taking into account
the simplicity leads to considering directed graphs induced by self - maps whose vertices , all but one , say @xmath9 , have valency one , and the valency of @xmath9 is greater than or equal to @xmath10 .
such directed graphs are described in theorem [ golfclub ] and remark [ val1 ] . in view of part 1 ) of remark [ val1 ] , the situation in which the valency of @xmath9 is equal to one is excluded by an unbounded variant of herrero s characterization of subnormal injective bilateral weighted shifts ( see ( * ? ? ? * theorem 5 ) ; see also ( * ? ? ? * theorem 3.2 ) for a recent approach ) .
if the valency of @xmath9 is strictly greater than one , then , by theorem [ golfclub ] , we have two cases .
the first , which is described in theorem [ golfclub ] , reduces to the directed tree @xmath6 , the case studied in @xcite .
unfortunately , the method invented in @xcite does not give any hope of answering our questions .
in the second case , which is described in theorem [ golfclub ] , the directed graph under consideration has exactly one circuit of length @xmath11 starting at @xmath9 and @xmath12 branches of infinite length attached to @xmath9 , where @xmath13 and @xmath14 ( see figure @xmath15 with @xmath16 ) ; denote it by @xmath17 .
the culminating result of the present paper , theorem [ czwarte ] , shows that there exists a non - hyponormal injective composition operator over the locally finite directed graph @xmath1 generating stieltjes moment sequences .
this answers our first question in the affirmative .
regarding simplicity , the only simpler directed graph which potentially may admit a composition operator with the above - mentioned properties is @xmath18 ( the subnormality over @xmath18 was studied in ( * ? ? ?
* section 3.4 ) ) .
however , so far this particular case remains unsolved because composition operators over @xmath18 obtained by our method are automatically subnormal ( cf .
theorem [ drugie1](iv ) ) .
a large part of the present paper is devoted to the study of subnormality of composition operators over the directed graph @xmath17 .
they all have the same symbol @xmath19 whereas masses attached to vertices that define the underlying @xmath0-space are subject to changes .
since the only known criterion for subnormality of unbounded composition operators relies on the consistency condition ( cf .
* theorem 9 ) ) , we first characterize families of borel probability measures ( on the positive half - line ) indexed by the vertices of @xmath17 which satisfy ( cf . theorem [ cc1 ] ) .
this enables us to model all such families via collections of measures indexed by the set @xmath20 which satisfy some natural conditions ( cf .
procedure [ proceed ] ) , where @xmath21
are ends of edges outgoing from @xmath16 not lying on the circuit ( see figure @xmath15 ) and @xmath22 is the set of all positive integers less than or equal to @xmath12 .
the end @xmath23 of the edge that outgoes from @xmath24 and lies on the circuit also plays an important role in our considerations .
namely , assuming both that the radon - nikodym derivatives @xmath25 ( see section [ sec3.1 ] ) calculated at @xmath23 and @xmath26 , @xmath27 , form stieltjes moment sequences and that appropriate sequences coming from @xmath28 are s - determinate , we show that the corresponding composition operator over @xmath17 is subnormal ( cf .
theorem [ suff ] ) .
the case of @xmath29 does not require any determinacy assumption and may be written purely in terms of the hankel matrices @xmath30_{i , j=0}^{\infty}$ ] and @xmath31_{i , j=0}^{\infty}$ ] ( cf.proposition [ eta1 ] ) .
the proofs of theorem [ suff ] and proposition [ eta1 ] rely on constructing families of measures satisfying .
these two results are in the spirit of lambert s characterization of subnormality of bounded composition operators ( cf .
@xcite ) which is no longer true for unbounded operators ( cf .
* theorem 4.3.3 ) and ( * ? ? ?
* section 11 ) ) .
the case of bounded composition operators over @xmath17 , which is also covered by lambert s criterion , follows easily from theorem [ suff ] ( cf .
proposition [ wn0gr ] ) .
the optimality of the assumptions of propositions [ eta1 ] and [ wn0gr ] is illustrated by examples [ final ] and [ prz2 ] .
it follows from ( * ? ? ?
* theorems 9 and 17 ) ( see also theorem [ glowne ] ) that under the assumption that @xmath32 takes finite values for all positive integers @xmath33 , any family of borel probability measures satisfying consists of representing measures of stieltjes moment sequences @xmath34 , where @xmath35 varies over the vertices of @xmath17 . in section [ s4.4 ]
we discuss the question of extending a given family @xmath36 of borel probability measures to a wider one ( indexed by @xmath17 ) satisfying the consistency condition . according to theorem [ main ] ,
such extension exists if and only if for every @xmath37 , @xmath38 is a stieltjes moment sequence represented by a measure @xmath39 satisfying the conditions , and of this theorem . the condition refers to moments of the measures @xmath39 , @xmath40 . the remaining two are of different nature , namely is a system of @xmath41 equations ( the case of @xmath42 is not excluded ) , while is a single inequality . in theorem
[ 6.2 ] we introduce the condition which is a weaker version of .
this turns out to be the key idea that leads to constructing exotic examples .
assuming the s - determinacy of the sequence @xmath43 for any @xmath44 , it is proved in theorem [ 6.2 ] that the conditions and are equivalent ( provided the remaining ones , and are satisfied ) .
however , this is no longer true if the s - determinacy assumption is dropped .
we show this by using procedure [ proceed2 ] that heavily depends on the existence of a pair of n - extremal measures satisfying some constraints ( cf .
lemma [ pierwsze ] ) .
the task of finding such a pair is challenging .
it is realized by transforming via special homotheties the krein and the friedrichs measures ( which are particular instances of n - extremal measures ) . the crucial properties of these transformations are described in lemma [ epslem ] .
the proof of the existence of the gap between and is brought to completion in theorems [ drugie ] ( the case of @xmath45 ) and [ drugie1 ] ( the case of @xmath46 ) . adapting the above technique , we show in theorem [ czwarte ] that for any integer @xmath47 , there exists a non - hyponormal injective composition operator over @xmath48 which generates stieltjes moment sequences . the case of @xmath49 is treated in theorem [ trzecie ] .
the parallel question of determinacy of moment sequences @xmath34 , @xmath50 , is studied in section [ sec5 ] by using the index of h - determinacy introduced by berg and duran in @xcite . as noted above , the proofs of the main results of the present paper ( theorems [ drugie ] , [ drugie1 ] , [ trzecie ] and [ czwarte ] ) essentially depend on subtle properties of n - extremal measures .
the question of determinacy of moment sequences is of considerable importance in our study as well .
therefore , for the sake of completeness , we collect in section [ sec3 ] basic concepts of the classical theory of moments and include some new results in this field . using ( * ? ? ?
* theorem 3.6 ) , we show that a measure which comes from an n - extremal measure by removing an infinite number of its atoms has infinite index of h - determinacy ( cf .
theorem [ b - d ] ) .
the carleman condition , which always guarantees the h - determinacy of stieltjes moment sequences , is investigated in section [ sec2.5 ] .
the process of transforming moment sequences and their representing measures , including n - extremal ones , via homotheties is described in section [ sec2.3 ] ( the particular case of transformations induced by translations has already been studied via different approaches in @xcite ) .
particular attention is paid to transforming the krein and the friedrichs measures ( cf .
theorem [ taso ] ) . as a consequence ,
a new way of parametrizing n - extremal measures of h - indeterminate stieltjes moment sequences is invented ( cf .
theorem [ tinfsup ] ) and an unexpected trichotomy property of n - extremal measures of h - indeterminate hamburger moment sequences is proven ( cf .
theorem [ trichotomy ] ) .
the n - extremal measures used in the proofs of theorems [ drugie ] , [ drugie1 ] and [ trzecie ] are derived from the krein and the friedrichs measures of an s - indeterminate stieltjes moment sequence , first by scaling them and then by transforming them via carefully chosen homotheties ( cf .
lemma [ epslem ] ) .
as for the proof of theorem [ czwarte ] , the above method requires the usage of the krein and the friedrichs measures coming from shifted al - salam - carlitz ( or , alternatively , from a quartic birth and death process , see remark [ piate ] ) .
the existence , determinacy and explicit form of orthogonalizing measures for al - salam - carlitz @xmath51 are discussed in section [ secasc ] .
it is worth mentioning that explicit examples of n - extremal measures such as those used in the present paper are to the best of our knowledge very rare ( see , e.g. , @xcite ) .
the necessary facts concerning composition operators in @xmath0-spaces over discrete measure spaces , including criteria for their hyponormality and subnormality , are recapitulated in section [ sec3.1 ] .
a variety of relations between radon - nikodym derivatives @xmath25 calculated in different vertices of the directed graph @xmath17 are established in section [ sec3rn ] .
denote by @xmath52 , @xmath53 , @xmath54 , @xmath55 , @xmath56 and @xmath57 the sets of complex numbers , real numbers , nonnegative real numbers , integers , nonnegative integers and positive integers , respectively . set @xmath58 and @xmath59 .
given @xmath60 , we write @xmath61 ( clearly @xmath62 ) .
the identity map on a set @xmath63 is denoted by @xmath64 .
we write @xmath65 for the cardinality of a set @xmath63 and @xmath66 for the characteristic function of a subset @xmath67 of @xmath63 .
the symbol `` @xmath68 '' denotes the disjoint union of sets . a mapping from @xmath63 to @xmath63
is called a _ self - map _ of @xmath63 .
the @xmath3-algebra of all borel subsets of a topological space @xmath63 is denoted by @xmath69 .
all measures considered in this paper are positive . since any finite borel measure @xmath70 on @xmath53 is automatically regular ( cf .
* theorem 2.18 ) ) , we can consider its closed support ; we denote it by @xmath71 . given @xmath72 ,
we write @xmath73 for the borel probability measure on @xmath53 such that @xmath74 . in this paper
we will use the notation @xmath75 and @xmath76 in place of @xmath77}$ ] and @xmath78 respectively ( @xmath79 ) .
we also use the convention that @xmath80 .
the ring of all complex polynomials in one real variable @xmath81 ( which in the context of @xmath0-spaces are regarded as equivalence classes ) is denoted by @xmath82 $ ] .
let @xmath83 be an operator in a complex hilbert space @xmath84 ( all operators considered in this paper are linear ) .
denote by @xmath85 the domain of @xmath83 .
if @xmath83 is closable , then the closure of @xmath83 is denoted by @xmath86 .
set @xmath87 with @xmath88 , where @xmath89 is the identity operator on @xmath84 .
we say that @xmath83 is _ positive _ if @xmath90 for all @xmath91 .
@xmath83 is said to be _ normal _ if it is densely defined , @xmath92 and @xmath93 for all @xmath91 ( or , equivalently , if and only if @xmath83 is closed , densely defined and @xmath94 , see ( * ? ? ?
* proposition , page 125 ) ) .
@xmath83 is called _ subnormal _ if @xmath83 is densely defined and there exists a normal operator @xmath95 in a complex hilbert space @xmath96 with @xmath97 ( isometric embedding ) such that @xmath98 and @xmath99 for all @xmath100 .
@xmath83 is said to be _ hyponormal _ if it is densely defined , @xmath101 and @xmath102 for all @xmath100 .
following @xcite , we say that @xmath83 _ generates stieltjes moment sequences _ if @xmath103 is dense in @xmath84 and @xmath104 is a stieltjes moment sequence for every @xmath105 ( see section [ secbc ] below for the definition and basic properties of stieltjes moment sequences ) .
it is known that if @xmath83 is subnormal and @xmath103 is dense in @xmath84 , then @xmath83 generates stieltjes moment sequences ( cf .
* proposition 3.2.1 ) ) .
however , the reverse implication is not true in general ( see @xcite ; see also @xcite ) . in what follows @xmath106 stands for the @xmath107-algebra of all bounded operators in @xmath84
whose domains are equal to @xmath84 .
denote by @xmath108 the set of all borel measures @xmath70 on @xmath53 such that @xmath109 for all @xmath110 . set @xmath111 .
a sequence @xmath112 is said to be a _ hamburger _
_ stieltjes _ ) _ moment sequence _ if there exists @xmath113 ( resp .
@xmath114 ) such that @xmath115 the set of all such measures , called _ h - representing _ ( resp . _
s - representing _ ) measures of @xmath116 , is denoted by @xmath117 ( resp.@xmath118 ) . a hamburger ( resp .
stieltjes ) moment sequence @xmath116 is said to be _ h - determinate _
_ s - determinate _ ) if @xmath119 ( resp.@xmath120 ) ; otherwise , we call it _
h - indeterminate _ ( resp . _ s - indeterminate _ ) .
we say that a measure @xmath121 ( resp .
@xmath122 ) is h - determinate ( resp .
s - determinate ) if the sequence @xmath123 is h - determinate ( resp.s - determinate ) .
similarly , we define h - indeterminacy and s - indeterminacy of measures .
clearly , an s - indeterminate stieltjes moment sequence is h - indeterminate .
it is well - known that a hamburger moment sequence which has a compactly supported h - representing measure is h - determinate ( cf.@xcite ) .
note that h - determinacy and s - determinacy coincide for stieltjes moment sequences having s - representing measures vanishing on @xmath124 ( see ( * ? ? ?
* corollary , p. 481 ) ; see also ( * ? ? ?
* lemma 2.2.5 ) ) .
let @xmath125 be a hamburger moment sequence .
a measure @xmath126 is called an _
n - extremal _ measure of @xmath116 if @xmath116 is h - indeterminate and @xmath82 $ ] is dense in @xmath127 .
we say that @xmath121 is an n - extremal measure if @xmath70 is an n - extremal measure of the hamburger moment sequence @xmath128 .
denote by @xmath129 the set of all n - extremal measures of @xmath116 and put @xmath130 . note that if @xmath125 is an h - indeterminate hamburger moment sequence , then @xmath131 ( cf .
* theorem 4 and remark , p. 96 ) ) .
moreover , we have : [ pron ] if @xmath125 is an h - indeterminate hamburger moment sequence , then @xmath132 , and the closed support of any @xmath133 is countably infinite with no accumulation point in @xmath53 .
now we state the m. riesz characterizations of h - determinacy and n - extremality ( see @xcite or ( * ? ? ?
* theorem , p. 58 ) ) and the berg - thill characterization of s - determinacy ( see ( * ? ? ?
* theorem 3.8 ) or ( * ? ?
* proposition 1.3 ) ) .
[ mrk ] _ ( i ) _ a measure @xmath134 is h - determinate @xmath135resp .
n - extremal@xmath136 if and only if @xmath82 $ ] is dense in @xmath137 @xmath135resp . @xmath82 $ ] is dense in @xmath127 and not dense in @xmath137@xmath136 . _
( ii ) _ a measure @xmath138 is s - determinate if and only if @xmath82 $ ] is dense in both @xmath139 and @xmath140 .
the above enables us to formulate a comparison test for determinacy .
[ krytnd ] let @xmath141 and @xmath70 be borel measures on @xmath53 @xmath135resp .
@xmath54@xmath136 such that @xmath121 @xmath135resp .
@xmath114@xmath136 and @xmath142 for every @xmath143 @xmath135resp .
@xmath144@xmath136 and for some @xmath145
. then @xmath146 @xmath135resp .
moreover , @xmath141 is h - determinate @xmath135resp .
s - determinate@xmath136 whenever @xmath70 is .
we deal only with the case of h - determinacy ; the other case can be treated similarly .
the standard measure - theoretic argument implies that @xmath148 .
since @xmath149 , we deduce from ( * ? ? ?
* theorem 3.13 ) that @xmath150 is a well - defined bounded operator with dense range .
hence , applying lemma [ mrk ] completes the proof .
[ snieg ] suppose that @xmath151 and @xmath152 are hamburger @xmath135resp .
stieltjes@xmath136 moment sequences such that @xmath153 is h - determinate @xmath135resp .
s - determinate@xmath136 .
then both @xmath151 and @xmath152 are h - determinate @xmath135resp .
s - determinate@xmath136 .
[ snieg2 ] it follows from corollary [ snieg ] that if @xmath154 is a stieltjes moment sequence such that @xmath155 is s - determinate for some @xmath44 , then @xmath154 is s - determinate .
this may suggest that if @xmath156 is an s - determinate stieltjes moment sequence , then so is @xmath157 for some @xmath158 .
however , in general , this is not true .
in fact , one can show more ; namely there exists an h - determinate stieltjes moment sequence @xmath154 such that @xmath155 is s - indeterminate for all @xmath44 .
indeed , as noticed by c. berg ( private communication ) , if @xmath70 is an n - extremal measure of an s - indeterminate stieltjes moment sequence such that @xmath159 ( e.g. , the orthogonalizing measure @xmath160 for the al - salam - carlitz @xmath51 with @xmath161 is n - extremal and its closed support is equal to @xmath162 , see section [ secasc ] ) , then the measure @xmath163 is h - determinate and for every @xmath164 , the measure @xmath165 is n - extremal ( see ( * ? ? ?
* theorem 3.6 and lemma 3.7 ) ) and consequently , since @xmath166 , it is s - indeterminate ( see ( * ? ? ?
* corollary , p. 481 ) or
* lemma 2.2.5 ) ) . the following lemma will be used in section [ sect4.3 ] .
[ kontr ] if @xmath167 , then the following conditions are equivalent _ : _ 1 .
@xmath156 is a stieltjes moment sequence which has an s - representing measure vanishing on @xmath168 , 2 .
@xmath169 for all finite sequences @xmath170 of complex numbers , 3 .
@xmath156 is a stieltjes moment sequence and @xmath171 for all finite sequences @xmath170 of complex numbers .
( i)@xmath172(ii ) obvious .
( ii)@xmath173(iii ) apply the stieltjes theorem ( cf .
* theorem 6.2.5 ) ) .
( ii)@xmath172(i ) let @xmath174 \to { \mathbb c}$ ] be a linear functional such that @xmath175 for all @xmath176 .
take @xmath177 $ ] which is nonnegative on @xmath178 . since @xmath179 is nonnegative on @xmath180
, there exist @xmath181 $ ] such that @xmath182 ( cf .
* problem 45 , p. 78 ) ) . hence @xmath183 . applying the riesz - haviland theorem ( cf .
@xcite ) completes the proof .
the question of when @xmath184 is nonempty has the following answer .
[ taso+ ] suppose @xmath116 is an h - indeterminate stieltjes moment sequence
. then @xmath185 . moreover , @xmath116 is s - determinate if and
only if @xmath186 . let @xmath83 be a symmetric operator in a complex hilbert space @xmath84 and @xmath187 be such that @xmath85 is the linear span of @xmath188 , and @xmath189 for all @xmath110 ( cf .
* ( 1.10 ) ) ) . by assumption
, @xmath83 is a positive operator with deficiency indices @xmath190 ( cf .
* theorem 2 and corollary 2.9 ) ) .
hence , the friedrichs extension @xmath191 of @xmath83 differs from @xmath192 ( cf .
* theorem 5.38 ) ) . as a consequence , @xmath193 , where @xmath194 is the spectral measure of @xmath191 ( cf .
this also proves necessity in the `` moreover '' part .
the sufficiency is a direct consequence of ( * ? ? ?
* theorem 4 ) . recall that if @xmath125 is an s - indeterminate stieltjes moment sequence , then @xmath195 and there exist distinct measures @xmath196 ( uniquely determined by ) such that for every @xmath197 , @xmath198 @xmath199 and @xmath200 are called the _
krein _ and the _ friedrichs _ measures of @xmath116 , respectively .
these two particular n - extremal measures come from the krein and the friedrichs extensions of a positive operator attached to @xmath116 .
the reader is referred to @xcite for the case of friedrichs extensions and to @xcite for a complete and up - to - date operator approach to moment problems ( see also ( * ? ? ? * section 2 ) ) . in this section
we investigate transformations acting on real sequences induced by homotheties of @xmath53 .
such transformations are shown to preserve many properties of hamburger and stieltjes moment sequences .
the particular case of transformations induced by translations has been considered in ( * ? ? ?
* section 3 ) and @xcite ( with different approaches ) .
fix @xmath201 and @xmath202 .
let us define the self - map @xmath203 of @xmath53 by @xmath204 note that @xmath203 is a strictly increasing homeomorphism of @xmath53 onto itself such that @xmath205 for all @xmath206 and @xmath207 .
next , we define the linear self - map @xmath208 of @xmath209 by @xmath210 the proof of lemma [ hg ] below , being elementary , is omitted .
[ hg ] the following hold for all @xmath211 and @xmath212 _ : _
@xmath213 in view of and lemma [ hg ] , the correspondence @xmath214 defines a faithful representation of the group of all strictly increasing homotheties of @xmath53 .
moreover , by and , we have , for all @xmath201 and @xmath215 , @xmath216 in lemma [ transh ] and theorem [ taso ] below we state properties of @xmath208 which are relevant for further considerations .
if @xmath70 is a borel measure on @xmath53 and @xmath217 is a homeomorphism of @xmath53 onto itself , then @xmath218 is the borel measure on @xmath53 given by @xmath219 [ transh ] let @xmath201 and @xmath202
. then 1 .
@xmath208 is a self - bijection on the set of all hamburger moment sequences , 2 .
if @xmath116 is a hamburger moment sequence , then the mapping @xmath220 is a well - defined bijection with the inverse given by @xmath221 in particular , @xmath116 is h - determinate if and only if @xmath222 is h - determinate , 3 .
if @xmath116 is an h - indeterminate hamburger moment sequence , then so is @xmath222 and the mapping defined by maps @xmath129 onto @xmath223 , 4 .
if @xmath116 is a nonzero hamburger moment sequence and @xmath224 , then @xmath225 with convention that @xmath226 , 5 .
if @xmath227 , @xmath116 is an h - indeterminate stieltjes moment sequence and @xmath228 , then @xmath222 is an h - indeterminate stieltjes moment sequence and @xmath229 .
( i)&(ii ) if @xmath116 is a hamburger moment sequence and @xmath224 , then , by the measure transport theorem ( cf .
* theorem 1.6.12 ) ) , we have @xmath230 which means that @xmath222 is a hamburger moment sequence and @xmath231 .
the above combined with and completes the proof of ( i ) and ( ii ) .
\(iii ) let @xmath116 be a hamburger moment sequence and @xmath224 .
since @xmath232 we deduce from the measure transport theorem that the mapping @xmath233 is a well - defined linear homeomorphism ( with the inverse @xmath234 ) such that @xmath235 ) = { \mathbb c}[t]$ ] .
similarly , @xmath236 is a unitary isomorphism such that @xmath237 ) = { \mathbb c}[t]$ ] .
this , ( ii ) and lemma [ mrk](i ) yield ( iii ) .
\(iv ) the equality is a direct consequence of the definition of the closed support of a measure ( see also ( * ? ? ?
* lemma 3.2 ) for a more general result ) . clearly , implies .
\(v ) apply ( iii ) and ( iv ) .
[ taso ] let @xmath238 and @xmath202 .
suppose @xmath125 is an s - indeterminate stieltjes moment sequence and @xmath200 is its friedrichs measure .
then @xmath239 is an h - indeterminate hamburger moment sequence and the following holds _ : _ 1 .
if @xmath240 , then @xmath241 is an s - indeterminate stieltjes moment sequence and @xmath242 is the friedrichs measure of @xmath241 , 2 .
if @xmath243 , then @xmath241 is an s - determinate stieltjes moment sequence , 3 .
if @xmath244 , then @xmath241 is not a stieltjes moment sequence , where @xmath245 . by lemma [ transh ] , @xmath241 is an h - indeterminate hamburger moment sequence such that @xmath246 note that @xmath247 indeed , otherwise , by and lemma [ transh ] , @xmath248 and @xmath249 , which contradicts .
\(i ) by and ( * ? ? ?
* corollary , p. 481 ) ( see also ( * ? ? ?
* lemma 2.2.5 ) ) , @xmath241 is an s - indeterminate stieltjes moment sequence .
denote by @xmath141 its friedrichs measure . if @xmath250 , then by and , @xmath251 , which would contradict .
\(ii ) by , @xmath241 is a stieltjes moment sequence and @xmath252 .
if @xmath222 were s - indeterminate and @xmath141 were its friedrichs measure , then by , @xmath253 , which , as in ( i ) , would contradict .
\(iii ) if @xmath241 were a stieltjes moment sequence , then by lemma [ taso+ ] , there would exist @xmath254 , and thus by , @xmath255 , which would contradict .
[ transh+ ] let @xmath256 .
suppose @xmath116 is an s - indeterminate stieltjes moment sequence and @xmath199 and @xmath200 are its krein and friedrichs measures , respectively .
then @xmath241 is an s - indeterminate stieltjes moment sequence , @xmath257 , @xmath242 is the friedrichs measure of @xmath241 and @xmath258 in particular , @xmath259 is neither the krein nor the friedrichs measure of @xmath241 . in view of and lemma [ transh ] , holds and @xmath260 .
this together with and theorem [ taso ] completes the proof . the particular case of theorem [ taso ] with @xmath261 ( without the statement that @xmath262 is the friedrichs measure of @xmath263 ) appeared in ( * ? ? ?
* theorem 3.3 ) with a very brief outline of the proof based on the von neumann theory of selfadjoint extensions of symmetric operators . in turn , the particular case of corollary [ transh+ ] with @xmath261 ( without any statement on the krein measure ) appeared in ( * ? ? ?
* proposition 3.5 ) with another approach based on the nevanlinna parametrization .
it follows from corollary [ transh+ ] that if @xmath264 , then the transformation @xmath208 preserves s - indeterminate stieltjes moment sequences , and the corresponding mapping defined by preserves the friedrichs measures ( but never the krein ones ) .
the situation changes drastically when @xmath265 ( for example , when we consider @xmath266 with @xmath267 ; see ) .
this is because the quantity @xmath268 may happen to be negative ( cf .
theorem [ taso ] ) .
the above - mentioned properties of self - maps @xmath208 enable us to parameterize n - extremal measures of h - indeterminate stieltjes moment sequences in a new way .
[ tinfsup ] suppose @xmath125 is an h - indeterminate stieltjes moment sequence .
set @xmath269 , where @xmath200 is either the friedrichs measure of @xmath116 if @xmath116 is s - indeterminate , or @xmath200 is the unique s - representing measure of @xmath116 otherwise
. then 1 .
if @xmath270 is s - determinate , then @xmath271 and @xmath272 , 2 . if @xmath270 is s - indeterminate , then @xmath273 , 3 . for every @xmath274 $ ]
there exists a unique @xmath275 such that @xmath276 4 . the mapping @xmath277 \ni t \mapsto
\nu_t \in { \mathscr m}_e({\boldsymbol \gamma})$ ] is a bijection , 5 .
@xmath278\}$ ] , 6 .
the closed support of each n - extremal measure of @xmath116 is bounded from below , 7 .
@xmath279=\{t\}$ ] for every @xmath280 $ ] , 8 .
@xmath281 \varsubsetneq \bigcup_{\mu \in { \mathscr m}_e^+({\boldsymbol \gamma } ) } { \mathrm{supp}(\mu)}$ ] and @xmath282 .
assume that @xmath116 is s - determinate .
then by ( * ? ? ?
* corollary , p. 481 ) ( see also ( * ? ? ?
* lemma 2.2.5 ) ) , @xmath283 and thus @xmath271 . in view of lemma [ taso+ ] ,
@xmath284 and the measure @xmath285 satisfies .
now take @xmath286 .
then by lemma [ transh ] , the sequence @xmath287 is h - indeterminate , @xmath288 and @xmath289 . applying ( * ? ? ?
* corollary , p. 481 ) again , we see that @xmath287 is s - indeterminate .
let @xmath290 be the krein measure of @xmath287 . since @xmath291 , we infer from lemma [ transh ] and that @xmath292 and @xmath293 .
assume now that @xmath116 is s - indeterminate .
let @xmath294 .
since @xmath295 we deduce from theorem [ taso](i ) that @xmath287 is s - indeterminate . taking the krein measure @xmath290 of @xmath287 and arguing as in the previous paragraph
, we see that @xmath296 satisfies . if @xmath297 , then @xmath298 does the job . in both cases , s - determinate and s - indeterminate , the uniqueness of @xmath275 satisfying follows from lemma [ pron ] .
altogether this proves ( i ) , ( ii ) and ( iii ) . clearly , by , the mapping @xmath277 \ni t \mapsto
\nu_t \in { \mathscr m}_e({\boldsymbol \gamma})$ ] is injective . to prove its surjectivity ,
take @xmath133 . by lemma [ pron ] and ,
there exists @xmath274 \cap { \mathrm{supp}(\nu)}$ ] .
then @xmath299 and so , by lemma [ pron ] , @xmath300 .
this proves ( iv ) and consequently ( v ) and ( vi ) .
the condition ( vii ) is a direct consequence of ( iii ) and lemma [ pron ] . to prove ( viii ) take any @xmath301 . by ( iii ) and
lemma [ pron ] , @xmath302 is a nonempty ( in fact , a countably infinite ) subset of @xmath54 which is disjoint with @xmath303 and the latter set being unbounded properly contains @xmath281 $ ] .
it follows from ( iv ) and lemma [ pron ] that the sets @xmath302 , @xmath304 , are nonempty and disjoint .
this completes the proof of ( viii ) and the theorem .
we conclude this section by stating the following unexpected trichotomy property of n - extremal measures of h - indeterminate hamburger moment sequences .
[ trichotomy ] suppose that @xmath125 is an h - indeterminate hamburger moment sequence .
then exactly one of the following three conditions holds _ : _ 1 . for every @xmath305 , @xmath306 and @xmath307 , 2 . for every @xmath305 , @xmath308 and @xmath307 , 3 . for every @xmath305 , @xmath306 and @xmath309 .
suppose that ( i ) does not hold .
then there exists @xmath305 such that either @xmath310 or @xmath311 . in the former case
, @xmath312 is an h - indeterminate stieltjes moment sequence ( cf .
lemma [ transh ] ) , and thus applying theorem [ tinfsup ] to @xmath312 and returning to @xmath116 via the self - map @xmath313 we get ( ii ) . considering the h - indeterminate hamburger moment sequence @xmath314 , one can show that the latter case leads to ( iii ) ( as in the proof of lemma [ transh ] , we verify that the mapping @xmath315 is a bijection which maps @xmath129 onto @xmath316 and @xmath317 for every @xmath318 , where @xmath319 is a self - map of @xmath53 given by @xmath320 for @xmath321 ) .
this completes the proof .
we refer the reader to ( * ? ? ? * and 94 ) ( see also ( * ? ? ?
* theorem 21.5.3 ) ) , where all n - extremal measures of the h - indeterminate hamburger moment sequence arising from polynomials with @xmath322 are explicitly calculated .
since their closed supports are bounded neither from below nor from above , we see that none of the conditions ( i ) , ( ii ) and ( iii ) of theorem [ trichotomy ] is redundant .
orthogonal @xmath323-polynomials introduced by al - salam and carlitz in @xcite give rise to examples of s - indeterminate stieltjes moment sequences for which some particular n - extremal measures are explicitly known .
these special measures help us to show that the directed graph @xmath1 admits a non - hyponormal composition operator generating stieltjes moment sequences ( cf .
theorem [ czwarte ] ) .
we begin by recalling the definition of symbol ( called also factorial ) . for @xmath324 , we write @xmath325 \prod_{j=1}^n ( 1-zq^{j-1 } ) & \text { if } n \in { \mathbb n } , \\[.7ex ] \prod_{j=1}^{\infty } ( 1-zq^{j-1 } ) & \text { if } n=\infty , |q| < 1 .
\end{cases } \end{aligned}\ ] ] ( see ( * ? ? ?
* section vii.1 ) for more on infinite products . )
following ( * ? ? ?
* section vi.10 ) , we extend the original definition of @xmath323-polynomials of al - salam and carlitz to cover the case of @xmath326 . given @xmath327 and @xmath328 , we define @xmath51 , the sequence of complex polynomials in one variable @xmath35 , by the recurrence formula @xmath329 the generating function for @xmath51 can be described as follows ( see @xcite and ( * ? ? ?
* section vi.10 ) ) . if @xmath330 , @xmath331 and @xmath332 , then @xmath333 \frac{\big(z\frac{1}{q};\frac{1}{q}\big)_{\infty } \big(z\frac{a}{q};\frac{1}{q}\big)_{\infty } } { \big(z\frac{x}{q } ; \frac{1}{q}\big)_{\infty } } & \text { if } |z| < \frac{|q|}{|x| } , |q| > 1 , \end{cases } \end{gathered}\ ] ] where @xmath334 with the convention that @xmath335 .
the function of one complex variable @xmath336 given by the right - hand side of is called the generating function for @xmath51 .
clearly , it is meromorphic and has simple poles because @xmath337 if and only if @xmath338 for some ( unique ) @xmath110 . by
, @xmath51 satisfies the following general recurrence relation @xmath339 where @xmath340 and @xmath341 are sequences of complex numbers ( @xmath342 can be chosen arbitrarily ) and @xmath343 are polynomials .
suppose that holds .
then @xmath343 is a hamel basis of @xmath344 $ ] and thus there exists a unique linear functional @xmath345 \to { \mathbb c}$ ] such that @xmath346 and @xmath347 for all @xmath348 ( or equivalently if and only if @xmath346 and @xmath349 for all @xmath350 such that @xmath351 ) .
we say that @xmath352 is an _ orthogonalizing measure _ for @xmath343 if @xmath353 for all @xmath354 $ ] . if this is the case , then @xmath355 is a hamburger moment sequence and @xmath356 is its h - representing measure ( clearly @xmath357 ) . by favard s theorem ( cf .
* theorems i.4.4 and ii.3.1 ) ) , the polynomials @xmath343 have an orthogonalizing measure if and only if @xmath358 and @xmath359 for all @xmath110 . applying the above , we obtain the following statement . @xmath360 as in @xcite we concentrate on the case of @xmath361 . then , by , the polynomials @xmath51 have an orthogonalizing measure if and only if @xmath362 . for such @xmath363 and @xmath323
, the question of determinacy of orthogonalizing measures for @xmath51 can be answered completely .
this is done below .
known orthogonalizing measures for @xmath51 are discussed in detail as well .
we refer the reader to figure @xmath10 which illustrates how determinacy depends on the parameters @xmath363 and @xmath323 .
+ determinacy of known orthogonalizing measures for @xmath51 .
al - salam and carlitz showed in @xcite that if @xmath362 and @xmath364 , then the measure @xmath365 is an orthogonalizing measure for @xmath51 ( if @xmath366 , then the right - hand side of either does not make sense or does not define a positive measure ) , and @xmath367:=\int_{{\mathbb r } } t^n \operatorname{d\hspace{-0.25ex}}\beta^{(a;q)}(t ) = \sum_{k=0}^{n } \frac{(q;q)_n q^{k(k - n)}}{(q;q)_k ( q;q)_{n - k } } a^k , \quad n\in { \mathbb z}_+ . \end{aligned}\ ] ] ( that @xmath160 is a probability measure was proved in ( * ? ? ? * theorem 5.1 ) . ) in turn , chihara essentially proved that if @xmath368 , then @xmath369\}_{n=0}^{\infty}$ ] is an s - indeterminate stieltjes moment sequence and @xmath160 is its friedrichs measure ( cf.@xcite ) .
this can be also deduced from lemma [ transh ] , ( * ? ? ?
* proposition 4.5.1 ) , ( * ? ? ?
* corollary , p. 481 ) ( see also ( * ? ? ?
* lemma 2.2.5 ) ) , ( * ? ? ? * , ( b ) ) and by considering the measure @xmath370 which coincides with the measure @xmath371 appearing in ( * ? ? ?
* proposition 4.5.1 ) .
a similar reasoning shows that if @xmath372 , then @xmath373\}_{n=0}^{\infty}$ ] is an s - indeterminate stieltjes moment sequence ( this was first noticed by chihara in @xcite ) and @xmath160 is its n - extremal measure which is neither the krein nor the friedrichs measure .
consider now the measure @xmath374 applying lemmata [ hg ] and [ transh ] , ( * ? ? ?
* proposition 4.5.1 ) and theorem [ taso ] , the latter to the measure @xmath375 which coincides with the measure @xmath376 appearing in ( * ? ? ?
* proposition 4.5.1 ) ( consult also remark [ fri ] ) , we deduce that if @xmath372 , then @xmath377={\sf m}_n[\gamma^{(a;q)}]$ ] for all @xmath110 , @xmath378 is an orthogonalizing measure for @xmath51 and @xmath378 is the friedrichs measure of @xmath373\}_{n=0}^{\infty}$ ] . finally , as shown in @xcite , if @xmath379 or @xmath380 , then the polynomials @xmath51 have an h - determinate orthogonalizing measure .
clearly , in the former case this measure coincides with @xmath160 defined by . to find the orthogonalizing measure for @xmath51 in the latter case , we follow an idea due to ismail , which in fact can be applied to wider set of parameters ( cf.@xcite ) . for this ,
note that if @xmath381 , then the polynomials @xmath382 satisfy the same recurrence relation as the polynomials @xmath383 ( cf . ) , and thus @xmath384 now suppose that @xmath385 and @xmath386 ( recall that @xmath387 ) . then using the measure transport theorem and the fact that @xmath388 is the orthogonalizing measure for @xmath389 , we get @xmath390 since @xmath391 is a probability measure , we deduce that it is an orthogonalizing measure for @xmath51 .
if additionally @xmath392 , then @xmath393 is another orthogonalizing measure for @xmath51 .
now there are three possibilities . if @xmath394 , then the measures @xmath160 and @xmath395 coincide .
if @xmath396 , then @xmath373\}_{n=0}^{\infty}$ ] is an s - indeterminate stieltjes moment sequence , @xmath160 is its friedrichs measure ( because @xmath397 ) and @xmath395 is its n - extremal measure which is neither the krein nor the friedrichs measure ( because of @xmath398 , and lemma [ transh ] ) . in turn , if @xmath399 , then @xmath373\}_{n=0}^{\infty}$ ] is an s - indeterminate stieltjes moment sequence , @xmath391 is its friedrichs measure ( because of @xmath400 and theorem [ taso ] ) which coincides with @xmath378 defined by , and @xmath160 is its n - extremal measure which is neither the krein nor the friedrichs measure ( because @xmath401 ) . finally , we note that if @xmath402 and @xmath380 ( i.e. , @xmath366 ) , then @xmath403 , and thus , by the above considerations , the measure @xmath391 is the unique orthogonalizing measure for @xmath51 .
one more observation is at hand .
namely , using the equality @xmath404 , we get @xmath405 now it is easily seen that for every @xmath406 such that @xmath407 , the right - hand side of the above equality defines the signed measure ( understood as in @xcite ) , call it @xmath408 , which is positive if and only if @xmath409 for some @xmath410 .
moreover , standard calculations show that if @xmath411 for some @xmath412 , then @xmath413 , which together with implies that @xmath414 . in a sense
, this means that all singularities appearing in can be removed .
the h - determinacy of the hamburger moment problem associated with the polynomials @xmath51 for @xmath415 was discussed by ismail in @xcite .
following @xcite , we define @xmath416 , the _ index of h - determinacy _ of an h - determinate measure @xmath141 at a point @xmath417 , by @xmath418 by the index of h - determinacy of an h - determinate stieltjes moment sequence @xmath419 at a point @xmath417 we understand the index of h - determinacy of a unique h - representing measure of @xmath116 at the point @xmath336 .
note that if @xmath141 is an h - determinate measure such that @xmath420 for some @xmath421 , then @xmath422 for all @xmath423 ( cf .
* corollary 3.4(1 ) ) ) .
if this is the case , then we say that @xmath141 ( or @xmath116 ) has _ infinite index of h - determinacy_. note also that the following holds .
@xmath424 indeed , this can be deduced from proposition [ krytnd ] and the inequality @xmath425 the following result can be thought of as a complement to ( * ? ? ? * theorem 3.6 ) .
[ b - d ] let @xmath70 be an n - extremal measure and @xmath426 be an infinite subset of @xmath427 . set @xmath428
@xmath135with @xmath429 if @xmath430@xmath136 .
then @xmath141 is an h - determinate measure such that @xmath431 for all @xmath417 .
fix @xmath348 .
take any subset @xmath432 of @xmath426 such that @xmath433 .
set @xmath434 .
note that for every borel function @xmath435 , @xmath436 by ( * ? ? ?
* theorem 3.6 ) , @xmath437 is h - determinate and @xmath438 for all @xmath439 . using proposition [ krytnd ] and applying the inequality first to @xmath440 and then to @xmath441
, we deduce that @xmath141 is h - determinate and @xmath442 for all @xmath439 .
this completes the proof .
suppose @xmath156 is a stieltjes moment sequence . clearly , the shifted sequence @xmath443 is a stieltjes moment sequence .
it turns out that if @xmath443 is s - determinate , then so is @xmath156 ( see ( * ? ? ?
* proposition 5.12 ) ; see also ( * ? ? ?
* lemma 2.4.1 ) ) .
the reverse implication is not true in general ( cf .
however , it is true if @xmath156 satisfies the carleman condition ( see proposition [ wouk ] below ) .
recall that a sequence @xmath444 satisfies the _ carleman condition _
if @xmath445 with the convention that @xmath335 . below we collect some properties of stieltjes moment sequences that satisfy the carleman condition .
[ wouk ] let @xmath125 be a stieltjes moment sequence
. then 1 .
if @xmath116 satisfies the carleman condition , then @xmath116 is h - determinate , 2 .
@xmath116 satisfies the carleman condition if and only if @xmath443 satisfies the carleman condition , 3 .
if @xmath116 satisfies the carleman condition , then so does for every @xmath446 , 4 .
@xmath116 satisfies the carleman condition if and only if @xmath447 satisfies the carleman condition for every @xmath448 @xmath135equivalently _ : _ for some @xmath448@xmath136 .
\(i ) see ( * ? ? ?
* corollary 4.5 ) .
\(ii ) this can be deduced from the equivalence ( d)@xmath173(e ) in ( * ? ? ? * theorem 19.11 ) ( the equivalence follows from the carleman inequality , cf .
@xcite ) .
\(iii ) let @xmath70 be an s - representing measure of @xmath156 .
set @xmath449 .
since the case of @xmath450 is obvious , we can assume that @xmath451 .
then @xmath452 which implies that @xmath157 satisfies the carleman condition .
\(iv ) this follows from ( * ? ? ?
* section 1 ) because , by the cauchy - schwarz inequality , @xmath453 for all nonnegative integers @xmath454 such that @xmath455 .
note that in view of proposition [ wouk ] , a stieltjes moment sequence which satisfies the carleman condition has infinite index of h - determinacy because its index of h - determinacy at @xmath456 is infinite .
in this paper we study composition operators in @xmath0-spaces over discrete measure spaces . by a _ discrete measure _ on a ( nonempty ) set @xmath63 we understand a @xmath3-finite measure @xmath356 on the @xmath3-algebra @xmath457 such that @xmath458 for every @xmath459 , with the convention @xmath460 note that if @xmath356 is a discrete measure on @xmath63 , then @xmath63 is at most countable and @xmath461 for every @xmath462 .
moreover , any discrete measure @xmath356 on @xmath63 is determined by a function @xmath463 via @xmath464 .
this one - to - one correspondence between discrete measures and positive functions is used frequently in the present paper .
let @xmath356 be a discrete measure on a set @xmath63 and let @xmath465 be a self - map of @xmath63 .
then the operator @xmath466 in @xmath467 given by @xmath468 is called a _ composition operator _ in @xmath467 with a _ symbol _ @xmath465 . since the measure @xmath469 given by @xmath470 for @xmath471 is absolutely continuous with respect to @xmath356
, we can consider the radon - nikodym derivative @xmath472 .
clearly @xmath473 hence , @xmath474 for every @xmath459 if and only if @xmath475 .
it is easily seen that ( cf.@xcite ) @xmath476 note also that @xmath477 applying ( * ? ? ?
* proposition 3.2 ) and the assertions ( ii ) and ( iv ) of ( * ? ? ?
* proposition 4.1 ) , we get a new criterion for the @xmath33th power of @xmath466 to be densely defined .
@xmath478 thus , if @xmath348 and @xmath479 for all @xmath462 , then @xmath480 for all @xmath481 and @xmath462 .
this hereditary property is no longer true for a single point @xmath462 ( see ( * ? ? ?
* example 4.2 ) ) .
in fact , given any nonempty subset @xmath482 of @xmath57 , we can construct a discrete measure @xmath356 on a set @xmath63 and a self - map @xmath465 of @xmath63 such that @xmath483 for some @xmath484 .
this is shown below .
fix @xmath485 .
let @xmath63 , @xmath486 , @xmath487 and @xmath465 be as in theorem [ golfclub2 ] with @xmath49 and @xmath488 for all @xmath489 .
suppose that @xmath482 is a nonempty subset of @xmath490 ( see section [ n&t ] for the definition of @xmath490 ) .
define the discrete measure @xmath356 on @xmath63 by @xmath491 it is a matter of routine to show that @xmath492 assume now that @xmath466 is densely defined ( as before , @xmath356 is a discrete measure on a set @xmath63 and @xmath465 is a self - map of @xmath63 ) , or equivalently , by and , that @xmath493 for all @xmath494 .
then for every function @xmath495 , there exists a unique @xmath496-measurable function @xmath497 such that @xmath498 or equivalently , such that for every @xmath499 , @xmath500 the above definition is correct because @xmath501 .
the function @xmath502 is called the _ conditional expectation _ of a function @xmath503 with respect to the @xmath3-algebra @xmath496 ( cf.@xcite ) . following @xcite
, we say that a family @xmath504 of borel probability measures on @xmath54 satisfies the _ consistency condition _
if @xmath505 in view of and , we see that @xmath504 satisfies the consistency condition if and only if @xmath506 the following characterization of hyponormality of @xmath507 can be deduced from , and ( * ? ?
* corollary 6.7 ) ( see also ( * ? ? ?
* lemma 2.1 ) ) .
[ buda ] let @xmath356 be a discrete measure on a set @xmath63 and @xmath465 be a self - map of @xmath63 .
assume that @xmath466 is densely defined .
then @xmath507 is hyponormal if and only if for every @xmath462 , @xmath474 and @xmath508 a criterion for subnormality of unbounded composition operators given in ( * ? ? ?
* theorems 9 and 17 ) takes in the present situation the following form ( recall that if @xmath466 is subnormal , then @xmath509 for every @xmath462 , or equivalently @xmath466 is injective , cf .
* section 6 ) ) .
[ glowne ] let @xmath356 be a discrete measure on a set @xmath63 and @xmath465 be a self - map of @xmath63 .
assume that @xmath466 is densely defined and @xmath474 for every @xmath462 .
suppose that there exists a family @xmath504 of borel probability measures on @xmath54 that satisfies .
then @xmath466 is subnormal and @xmath510 let @xmath511 be a directed graph ( i.e. , @xmath512 is a nonempty set called the _ vertex set _ of @xmath7 and @xmath194 is a subset of @xmath513 called _ the edge set _ of @xmath7 ) .
we say that a vertex @xmath514 is a _ parent _ of a vertex @xmath5 , and write @xmath515 , if @xmath516 and @xmath517 whenever @xmath518 .
the directed graph @xmath7 is said to be _ connected _ if for every pair @xmath519 of distinct vertices there exists an _
undirected path _ joining @xmath514 and @xmath5 , i.e. , a finite sequence @xmath520 of vertices with @xmath521 such that @xmath522 , @xmath523 and for every @xmath524 , either @xmath525 or @xmath526 .
we say that a finite sequence @xmath527 of distinct vertices is a _ circuit _ if @xmath528 , @xmath529 for all @xmath530 and @xmath531 . by a _ rootless directed tree _
we mean a directed graph @xmath532 which is connected , has no circuits , each vertex of @xmath4 has a parent and @xmath533 for every @xmath534 . in this case , obviously , the partial function @xmath535 is a self - map of @xmath512 . given a rootless directed tree @xmath536 and a family @xmath537 of complex numbers , we define a _ weighted shift _ @xmath538 on @xmath4 with weights @xmath539 via @xmath540 where @xmath541 for @xmath542 .
we refer the reader to @xcite for more information on directed trees and weighted shifts on them .
it follows from ( * ? ? ?
* lemma 4.3.1 ) and ( * ? ? ?
* theorem 3.2.1 ) that each weighted shift @xmath538 on a countable rootless directed tree @xmath536 with nonzero weights is unitarily equivalent to a composition operator @xmath543 in @xmath544 for some discrete measure @xmath356 on @xmath512 ( the assumption that @xmath512 is infinite made in ( * ? ? ?
* lemma 4.3.1 ) is redundant ) . as explicated below , the proof of ( * ? ? ?
* lemma 4.3.1 ) contains more information .
[ jfa - ex ] let @xmath545 be a rootless directed tree .
suppose that @xmath356 is a discrete measure on @xmath512 .
then the composition operator @xmath543 in @xmath467 is unitarily equivalent to a weighted shift on @xmath4 with positive weights .
since @xmath356 is a discrete measure on @xmath512 , we see that @xmath546 .
consider the weighted shift @xmath538 on the directed tree @xmath4 with weights @xmath547 . a careful inspection of the proof of ( * ?
* lemma 4.3.1 ) reveals that the composition operator @xmath543 is unitarily equivalent to @xmath538 via the unitary isomorphism @xmath548 defined in ( * ? ? ?
* ( 4.3.4 ) ) . in this section
we classify connected directed graphs induced by self - maps whose vertices , all but one , have valency one and the valency of the remaining vertex is nonzero ( see theorem [ golfclub ] below ; see also figures @xmath15 and @xmath549 which illustrate this theorem ) .
let @xmath63 be a nonempty set and @xmath465 be a self - map of @xmath63 .
set @xmath550 then @xmath551 is a directed graph which we call the _ directed graph induced by _ @xmath465 .
note that the valency of a vertex @xmath462 is equal to @xmath552 and that @xmath553 is the parent of @xmath35 .
we will write @xmath554 whenever @xmath176 and @xmath555 .
= [ circle , fill = gray!15 ] ( e0)[font= ] at ( 2,-1 ) @xmath23 ; ( e-1)[font= ] at ( 2,3 ) @xmath556 ; ( e-2)[font= ] at ( -1,3)@xmath557 ; ( e1)[font= ] at ( -1,-1 ) @xmath558 ; ( ek)[font= ] at ( 3.5,1 ) @xmath559 ; ( e11)[font= ] at ( 7,3 ) @xmath560 ; ( e12)[font= ] at ( 10.5,3 ) @xmath561 ; ( e13)[font= ] at ( 14,3 ) @xmath562 ; ( e1n ) at ( 16,3 ) ; ( e21)[font= ] at ( 7,1 ) @xmath563 ; ( e22)[font= ] at ( 10.5,1 ) @xmath564 ; ( e23)[font= ] at ( 14,1 ) @xmath565 ; ( e2n)[font= ] at ( 16,1 ) ; ( e 5 1)[font= ] at ( 7,-3 ) ; ( e 5 n)[font= ] at ( 16,-3 ) ; ( e31)[font= ] at ( 7,-1 ) @xmath566 ; ( e32)[font= ] at ( 10.5,-1 ) @xmath567 ; ( e33)[font= ] at ( 14,-1 ) @xmath568 ; ( e3n)[fill = none ] at ( 16,-1 ) ; ( e11)(ek ) node[pos=0.5,above = 0pt , fill = none ] ; ( e21)(ek ) node[pos=0.5 , above = 0pt , sloped , fill = none ] ; ( e 5 1)(ek ) node [ pos=0.5,below=-10pt , sloped , fill = none ] ; ( e31)(ek ) node[pos=0.5 , below=0pt , fill = none ] ; ( e12)(e11 ) node[pos=0.5,above = 0pt , fill = none ] ; ( e22)(e21 ) node[pos=0.5,above = 0pt , fill = none ] ; ( e32)(e31 ) node[pos=0.5,below=0pt , fill = none ] ; ( e13)(e12 ) node[pos=0.5,above = 0pt , fill = none ] ; ( e23)(e22 ) node[pos=0.5,above=0pt , fill = none ] ; ( e33)(e32 ) node[pos=0.5,below = 0pt , fill = none ] ; ( e 5 n)(e 5 1 ) node[pos=0.5,above = -10pt , fill = none ] ; ( ek ) to [ bend right ] ( e-1 ) node[pos=0.5,above = 0pt , fill = none ] ; ( e1n)(e13 ) ; ( e2n)(e23 ) ; ( e3n)(e33 ) ; ( e-2 ) to [ bend right ] ( e1 ) ; ( e0 ) to [ bend right ] ( ek ) ; ( e-1)(e-2 ) node[pos=0.5,above = 0pt , fill = none ] ; ( e1)(e0 ) node[pos=0.5,above = 0pt , fill = none ] ; . the directed graph @xmath551 in the case of ( the self - map + @xmath465 acts in accordance with the reverted arrows ) .
= [ circle , fill = gray!15 ] ( e-1)[font= ] at ( 1,1 ) @xmath569 ; ( e-2)[font= ] at ( -1.5,1)@xmath570 ; ( e1)[font= ] at ( -3.5,1 ) ; ( ek)[font= ] at ( 3.5,1 ) @xmath23 ; ( e11)[font= ] at ( 7,3 ) @xmath560 ; ( e12)[font= ] at ( 10.5,3 ) @xmath561 ; ( e13)[font= ] at ( 14,3 ) @xmath562 ; ( e1n ) at ( 16,3 ) ; ( e21)[font= ] at ( 7,1 ) @xmath563 ; ( e22)[font= ] at ( 10.5,1 ) @xmath564 ; ( e23)[font= ] at ( 14,1 ) @xmath565 ; ( e2n)[font= ] at ( 16,1 ) ; ( e 5 1)[font= ] at ( 7,-3 ) ; ( e 5 n)[font= ] at ( 16,-3 ) ; ( e31)[font= ] at ( 7,-1 ) @xmath566 ; ( e32)[font= ] at ( 10.5,-1 ) @xmath567 ; ( e33)[font= ] at ( 14,-1 ) @xmath568 ; ( e3n)[fill = none ] at ( 16,-1 ) ; ( e11)(ek ) node[pos=0.5,above = 0pt , fill = none ] ; ( e21)(ek ) node[pos=0.5 , above = 0pt , sloped , fill = none ] ; ( e 5 1)(ek ) node [ pos=0.5,below=-10pt , sloped , fill = none ] ; ( e31)(ek ) node[pos=0.5 , below=0pt , fill = none ] ; ( e12)(e11 ) node[pos=0.5,above = 0pt , fill = none ] ; ( e22)(e21 ) node[pos=0.5,above = 0pt , fill = none ] ; ( e32)(e31 ) node[pos=0.5,below=0pt , fill = none ] ; ( e13)(e12 ) node[pos=0.5,above = 0pt , fill = none ] ; ( e23)(e22 ) node[pos=0.5,above=0pt , fill = none ] ; ( e33)(e32 ) node[pos=0.5,below = 0pt , fill = none ] ; ( e 5 n)(e 5 1 ) node[pos=0.5,above = -10pt , fill = none ] ; ( ek ) to [ ] ( e-1 ) ; ( e1n)(e13 ) ; ( e2n)(e23 ) ; ( e3n)(e33 ) ; ( e-2 ) to [ ] ( e1 ) ; ( e-1)(e-2 ) node[pos=0.5,above = 0pt , fill = none ] ; . the directed graph @xmath551 in the case of .
[ golfclub ] let @xmath63 and @xmath465 be as above and let @xmath571 .
then the following two conditions are equivalent _ : _ 1 .
the directed graph @xmath551 is connected and there exists @xmath572 such that @xmath573 and @xmath574 for every @xmath575 , 2 .
one of the following two conditions is satisfied _ : _ 1 .
there exist @xmath576 and two disjoint systems @xmath577 and @xmath578 of distinct points of @xmath63 such that @xmath579 2 .
there exist two disjoint systems @xmath486 and @xmath580 of distinct points of @xmath63 such that @xmath581 clearly , we need only to prove the implication ( i)@xmath172(ii ) .
we do it in five steps .
the proof of step 1 being simple is omitted .
if @xmath572 is such that @xmath582 for all @xmath583 , then the restriction of @xmath465 to @xmath584 is injective .
if @xmath572 and @xmath585 are such that @xmath586 , @xmath587 , @xmath588 , @xmath574 for every @xmath575 and @xmath589 , then 1 . @xmath590 provided @xmath176 and @xmath591 , 2 .
@xmath592 provided @xmath176 and @xmath593 , where @xmath594 is a unique element of @xmath63 such that @xmath595 ( see ( s1 ) ) , 3 .
@xmath596 and @xmath597 provided @xmath350 and @xmath598 , 4 .
@xmath599 are distinct points of @xmath63 provided @xmath600 , 5 . @xmath601 provided @xmath602 and @xmath603 . for this , take @xmath598 .
by assumption , @xmath574 and thus there exists a unique @xmath604 such that @xmath605 . since @xmath587 and @xmath606
, we deduce that @xmath607 . an induction argument yields ( s1 ) , ( s2 ) and
( s3 ) . to prove ( s4 ) and ( s5 ) , suppose that @xmath602 and @xmath608 for some integers @xmath609 . if @xmath610 , then by ( s2 ) and ( s3 ) , we have @xmath611 and so @xmath612 which yields @xmath613 , a contradiction . finally , if @xmath614 , then @xmath615 which completes the proof of ( s4 ) and ( s5 ) .
step 3 . if @xmath551 is connected and @xmath616 is a nonempty subset of @xmath63 such that @xmath617 and @xmath618 , then @xmath619 . indeed , otherwise there exists @xmath620 .
take @xmath621 .
since the graph @xmath551 is connected , there exists a finite sequence @xmath520 of elements of @xmath63 with @xmath521 such that @xmath622 , @xmath623 and for every @xmath624 , either @xmath625 or @xmath626
. then there exists @xmath627 such that @xmath628 and @xmath629 .
thus either @xmath630 or @xmath631 , a contradiction .
step 4 . if @xmath551 is connected and @xmath572 is such that @xmath573 , @xmath574 for every @xmath575 and @xmath632 for some @xmath348 , then holds . for this
, we set @xmath633 first we show that @xmath634 is a sequence of distinct points of @xmath63 . indeed , if @xmath635 for some integers @xmath636 , then @xmath637 and @xmath638 which contradicts .
set @xmath639 with @xmath640 for @xmath641 , and @xmath642 .
then clearly @xmath643 and @xmath644 for @xmath645 , and @xmath646 .
since @xmath647 is injective , we see that @xmath648 .
this and @xmath573 imply that @xmath649 .
thus , by step 2 , the conditions ( s1)-(s5 ) hold .
let @xmath650 be a sequence of distinct points of @xmath67 .
setting @xmath651 for @xmath652 and @xmath653 , we verify that @xmath577 and @xmath578 are disjoint systems of distinct points of @xmath63 which satisfy .
hence , @xmath654 and @xmath655 for every @xmath656 , where @xmath657 since @xmath574 for every @xmath658 and @xmath573 , we deduce that @xmath659 , or equivalently that @xmath660 .
this together with step 3 completes the proof of step 4 .
step 5 . if @xmath551 is connected and @xmath572 is such that @xmath573 , @xmath574 for every @xmath575 and @xmath661 for every @xmath348 , then ( ii - b ) holds .
we begin by proving that @xmath662 for this , suppose that @xmath663 for some integers @xmath664 .
since , by assumption @xmath665 for all @xmath410 , we can assume that @xmath666 .
if @xmath667 , then @xmath668 , which in view of step 1 implies that @xmath669 .
an induction argument shows that @xmath670 .
clearly , the last equality remains valid if @xmath671 .
hence , in both cases , we see that @xmath672 . as @xmath673 and
consequently @xmath674 , we deduce that @xmath675 , a contradiction .
this proves .
set @xmath676 with @xmath677 , and @xmath678 .
clearly @xmath586 , @xmath679 and @xmath680 .
this yields @xmath681 .
therefore , by step 2 , the conditions ( s1)-(s5 ) hold .
let @xmath682 be a sequence of distinct points of @xmath67 .
setting @xmath651 for @xmath683 and @xmath684 , we verify that @xmath486 and @xmath580 are disjoint systems of distinct points of @xmath63 which satisfy .
now , arguing as in the final stage of the proof of step 4 with @xmath685 we complete the proof of step 5 and theorem [ golfclub ] .
now we make some comments on theorem [ golfclub ] that support the idea of studying composition operators with symbols of type .
we also shed more light on the question of simplicity of directed graphs discussed in section [ subsec1 ] .
[ val1 ] 1 ) suppose that the directed graph @xmath551 is connected and @xmath466 is a composition operator in @xmath686 , where @xmath356 is a discrete measure on @xmath63 . by and ( * ? ? ?
* proposition 6.2 ) , @xmath466 is injective if and only if @xmath687 to guarantee injectivity of @xmath466 , we assume that holds . we also exclude the ( more complex ) case when the directed graph @xmath551 has more than one vertex of valency greater than @xmath10 .
the case when @xmath551 has exactly one vertex of valency greater than @xmath10 has been described in theorem [ golfclub ] . if @xmath574 for every @xmath459 ( the _ flat _ case ) , then , by ( * ? ?
* proposition 2.4 ) and step 3 of the proof of theorem [ golfclub ] , @xmath465 is bijectively isomorphic to the mapping @xmath688 acting on @xmath689 for some @xmath690 . hence , the composition operators @xmath466 is unitarily equivalent to an injective bilateral weighted shift ( the case of @xmath55 ) or to a bijective finite dimensional weighted shift ( the case of @xmath691 for @xmath410 ) .
\2 ) now we assume that the directed graph @xmath551 is not connected .
note that if there exists @xmath572 such that @xmath573 for some @xmath571 and @xmath574 for every @xmath575 ( hence holds ) , then @xmath692 , where @xmath616 is given either by or by depending on whether @xmath693 for some @xmath348 or not ( see the proofs of steps 4 and 5 of theorem [ golfclub ] ) .
indeed , this is a consequence of the easily verifiable fact that @xmath694 is a connected subgraph of @xmath551 .
next observe that @xmath695 is a bijective self - map of @xmath696 and thus it is bijectively isomorphic to a disjoint sum of a number of self - maps @xmath697 of @xmath698 , where @xmath699 ( see ( * ? ? ?
* proposition 2.4 ) ) . clearly , the directed graph @xmath694 satisfies the condition ( i ) of theorem [ golfclub ] .
hence , the composition operator @xmath466 is unitarily equivalent to an orthogonal sum of composition operators whose symbols are described above ( cf .
* appendix c ) ) .
one can draw a similar conclusion for symbols @xmath465 discussed in the flat case in 1 ) .
\3 ) the directed graph @xmath551 appearing in ( ii - b ) ( see figure @xmath549 ) is isomorphic to the directed tree @xmath700 defined in ( * ? ? ?
* ( 6.2.10 ) ) .
by lemma [ jfa - ex ] , the corresponding composition operator @xmath466 is unitarily equivalent to a weighted shift on @xmath700 with nonzero weights .
subnormality of such operators has been studied in @xcite .
if we allow the directed graph @xmath701 to have vertices of valency @xmath456 , still preserving the simplicity requirement , we exclude from our considerations the case when @xmath701 has more than one vertex of valency greater than @xmath10 . ] , then the question is how many such vertices can there be .
the answer is given in proposition [ ibj1 ] and in the proof of theorem [ golfclub2 ] ( see ) .
the question becomes especially interesting when the composition operator @xmath466 generates stieltjes moment sequences . in this version ,
the question is a particular case of a more general problem , called here the _ injectivity problem _ ( see problem [ open2 ] ) . in the present section
we investigate the injectivity problem in the context of directed graphs @xmath551 having at most one vertex of valency greater than @xmath10 . the following assumption will be used many times in this section .
@xmath702 if holds , then we set @xmath703 recall that the case of @xmath704 has been discussed in remark [ val1 ] .
we begin by considering the situation where the valency of each vertex of the directed graph @xmath551 does not exceed @xmath10 .
[ ibj1 ] assume that holds . if the directed graph @xmath551 is connected and @xmath705 for every @xmath462 , then 1 .
@xmath706 , 2 .
@xmath466 is injective whenever @xmath466 generates stieltjes moment sequences .
\(i ) for if not , there are two distinct vertices @xmath707 . by the connectivity of @xmath551 , there exists an undirected path @xmath708 joining @xmath514 and @xmath5 of smallest possible length @xmath709 ( with @xmath710 and @xmath711 ) .
it is easily seen that the sequence @xmath520 is injective .
hence , since @xmath712 , we see that @xmath713 , @xmath714 and @xmath715 . by induction , there exists @xmath716 such that @xmath717 and @xmath718 . as a consequence , @xmath719 , which contradicts the inequality @xmath720 .
this proves ( i ) .
\(ii ) suppose , on the contrary , that @xmath466 is not injective , or equivalently that @xmath721 . by ( i ) , @xmath722 for some @xmath572 , and thus , by ( * ? ? ?
* proposition 2.4 ) and steps 1 and 3 of the proof of theorem [ golfclub ] , @xmath465 is bijectively isomorphic to the mapping @xmath723 acting on @xmath56 .
hence , without loss of generality , we can assume that @xmath724 and @xmath725 for all @xmath726 .
then @xmath727 , @xmath728 and @xmath729 .
this contradicts ( * ? ? ?
* lemma 1.1(ii ) ) because @xmath466 generates stieltjes moment sequences .
it remains to consider the case when the directed graph @xmath551 has exactly one vertex of valency greater than @xmath10 . the following theorem which describes such graphs ( see figures @xmath730 and @xmath731 )
will be deduced from theorem [ golfclub ] .
= [ circle , fill = gray!15 ] ( e0)[font= ] at ( 2,-1 ) @xmath23 ; ( e-1)[font= ] at ( 2,3 ) @xmath556 ; ( e-2)[font= ] at ( -1,3)@xmath557 ; ( e1)[font= ] at ( -1,-1 ) @xmath558 ; ( ek)[font= ] at ( 3.5,1 ) @xmath559 ; ( e11)[font= ] at ( 7,3 ) @xmath560 ; ( e12)[font= ] at ( 10.5,3 ) @xmath561 ; ( e21)[font= ] at ( 7,1 ) @xmath563 ; ( e 5 1)[font= ] at ( 7,-3 ) ; ( e 5 n)[font= ] at ( 16,-3 ) ; ( e31)[font= ] at ( 7,-1 ) @xmath566 ; ( e32)[font= ] at ( 10.5,-1 ) @xmath567 ; ( e33)[font= ] at ( 14,-1 ) @xmath568 ; ( e3n)[fill = none ] at ( 16,-1 ) ; ( e11)(ek ) node[pos=0.5,above = 0pt , fill = none ] ; ( e21)(ek ) node[pos=0.5 , above = 0pt , sloped , fill = none ] ; ( e 5 1)(ek ) node [ pos=0.5,below=-10pt , sloped , fill = none ] ; ( e31)(ek ) node[pos=0.5 , below=0pt , fill = none ] ; ( e12)(e11 ) node[pos=0.5,above = 0pt , fill = none ] ; ( e32)(e31 ) node[pos=0.5,below=0pt , fill = none ] ; ( e33)(e32 ) node[pos=0.5,below = 0pt , fill = none ] ; ( e 5 n)(e 5 1 ) node[pos=0.5,above = -10pt , fill = none ] ; ( ek ) to [ bend right ] ( e-1 ) node[pos=0.5,above = 0pt , fill = none ] ; ( e3n)(e33 ) ; ( e-2 ) to [ bend right ] ( e1 ) ; ( e0 ) to [ bend right ] ( ek ) ; ( e-1)(e-2 ) node[pos=0.5,above = 0pt , fill = none ] ; ( e1)(e0 ) node[pos=0.5,above = 0pt , fill = none ] ; . the directed graph @xmath551 in the case of with @xmath732 , + @xmath733 , @xmath734 , @xmath735 .
= [ circle , fill = gray!15 ] ( e-1)[font= ] at ( 1,1 ) @xmath569 ; ( e-2)[font= ] at ( -1.5,1)@xmath570 ; ( e1)[font= ] at ( -3.5,1 ) ; ( ek)[font= ] at ( 3.5,1 ) @xmath23 ; ( e11)[font= ] at ( 7,3 ) @xmath560 ; ( e12)[font= ] at ( 10.5,3 ) @xmath561 ; ( e21)[font= ] at ( 7,1 ) @xmath563 ; ( e22)[font= ] at ( 10.5,1 ) @xmath564 ; ( e23)[font= ] at ( 14,1 ) @xmath565 ; ( e2n)[font= ] at ( 16,1 ) ; ( e 5 1)[font= ] at ( 7,-3 ) ; ( e 5 n)[font= ] at ( 16,-3 ) ; ( e31)[font= ] at ( 7,-1 ) @xmath566 ; ( e11)(ek ) node[pos=0.5,above = 0pt , fill = none ] ; ( e21)(ek ) node[pos=0.5 , above = 0pt , sloped , fill = none ] ; ( e 5 1)(ek ) node [ pos=0.5,below=-10pt , sloped , fill = none ] ; ( e31)(ek ) node[pos=0.5 , below=0pt , fill = none ] ; ( e12)(e11 ) node[pos=0.5,above = 0pt , fill = none ] ; ( e22)(e21 ) node[pos=0.5,above = 0pt , fill = none ] ; ( e23)(e22 ) node[pos=0.5,above=0pt , fill = none ] ; ( e 5 n)(e 5 1 ) node[pos=0.5,above = -10pt , fill = none ] ; ( ek ) to [ ] ( e-1 ) ; ( e2n)(e23 ) ; ( e-2 ) to [ ] ( e1 ) ; ( e-1)(e-2 ) node[pos=0.5,above = 0pt , fill = none ] ; .
the directed graph @xmath551 in the case of with @xmath732 , + @xmath736 , @xmath737 , @xmath735 .
[ golfclub2 ] assume that holds and @xmath738 .
then the following two conditions are equivalent _ : _ 1 .
the directed graph @xmath551 is connected and there exists @xmath572 such that @xmath573 and @xmath705 for every @xmath575 , 2 .
one of the following two conditions is satisfied _ : _ 1 .
there exist @xmath576 , a sequence @xmath739 and two disjoint systems @xmath577 and @xmath740 of distinct points of @xmath63 such that @xmath741 2 .
there exist a sequence @xmath742 and two disjoint systems @xmath486 and @xmath487 of distinct points of @xmath63 such that @xmath743 ( ii)@xmath172(i ) obvious .
( i)@xmath172(ii ) in view of theorem [ golfclub ] , we may assume that @xmath721 .
let @xmath744 be a family of pairwise disjoint countably infinite sets such that @xmath745 .
set @xmath746 . for every @xmath747 ,
let @xmath748 be a sequence of distinct points of @xmath749 such that @xmath750 .
define the self - map @xmath751 of @xmath752 by @xmath753 it is a matter of routine to verify that @xmath754 ( i.e. , @xmath755 extends @xmath465 ) , the directed graph @xmath756 is connected , @xmath757 and @xmath758 for every @xmath759 .
hence , by theorem [ golfclub ] , @xmath760 takes the form or ( with @xmath761 in place of @xmath762 ) .
set @xmath763 if @xmath760 takes the form and @xmath764 otherwise . since @xmath757
, we deduce that @xmath765 for every @xmath766 .
this , the explicit description of @xmath767 and an induction argument combined with @xmath768 imply that there exists a sequence @xmath769 such that ( ii ) holds .
in particular , we have @xmath770 this completes the proof .
now we take a closer look at the directed graphs described by parts and of theorem [ golfclub2 ] which admit composition operators generating stieltjes moment sequences .
[ golfclub3 ] assume that holds . if @xmath551 is as in theorem _ [ golfclub2 ] _ with @xmath571 and @xmath466 generates stieltjes moment sequences , then 1 .
@xmath771 for every @xmath652 , 2 .
@xmath772 , 3 .
@xmath466 is injective whenever the stieltjes moment sequence @xmath773 is s - determinate .
\(i ) indeed , otherwise @xmath774 for some @xmath775 , which implies that @xmath776 , @xmath777 and @xmath778 .
this contradicts ( * ? ? ?
* lemma 1.1(ii ) ) because @xmath466 generates stieltjes moment sequences .
\(ii ) suppose , on the contrary , that @xmath779
. then , by , @xmath780 and consequently @xmath781 and @xmath466 is not injective . in view of lambert s theorem ( we use ( * ? ? ? * theorem 7 ) ) , @xmath466 is subnormal and , as such , is injective ( see ( * ? ? ?
* theorem 9d ) ) .
this gives a contradiction and proves ( ii ) .
\(iii ) by ( * ? ? ?
* theorem 10.4 ) , the sequence @xmath782 is a stieltjes moment sequence for every @xmath462 .
suppose , on the contrary , that @xmath466 is not injective . using ( i ) , we see that there exists @xmath652 such that @xmath783 .
we easily verify that @xmath784 is an s - representing measure of the stieltjes moment sequence @xmath785 ( cf . ) . applying ( * ? ? ?
* lemma 38 ) to @xmath786 , we are led to a contradiction .
the following result is an analog of proposition [ golfclub3 ] .
[ golfclub4 ] assume that holds . if @xmath551 is as in theorem _
[ golfclub2 ] _ with @xmath571 and @xmath466 generates stieltjes moment sequences , then 1 .
@xmath771 for every @xmath683 , 2 .
@xmath787 , 3 .
@xmath466 is injective whenever the stieltjes moment sequence @xmath788 is s - determinate .
arguing as in the proof of proposition [ golfclub3 ] ( but now applying ( * ? ? ?
* lemma 38 ) to @xmath789 ) , we get ( i ) and ( iii ) .
\(ii ) suppose , on the contrary , that @xmath790 . then , by , @xmath791 and @xmath792 .
it follows from ( i ) that @xmath783 for every @xmath683 . as a consequence , @xmath793 , @xmath794 and @xmath795 , which contradicts ( *
* lemma 1.1(ii ) ) because @xmath466 generates stieltjes moment sequences . similar reasoning as in the proof of part ( iii ) of proposition [ golfclub3 ] gives a criterion for injectivity of @xmath466 in a more general situation .
[ sdeterm ] suppose that holds . if @xmath466 generates stieltjes moment sequences and the stieltjes moment sequence @xmath796 is s - determinate for every @xmath462 , then @xmath466 is injective .
the above considerations leads us to the following injectivity problem ( see @xcite for the necessary definitions ) .
[ open2 ] suppose that @xmath797 is a @xmath3-finite measure space , @xmath465 is a nonsingular self - map of @xmath63 and @xmath466 is a composition operator in @xmath467 with symbol @xmath465 generating stieltjes moment sequences .
is it true that @xmath466 is injective ?
problem [ open2 ] seems to be hard to solve . to shed more light on this we make the following remark .
first , we note that problem [ open2 ] has an affirmative solution for bounded composition operators . indeed , by lambert s theorem ( we use ( * ? ?
* theorem 7 ) again ) , a bounded composition operator in an @xmath0-space generating stieltjes moment sequences is subnormal and consequently , by ( * ? ? ?
* theorem 9d ) , it is injective .
if the composition operator in question is over a rootless directed tree and it has sufficiently many quasi - analytic vectors , then the property of generating stieltjes moment sequences is equivalent to subnormality ( use lemma [ jfa - ex ] , its proof and ( * ? ? ?
* theorem 5.3.1 ) ) .
hence , in view of ( * ? ? ?
* corollary 6.3 ) , in this particular case , problem [ open2 ] has an affirmative solution as well ( this can be also deduced from proposition [ sdeterm ] by applying and proposition [ wouk ] ) .
propositions [ ibj1 ] , [ golfclub3 ] , [ golfclub4 ] and [ sdeterm ] provide yet another examples for which problem [ open2 ] has an affirmative solution .
the following assumption will be used frequently through this paper . @xmath798
it is easily seen that @xmath799 we begin by deriving a formula for iterated inverses of @xmath465 at the point @xmath24 . [ gc1 ] suppose holds . if @xmath800 for some @xmath801 and @xmath802 , then @xmath803 we proceed by induction on @xmath804 .
the case of @xmath805 is easily verified .
this and imply that @xmath806 suppose that holds for a fixed @xmath807
. then , by and induction hypothesis , we have @xmath808 which completes the proof .
applying to @xmath465 and @xmath809 and using lemma [ gc1 ] , we can easily calculate the radon - nikodym derivatives @xmath810 and @xmath811 .
[ hfi2 ] suppose holds . then for every @xmath462 , @xmath812 \frac{\mu(x_0 ) + \sum_{i=1}^{\eta } \mu(x_{i,1})}{\mu(x_{\kappa } ) } & \text{if } x = x_{\kappa } , \\[1ex ] \frac{\mu(x_{i , j+1})}{\mu(x_{i , j } ) } & \text{if } x = x_{i , j } \text { with } i \in j_{\eta } \text { and } j \in { \mathbb n}. \end{cases } \end{aligned}\ ] ] if @xmath800 for some @xmath801 and @xmath813 , then @xmath814 1 + \sum_{i=1}^{\eta } \sum_{l=1}^j \frac{\mu(x_{i , l(\kappa+1)})}{\mu(x_{\kappa } ) } & \text { if } j { \geqslant}1 \text { and } r=0 , \\[1ex ] \frac{\mu(x_{r-1})}{\mu(x_{\kappa } ) } + \sum_{i=1}^{\eta } \sum_{l=0}^j \frac{\mu(x_{i , l(\kappa+1)+r})}{\mu(x_{\kappa } ) } & \text { if } r \in j_{\kappa}. \end{cases } \end{aligned}\ ] ] now we calculate the radon - nikodym derivatives @xmath32 , @xmath815 , at the vertices lying on the circuit .
suppose holds . then @xmath816 if @xmath817 and @xmath176 , then by and we have @xmath818 which gives . applying induction on @xmath819 and , we obtain .
the subsequent lemma plays an essential role in the present paper .
[ fund ] if holds , then @xmath820 observe that by and we have @xmath821 which completes the proof of .
the question of density of domains of powers of @xmath466 can be answered in terms of the radon - nikodym derivatives @xmath32 , @xmath815 , calculated at @xmath23 .
[ hfi4 ] suppose holds and @xmath348
. then the following conditions are equivalent _ : _ 1 .
@xmath822 is dense in @xmath467 , 2 .
@xmath823 for every @xmath802 . moreover , if @xmath824 , then the following conditions are equivalent _ : _ 1 .
@xmath825 is dense in @xmath467 , 2 .
@xmath826 is dense in @xmath467 for all @xmath827 , 3 .
@xmath828 for all @xmath829 and @xmath462 , 4 .
@xmath830 for all @xmath653 .
( i)@xmath172(ii ) by , @xmath831 for all @xmath832 , and thus , by with @xmath833 , @xmath834 for all @xmath835 . since , by ,
@xmath836 for every @xmath837 , we see that @xmath836 for every @xmath838 .
( ii)@xmath172(i ) applying , we deduce that @xmath839 for every @xmath840 .
it follows from that @xmath841 for all @xmath775 and @xmath827 .
this , and yield ( i ) .
now we prove the `` moreover '' part . by and (
* theorem 4.7 ) , it suffices to prove that ( vi ) implies ( iv ) .
it follows from that @xmath842 for all integers @xmath843 . applying
, we deduce that @xmath842 for all @xmath801 .
hence , by the implication ( ii)@xmath172(i ) , @xmath826 is dense in @xmath467 for all @xmath684 .
[ porow ] suppose holds and @xmath348
. then the following conditions are equivalent _ : _ 1 .
@xmath844 and @xmath845 , 2 .
@xmath823 for every @xmath802 and @xmath846 .
this section deals with the consistency condition which , according to theorem [ glowne ] , automatically implies subnormality of @xmath847 ( because , under our standing assumption , @xmath848 for all @xmath462 ) .
[ cc1 ] suppose holds , @xmath466 is densely defined and @xmath504 is a family of borel probability measures on @xmath54
. then @xmath504 satisfies if and only if the following three conditions are satisfied _ : _ 1 .
@xmath849 for all @xmath850 and @xmath144 , 2 .
@xmath851 for all @xmath144 , 3 .
@xmath852 for all @xmath40 , @xmath853 and @xmath854 .
moreover , if @xmath504 satisfies , then 1 .
@xmath855 ) = 0 $ ] for all @xmath856 , @xmath40 and @xmath829 , 2 .
@xmath857 for all @xmath144 , 3 .
@xmath858 , 4 .
@xmath859 , 5 .
@xmath860 for all @xmath861 , 6 .
@xmath862 . since , by , @xmath863 , we infer from proposition [ hfi2 ] that @xmath864 assume now that @xmath504 satisfies .
substituting @xmath865 with @xmath866 into , we get @xmath867 an induction argument shows that ( i ) holds . substituting @xmath868 into and using ( i ) , we obtain @xmath869 this implies ( ii ) . substituting @xmath870 into yields @xmath871 an induction argument leads to ( iii ) .
similar reasoning shows that the conditions ( i)-(iii ) imply that @xmath504 satisfies . to prove the `` moreover '' part
, we assume that @xmath504 satisfies the condition .
since @xmath872 for all @xmath459 , we deduce from ( ii ) and that @xmath873 . hence , by ( ii ) again , @xmath874 ) = 0 $ ] for all @xmath652 and @xmath875 . applying ( i ) and ( iii ) gives ( iv ) .
it follows from ( ii ) that @xmath876 using ( iv ) and and integrating the function @xmath877 with respect to the measure @xmath878 , we obtain ( v ) . since @xmath879
, the conditions ( vi ) and ( vii ) follow from ( v ) .
the equality ( viii ) is a direct consequence of ( iv ) , ( v ) and ( vii ) . finally ,
integrating the function @xmath880 with respect to the ( positive ) measure @xmath881 , we deduce from ( v ) and that @xmath882 applying proposition [ hfi4 ] with @xmath883 and @xmath833 , we get ( ix ) .
the following proposition provides new criteria for @xmath847 to have densely defined @xmath33th power ( cf . proposition [ hfi4 ] ) .
[ wk ] suppose holds , the composition operator @xmath466 is densely defined and @xmath504 is a family of borel probability measures on @xmath54 that satisfies
. then for every @xmath348 , the following conditions are equivalent _ : _ 1 .
@xmath884 is densely defined , 2 .
@xmath885 , 3 .
@xmath886 . by and theorem [ cc1](iv )
, the sequence @xmath887 is monotonically increasing .
hence , by proposition [ hfi4 ] , the conditions ( i ) and ( ii ) are equivalent . integrating the function @xmath888 with respect to the measure @xmath889 and using and theorem [ cc1](viii )
, we deduce that the conditions ( ii ) and ( iii ) are equivalent . as a consequence of theorem [ cc1](iv ) and proposition [ wk ]
, we have the following corollary ( cf .
corollary [ porow ] ) .
[ re - m ] suppose holds , @xmath466 is densely defined and @xmath504 is a family of borel probability measures on @xmath54 that satisfies .
let @xmath348 .
then @xmath890 if and only if the following two conditions hold : 1 . @xmath886 , 2 . @xmath891 . in procedure
[ proceed ] below , we propose a method of constructing all possible subnormal composition operators @xmath847 in @xmath892 that satisfy in the meaning that they admit families of probability measures satisfying .
the starting point of our procedure is a family @xmath893 of borel probability measures on @xmath54 that satisfies the conditions - below .
let us point out that if a densely defined @xmath847 admits a family @xmath894 of borel probability measures on @xmath54 that satisfies , then , by theorem [ cc1 ] , the measures @xmath895 , @xmath652 , satisfy the conditions - .
[ proceed ] let @xmath571 , @xmath896 , @xmath465 be a self - map of a set @xmath63 , @xmath577 and @xmath578 be two disjoint systems of distinct points of @xmath63 that satisfy and .
let @xmath897 be a family of borel probability measures on @xmath54 that satisfies the following three conditions : @xmath898 ) = 0 ,
\quad i \in j_{\eta } , \\
\label{i11 } \int_0^\infty t^j p(x_{i,1 } , \operatorname{d\hspace{-0.25ex}}t ) < \infty , \quad j \in
{ \mathbb n } , \,i \in j_{\eta } , \\
\label{i12 } \int_0^\infty \frac{t^{\kappa+1}}{t^{\kappa+1}-1 } p(x_{i,1 } , \operatorname{d\hspace{-0.25ex}}t ) < \infty ,
\quad i \in j_{\eta}. \end{gathered}\ ] ] let @xmath899 be a family of positive real numbers such that @xmath900 it follows from and that @xmath901 using and , we get @xmath902 now , by , we can take @xmath903 such that @xmath904 then we define the borel measure @xmath905 on @xmath54 by @xmath906 by and , @xmath905 is a probability measure such that @xmath907 .
moreover , @xmath905 satisfies the condition ( ii ) of theorem [ cc1 ] .
since @xmath895 , @xmath908 , are probability measures , we infer from and that @xmath909 for all @xmath829 and @xmath40 .
this enables us to define the family @xmath910 of positive real numbers by @xmath911 and the family @xmath912 of borel measures on @xmath54 by @xmath913 in view of , the family @xmath912 consists of probability measures . according to , and , @xmath914 for every @xmath813 .
hence , we can define positive real numbers @xmath915 , @xmath817 , via @xmath916 as a consequence , the measures @xmath917 , @xmath918 , defined by @xmath919 are borel probability measures on @xmath54 .
let @xmath356 be the discrete measure on @xmath63 such that @xmath920 and @xmath921 for all @xmath922 , @xmath652 and @xmath827 , and let @xmath507 be the corresponding composition operator in @xmath467 with @xmath923 . by , and , @xmath466 is densely defined . applying theorem [ cc1 ]
, we see that the family @xmath504 satisfies .
hence , by theorem [ glowne ] , @xmath466 is subnormal .
our procedure enables us to model all subnormal composition operators @xmath847 that satisfy and have densely defined @xmath33th power ( @xmath33 is a fixed positive integer ) .
it suffices to replace by the condition @xmath924 ( leaving the assumptions and unchanged ) and to choose a family @xmath925 that satisfies , in place of , the following inequality @xmath926 indeed , by , the conditions and imply and , respectively . on the other hand , under the assumptions - , @xmath884 is densely defined if and only if holds ( cf.proposition [ wk ] ) . as a consequence
( see also corollary [ re - m ] ) , @xmath927 if and only if both and hold , where @xmath928 using procedure [ proceed ] , we will show that for every @xmath348 , there exists a subnormal composition operator @xmath466 such that @xmath884 is densely defined , while @xmath929 is not .
examples of this kind have been given in @xcite by using weighted shifts on directed trees ( see also a recent paper @xcite for more subtle examples ) .
fix @xmath930 .
consider a sequence @xmath931 of borel probability measures on @xmath54 given by @xmath932 for all @xmath861 and @xmath933 .
set @xmath934 for @xmath933 .
it is now a routine matter to verify that the conditions , , and hold for @xmath49 and for arbitrary @xmath576 .
hence , applying procedure [ proceed ] , we get a composition operator @xmath466 with the required properties , i.e. , @xmath890 .
note that by and proposition [ hfi4 ] , @xmath935 is densely defined for every @xmath829 whenever @xmath936 . in this section
we give criteria for subnormality of composition operators @xmath466 in @xmath937 with @xmath938 and @xmath923 written in terms of the radon - nikodym derivatives @xmath25 calculated at the points @xmath23 and @xmath26 , @xmath27 ( cf .
theorem [ suff ] ) .
surprisingly , in the case of @xmath46 the subnormality of @xmath466 can be inferred from the behaviour of @xmath25 only at the point @xmath23 ( cf .
proposition [ eta1 ] ) .
we begin by stating two necessary lemmata .
[ det ] let @xmath156 be a stieltjes moment sequence and let @xmath939 .
then @xmath447 is a stieltjes moment sequence and the following assertions hold _ : _ 1 . if @xmath447 is s - determinate , then so is @xmath156 , 2 . if @xmath447 is s - determinate and @xmath940 is a stieltjes moment sequence , then @xmath156 is s - determinate and its unique s - representing measure vanishes on @xmath168 , 3 .
if @xmath941 is a stieltjes moment sequence , then @xmath942 satisfies the carleman condition if and only if @xmath943 satisfies the carleman condition .
\(i ) let @xmath141 be an s - representing measure of @xmath156 , @xmath944 be a function given by @xmath945 for @xmath946 , and @xmath947 be a borel measure on @xmath54 given by @xmath948 for @xmath861 . using the measure transport theorem
, we see that @xmath447 is a stieltjes moment sequence with the s - representing measure @xmath947 .
if @xmath949 is another s - representing measure of @xmath156 , then the measure @xmath950 , being an s - representing measure of @xmath447 , coincides with @xmath947 , and consequently @xmath951 .
\(ii ) in view of ( i ) , @xmath156 is s - determinate .
denote by @xmath141 its unique s - representing measure .
let @xmath70 be an s - representing measure of @xmath940 .
then @xmath952 for all finite sequences @xmath953 . by lemma [ kontr ] and the s - determinacy of @xmath447
, we deduce that @xmath954 .
hence @xmath955 .
\(iii ) set @xmath956 for @xmath176 . since @xmath957 for all @xmath176
, we infer from proposition [ wouk](ii ) that if @xmath942 satisfies the carleman condition , then so does @xmath958 . to prove the converse implication assume that @xmath958 satisfies the carleman condition .
note that @xmath959 let @xmath960 be an s - representing measure of @xmath958 .
set @xmath961 for @xmath176 . if @xmath962 , then , by , @xmath963 ) + \gamma_0\big ) , \quad n \in { \mathbb n}. \end{aligned}\ ] ] if @xmath964 , then using and the fact that the sequence @xmath965 is monotonically increasing , we obtain @xmath966 ) + \gamma_0 + \vardelta_n^{\prime}\big ) { \leqslant}n \bigg(\frac{\tau([0,1 ] ) + \gamma_0}{\tau((1,\infty ) ) } + 1\bigg ) \vardelta_n , \quad n \in { \mathbb n}. \end{aligned}\ ] ] combining and completes the proof .
the next lemma is a direct consequence of lemma [ fund ] .
[ suf1 ] assume that holds , @xmath967 for every @xmath348 and @xmath38 is a stieltjes moment sequence with an s - representing measure @xmath895 for every @xmath652 .
then @xmath968 is a stieltjes moment sequence with an s - representing measure @xmath70 given by @xmath969 the above enables us to prove the aforementioned criteria for subnormality .
[ suff ] assume that holds and @xmath34 is a stieltjes moment sequence for every @xmath970 .
if one of the following four conditions is satisfied , @xmath968 is a stieltjes moment sequence . ]
_ : _ 1 .
@xmath971 is s - determinate and there exists an s - representing measure @xmath905 of @xmath28 such that @xmath972 , 2 .
@xmath973 and @xmath974 are s - determinate , 3 .
@xmath974 satisfies the carleman condition , 4 .
@xmath975 satisfies the carleman condition , then @xmath466 is subnormal . by proposition [ hfi4 ]
, @xmath466 is densely defined .
it follows from lemmata [ det ] and [ suf1 ] that the conditions ( iii ) and ( iv ) are equivalent . if ( iv ) holds , then by proposition [ wouk](i ) and lemma [ det](i ) , @xmath976 is s - determinate , and thus , because of ( iv)@xmath172(iii ) , the condition ( ii ) holds .
applying lemma [ det](ii ) , we see that ( ii ) implies ( i ) .
all this means that it suffices to prove that ( i ) implies the subnormality of @xmath466 .
to this end , assume that ( i ) holds .
let @xmath977 be an s - representing measure of @xmath38 for @xmath775 .
note that @xmath978 as @xmath979 , the set - function @xmath980 is a ( positive ) measure .
applying lemma [ suf1 ] and using the s - determinacy assumption , we deduce that the condition ( ii ) of theorem [ cc1 ] holds .
now we define the measures @xmath981 and @xmath982 by the conditions ( i ) and ( iii ) of theorem [ cc1 ] , respectively . using and the fact that @xmath983 is an s - representing measure of @xmath34 for every @xmath984
, we verify that @xmath985 is a family of borel probability measures on @xmath54 that satisfies . applying theorem [ glowne ]
, we conclude that @xmath466 is subnormal .
the situation changes drastically if @xmath12 equals @xmath10 .
[ eta1 ] suppose holds and @xmath46
. then @xmath986 , @xmath987 for every @xmath462 and the following conditions are equivalent _ : _ 1 .
there exists a family @xmath985 of borel probability measures on @xmath54 that satisfies , 2 .
@xmath28 is a stieltjes moment sequence which has an s - representing measure @xmath141 vanishing on @xmath168 , 3 .
@xmath988 for all finite sequences @xmath170 of complex numbers . moreover , if any of the above conditions holds , then @xmath466 is subnormal .
it follows from , and proposition [ hfi4 ] that @xmath986 and @xmath987 for every @xmath462 .
( i)@xmath172(ii ) apply theorem [ glowne ] and theorem [ cc1](iv ) .
( ii)@xmath172(i ) set @xmath989 . by lemma [ fund ]
, @xmath990 is a stieltjes moment sequence with an s - representing probability measure @xmath991 given by @xmath992 clearly , the condition ( ii ) of theorem [ cc1 ] holds .
next , we define @xmath993 , the borel measures on @xmath54 , using the condition ( i ) of theorem [ cc1 ] . since @xmath905 is an s - representing measure of @xmath28 , we deduce from that the so - defined measures are probabilistic .
finally , we define @xmath994 , the borel measures on @xmath54 , using the condition ( iii ) of theorem [ cc1 ] . noting that @xmath995 we see that the measures @xmath994 are probabilistic .
now , applying theorem [ cc1 ] , we conclude that @xmath996 satisfies .
( ii)@xmath173(iii ) apply lemma [ kontr ] .
the `` moreover '' part is a direct consequence of theorem [ glowne ] .
regarding the implication ( ii)@xmath172(i ) of proposition [ eta1 ] , we note that the assumption that @xmath28 is a stieltjes moment sequence is not sufficient for @xmath466 to be subnormal , even if @xmath997 .
[ final ] first , we show that if @xmath998 then there exists a discrete measure @xmath356 on @xmath999 such that @xmath1000 for every @xmath176 with @xmath1001 . for this , take any @xmath1002 and set @xmath1003 for every @xmath918 .
next we put @xmath1004 since @xmath1005 for every @xmath110 , we see that @xmath1006 for every @xmath827 . clearly , by , @xmath1007 for every @xmath813 . using induction , lemma [ fund ] , and
, we verify that @xmath1000 for every @xmath1008 .
it is worth mentioning that if holds and @xmath46 , then the sequence @xmath156 defined by @xmath1009 for every @xmath176 satisfies .
indeed , this is a direct consequence of proposition [ eta1 ] , lemma [ fund ] and .
now , we can apply the above procedure as follows . take @xmath1010 and set @xmath1011 for every @xmath176 .
clearly , @xmath156 is an s - determinate stieltjes moment sequence with the s - representing measure @xmath1012 . since @xmath1013
we see that @xmath1005 for every @xmath110 . applying our procedure
, we get a discrete measure @xmath356 on @xmath63 such that @xmath1000 for every @xmath176 .
note that @xmath1014 .
indeed , by and , we have @xmath1015 this , combined with and , yields @xmath1014 .
it is worth mentioning that @xmath466 is not subnormal ( hence , by proposition [ wn0gr ] below , @xmath990 is not a stieltjes moment sequence ) . indeed , otherwise , by ( * ?
* theorem 13 ) , @xmath466 admits a family @xmath504 of borel probability measures on @xmath54 that satisfies .
this , together with proposition [ eta1 ] , contradicts the s - determinacy of @xmath156 .
the question of characterizing subnormality of bounded composition operators of the form @xmath847 has a simple solution .
[ wn0gr ] suppose holds and @xmath1016
. then @xmath466 is subnormal if and only if @xmath34 is a stieltjes moment sequence for every @xmath1017 .
since , by , @xmath1018 for all @xmath176 , the sufficiency follows from theorem [ suff](iii ) . the sufficiency can also be deduced from lambert s theorem ( cf .
@xcite ) by using and . the necessity is a direct consequence of the corresponding part of lambert s theorem .
now we discuss the question of `` optimality '' of the assumptions of proposition [ wn0gr ] . as shown in example
[ final ] , for @xmath46 and for every @xmath576 , there exists a non - subnormal @xmath781 such that @xmath28 is a stieltjes moment sequence . in example
[ prz2 ] below , we will show that for @xmath46 and @xmath42 the assumption that @xmath990 is a stieltjes moment sequence is not sufficient for @xmath466 to be subnormal , even if @xmath781 .
[ prz2 ] assume that @xmath46 and @xmath42 .
take any positive real numbers @xmath1019 and @xmath1020 .
let @xmath156 be a stieltjes moment sequence with an s - representing measure @xmath70 such that @xmath1021 , @xmath1022 and @xmath1023 .
then @xmath1024 which implies that @xmath156 is s - determinate ( see e.g. , proposition [ wouk](i ) ) .
set @xmath1025 for every @xmath1026 .
clearly , @xmath1027 for all @xmath110 , which means that @xmath990 is a stieltjes moment sequence . since @xmath942 is monotonically decreasing , we see that @xmath1028 for every @xmath829 .
this , together with , implies that @xmath1029 .
note that @xmath1030 this and imply that @xmath1031 now we show that @xmath28 is not a stieltjes moment sequence .
indeed , otherwise by and ( * ? ? ?
* exercise 4(e ) , p. 71 ) , we see that @xmath1032 , where @xmath141 is an s - representing measure of @xmath28 .
hence @xmath1033 which contradicts @xmath1023 .
by proposition [ wn0gr ] , @xmath466 is not subnormal .
we begin this section by providing necessary and sufficient conditions for the extendibility of a given family of borel probability measures on @xmath54 indexed by @xmath1034 to a family of borel probability measures on @xmath54 satisfying . [
main ] suppose holds and @xmath466 is densely defined
. then the following assertions hold . 1 .
if @xmath504 is a family of borel probability measures on @xmath54 that satisfies , then 1 .
@xmath38 is a stieltjes moment sequence for every @xmath652 , 2 . for every @xmath652 , the measure @xmath895 is an s - representing measure of @xmath38 vanishing on @xmath1035 $ ] , 3 .
@xmath1036 for every @xmath817 , 4 .
( i - a ) _ holds and @xmath897 is a family of borel probability measures on @xmath54 satisfying , and , then there exists a family @xmath1037 of borel probability measures on @xmath54 such that @xmath504 satisfies .
\(i ) the conditions and follow from , theorem [ glowne ] and theorem [ cc1](iv ) .
the condition is a direct consequence of theorem [ cc1](vi ) . using the conditions ( i ) , ( vii ) and ( viii ) of theorem [ cc1 ] and integrating the function @xmath1038 with respect to the measure @xmath905
, we obtain @xmath1039 which implies .
\(ii ) for @xmath652 and @xmath1026 , we define the measure @xmath1040 by the condition ( iii ) of theorem [ cc1 ] .
note that @xmath1041 because @xmath1042 next , we define the measures @xmath889 , , @xmath1043 by @xmath1044 for @xmath144 and @xmath1045 with @xmath1046 . by , and , the quantity @xmath1047 is well - defined , @xmath1048 $ ] and @xmath1049 is a well - defined finite measure .
that it is probabilistic follows from the equalities @xmath1050 now , integrating the function @xmath1038 with respect to @xmath889 , we see that the condition ( i ) of theorem [ cc1 ] is satisfied . finally , integrating the function @xmath1051 ( which is positive on @xmath1052 ) with respect to @xmath889
, we deduce that the condition ( ii ) of theorem [ cc1 ] is satisfied .
this combined with theorem [ cc1 ] completes the proof .
below we introduce the condition , a weaker version of the condition of theorem [ main ] that will lead us to constructing exotic examples ( cf .
section [ sec5 ] ) . for more information concerning the conditions ( e1 ) and ( e2 ) below
, the reader is referred to lemma [ snieg ] and remark [ snieg2 ] . [
6.2 ] suppose holds , the condition _
( i - a ) _ of theorem _ [ main ] _ is satisfied and @xmath1053 is a family of borel probability measures on @xmath54 satisfying the conditions and of theorem _ [ main ] _ and the condition below _ : _ 1 .
@xmath1054 .
consider the following three conditions _ : _ 1 .
@xmath1055 is an s - determinate stieltjes moment sequence for every @xmath446 , 2 .
@xmath1056 is an s - determinate stieltjes moment sequence for every @xmath446 , 3 .
@xmath1057 is a stieltjes moment sequence that satisfies the carleman condition . then _
( e3)_@xmath172_(e2)_@xmath172_(e1 ) _ and , if _ ( e1 ) _ holds , then the composition operator @xmath466 is densely defined and there exists a family @xmath1058 of borel probability measures on @xmath54 such that @xmath504 satisfies . moreover , if _ ( e3 ) _ holds , then @xmath1059 satisfies the carleman condition for every @xmath1060 .
it follows from and that the quantity @xmath1061 defined below @xmath1062 is a real number . by , and , we have @xmath1063 in particular , the above series are convergent in @xmath54 . first , we show that for all @xmath110 and @xmath652 , and in the case of @xmath49 , @xmath1064 for any integer @xmath33 such that @xmath1065 .
however , this series may be divergent to infinity for some integers @xmath1066 . ]
@xmath1067 to this end , we fix @xmath110 .
then @xmath1068 for some @xmath1069 and @xmath1070 .
we begin by considering the case of @xmath1071 . since then @xmath1072 is the remainder of the division of @xmath1073 by @xmath1074 , we infer from that @xmath1075 if @xmath833 , then mimicking the above proof we get @xmath1076 which proves .
it follows from that @xmath1077 ( the case of @xmath1078 can be deduced from with @xmath833 . ) ( e3)@xmath172(e2 ) apply the assertions ( i)-(iii ) of proposition [ wouk ] .
( e2)@xmath172(e1 ) since the class of stieltjes moment sequences is closed under the operation of taking pointwise limits ( which follows from the stieltjes theorem , cf .
* theorem 6.2.5 ) ) and @xmath1079 for all @xmath110 , we see that @xmath1080 is a stieltjes moment sequence .
let @xmath141 be an s - representing measure of @xmath1080 .
then , by , we have @xmath1081 where @xmath70 is the borel measure on @xmath54 given by @xmath1082 now we prove that the condition of theorem [ main ] is satisfied . indeed , otherwise @xmath1083 . by ( e2 ) ,
the stieltjes moment sequence @xmath1084 using and ( i - b ) , and substituting @xmath1085 into , we deduce that @xmath1086 , a contradiction .
this shows that the condition of theorem [ main ] is satisfied . since @xmath1087 for all @xmath1008
, we infer from proposition [ hfi4 ] that @xmath825 is dense in @xmath467 .
hence , by theorem [ main](ii ) , there exists @xmath1037 , a family of borel probability measures on @xmath54 , such that @xmath504 satisfies . by theorem [ glowne ] , @xmath1088 for all @xmath176 .
this means that for every @xmath164 , @xmath1055 is a stieltjes moment sequence and , by ( e2 ) , @xmath1056 is an s - determinate stieltjes moment sequence , which implies that @xmath1055 is s - determinate for every @xmath164 ( see ( * ? ? ? * proposition 5.12 ) ; see also ( * ? ? ?
* lemma 2.4.1 ) ) .
assume now that ( e1 ) holds .
passing to the limit , as in the proof of ( e2)@xmath172(e1 ) , we see that @xmath1057 is a stieltjes moment sequence .
let @xmath949 be an s - representing measure of @xmath1057 .
then , by , we have @xmath1089 where @xmath1090 is the borel measure on @xmath54 given by @xmath1091 arguing as in the paragraph containing ( using ( e1 ) in place of ( e2 ) ) , we deduce that the condition of theorem [ main ] is satisfied . according to proposition [ hfi4 ] ,
@xmath466 is densely defined . applying theorem [ main](ii ) , we get the required family of borel probability measures on @xmath54 that satisfies .
now we prove the `` moreover '' part .
assume ( e3 ) holds .
first note that , by what has been proved above , the assumptions of theorem [ glowne ] are satisfied .
hence , by proposition [ hfi4 ] , @xmath34 is a stieltjes moment sequence for every @xmath462 .
since @xmath1057 satisfies the carleman condition , we deduce from and proposition [ wouk](ii ) that @xmath1092 satisfies the carleman condition as well .
an induction argument completes the proof .
the following is related to the `` moreover '' part of the conclusion of theorem [ 6.2 ] .
suppose holds , @xmath1059 is a stieltjes moment sequence for every @xmath1093 , and @xmath1057 is s - determinate . then for every @xmath1093 , @xmath1094 is s - determinate .
it follows from applied to @xmath833 that the stieltjes moment sequence @xmath1095 is s - determinate .
this combined with ( * ? ? ?
* proposition 5.12 ) ( see also ( * ? ? ?
* lemma 2.1.1 ) ) implies that @xmath1096 is s - determinate as well .
thus , an induction argument completes the proof .
in the last part of the paper we construct non - hyponormal injective composition operators generating stieltjes moment sequences .
the construction relies on the key observation that there is a gap between the conditions and of theorems [ main ] and [ 6.2 ] .
the indices of h - determinacy of the sequences @xmath34 , @xmath459 , are discussed as well .
we begin by introducing for any @xmath738 and @xmath576 a class of composition operators over the directed graph @xmath1097 which admit families @xmath897 of borel probability measures on @xmath54 satisfying the conditions , , and of theorems [ main ] and [ 6.2 ] , but not of theorem [ main ] ( see procedure [ proceed2 ] and lemma [ pierwsze ] ) . the construction of these operators essentially depends on a choice of specific n - extremal measures @xmath70 and @xmath960 , and a partition @xmath1098 of @xmath1099 .
the fundamental properties of the so - constructed operators , including the characterization of their hyponormality for @xmath42 , are proven in lemma [ pierwsze ] .
section [ s5.3 ] shows that the gap between the conditions and does exist .
theorem [ czwarte ] is the culminating result of the present paper .
it shows that there exists a non - hyponormal composition operator generating stieltjes moment sequences over the locally finite directed graph @xmath1 .
its proof heavily depends on the existence of n - extremal probability measures @xmath1100 and @xmath141 satisfying a restrictive condition which is not easy to deal with . fortunately , there are n - extremal probability measures coming from shifted al - salam - carlitz or from a quartic birth and death process that satisfy this condition .
it is worth pointing out that , without the use of these special n - extremal measures , step 1 of the proof of theorem [ czwarte ] combined with implies that for every sufficiently large integer @xmath12 , there exists a non - hyponormal composition operator generating stieltjes moment sequences over the locally finite directed graph @xmath48 .
[ proceed2 ] fix @xmath571 and @xmath896 .
suppose that @xmath1101 \label{gs2 } 1=\inf{\mathrm{supp}(\nu ) } < \inf{\mathrm{supp}(\tau ) } , \\
\label{gs3 } \nu({\mathbb r}_+ ) = 1 + \nu(\{1\ } ) , \\
\label{gs4 } 1 + \int_0^\infty \frac{1}{t^{\kappa } } \operatorname{d\hspace{-0.25ex}}\tau(t ) > \tau({\mathbb
r}_+ ) , \end{gathered}\ ] ] and @xmath1102 ( a partition of a nonempty set is always assumed to consist of nonempty sets . ) since , by lemma [ pron ] , @xmath1103 , such partition always exists .
it follows from lemma [ pron ] , and that @xmath1104 for every @xmath652 ( see section [ n&t ] for the definition of @xmath1105 ) . hence , @xmath1106 set @xmath1107 clearly , @xmath895 is a borel probability measure on @xmath54 such that @xmath1108 for all @xmath1008 and @xmath652 ( use and ) .
take any @xmath1109 and define the family @xmath1110 of positive real numbers by @xmath1111 next , we define the family @xmath1112 of positive real numbers and the family @xmath1113 of borel probability measures on @xmath54 by @xmath1114 let @xmath1115 be the borel measure on @xmath54 given by @xmath1116 by and , @xmath1115 is a probability measure .
it is also clear that @xmath1117 for every @xmath110 . in view of and
, we have @xmath1118 and @xmath1119 if @xmath1120 , then we set @xmath1121 it follows from and that @xmath1122 for every @xmath1123 . finally , let @xmath356 be the ( unique ) discrete measure on @xmath938 such that @xmath464 for every @xmath459 ( we follow the convention ) , and let @xmath466 be the corresponding composition operator in @xmath467 with the symbol @xmath923 . since @xmath475 , we infer from that @xmath474 for every @xmath1124 .
[ pierwsze ] let @xmath41 , @xmath12 , @xmath70 , @xmath960 , @xmath1098 , @xmath63 , @xmath356 , @xmath1115 , @xmath1125 and @xmath466 be as in procedure _ [ proceed2]_. then @xmath1126 and the following holds _ : _ 1 .
@xmath1057 is an h - determinate stieltjes moment sequence with the s - representing measure @xmath1115 whose index of h - determinacy at @xmath456 is @xmath456 , 2 . for all @xmath652 and @xmath827
, @xmath1127 is a stieltjes moment sequence with the s - representing measure @xmath1040 , 3 .
the condition of theorem _ [ main ] _ holds and the family @xmath897 satisfies the conditions , and of theorems _ [ main ] _ and _ [ 6.2 ] _ , 4 .
@xmath1053 does not satisfy the condition _ _ of theorem _ [ main ] _ , 5 . if @xmath42 , then @xmath466 is hyponormal if and only if @xmath1128 6 . if @xmath42 , then @xmath466 generates stieltjes moment sequences , 7 . if @xmath42 and @xmath46 , then there exists a family @xmath1129 of borel probability measures on @xmath54 that satisfies and thus @xmath466 is subnormal .
\(ii ) it follows from and that @xmath1130 moreover , we have @xmath1131 altogether this implies that ( ii ) holds .
in particular , the condition of theorem [ main ] holds . by , , and
, the family @xmath897 satisfies the condition of theorem [ main ] .
\(iv ) the condition of theorem [ main ] , , and yield @xmath1132 hence , the family @xmath1133 satisfies the condition of theorem [ 6.2 ] and the quantity @xmath1061 defined by is a real number .
since the closed support of an n - extremal measure has no accumulation point in @xmath53 ( cf.lemma [ pron ] ) , we infer from and that @xmath1134 thus , noting that @xmath1135 1 - \nu({\mathbb r}_+ ) \overset{\eqref{gs3 } } = - \nu(\{1\ } ) \text { if } \kappa=0 , \end{cases } \end{aligned}\ ] ] where ( @xmath1136 ) refers to if @xmath1120 and to if @xmath42 , we obtain @xmath1137 it follows from that the assertion ( iv ) holds .
\(iii ) in view of what has been done , it remains to show that the condition of theorem [ main ] holds for @xmath1138 .
suppose @xmath1138 and @xmath1139 .
using , and , we get @xmath1140 1 - \nu({\mathbb r}_+ ) \overset { \eqref{gs3 } } = - \nu(\{1\ } ) \text { if } r=\kappa , \end{cases } \end{aligned}\ ] ] where ( @xmath1141 ) refers to and .
this and yield @xmath1142 it follows from and that the family @xmath1053 satisfies the condition of theorem [ main ] . therefore ( iii ) holds .
\(i ) arguing as in the proof of theorem [ 6.2 ] , we verify that is satisfied .
hence , applying , and we get @xmath1143 this means that @xmath1057 is a stieltjes moment sequence with the s - representing measure @xmath1115 . employing , , and ( * ? ? ?
* theorem 3.6 ) , we deduce that @xmath1057 is h - determinate , @xmath1144 if @xmath1145 and @xmath1146 if @xmath1147 .
hence , by , @xmath1148 .
\(v ) it follows from the cauchy - schwarz inequality that @xmath1149 hence , the inequality holds for every @xmath1150 .
note that @xmath1151 therefore , the inequality holds for @xmath789 if and only if is satisfied .
this combined with proposition [ buda ] yields ( v ) .
[ epslem ] let @xmath1152 be such that @xmath1153 , @xmath1154 and @xmath1155 , where @xmath1156 is an injective sequence .
set @xmath1157 and @xmath1158 for @xmath933 and @xmath402 @xmath135see and for the necessary definitions@xmath136
. then the following holds _ : _ 1 .
@xmath1159 and @xmath1160 for all @xmath402 , 2 .
there exists @xmath1161 such that @xmath1162 for every integer @xmath1163 , 3 . if @xmath1161 is such that @xmath1164 , then there exists @xmath1165 such that for every @xmath1166 , @xmath1167 4 .
there exists @xmath1168 such that for every @xmath1169 , @xmath1170 5 . if @xmath1171 , then there exists @xmath1172 such that for every @xmath1173 , @xmath1174 \(iii ) it follows from our assumptions that @xmath1176 and @xmath1177 .
hence , there exists @xmath1178 such that @xmath1179 using , we easily verify that @xmath1180 applying the measure transport theorem , we obtain @xmath1181 combining , and yields ( iii ) . since the function @xmath1190 maps @xmath55 injectively into @xmath56
, we deduce from euler s pentagonal - number theorem ( cf .
* theorem 14.3 ) ) that @xmath1191 it is easily seen that there exists @xmath1192 such that @xmath1193 hence , by , we have @xmath1194 this combined with @xmath1195 and @xmath1196 imply that @xmath1197 which completes the proof . in this section , we show that the assertion ( ii ) of theorem [ main ] is no longer true if the condition of theorem [ main ] is replaced by the condition of theorem [ 6.2 ] .
as shown in theorem [ drugie1 ] below , this can happen even for subnormal composition operators .
in fact , this phenomenon is independent of whether the operator in question is subnormal or not ( cf .
remark [ whynot ] ) .
[ drugie ] let @xmath576 and @xmath1198 .
then there exists a discrete measure @xmath356 on @xmath938 such that the composition operator @xmath466 in @xmath467 with @xmath923 has the property that @xmath986 and the following conditions hold _ : _ 1 .
@xmath1057 is an h - determinate stieltjes moment sequence with index of h - determinacy at @xmath456 equal to @xmath456 , 2 . for all @xmath652 and @xmath827
, @xmath1127 is an h - determinate stieltjes moment sequence with infinite index of h - determinacy ; in particular , the condition of theorem _ [ main ] _ holds , 3 .
there exists a unique family @xmath897 of borel probability measures on @xmath54 that satisfies the conditions , and of theorems _ [ main ] _ and _ [ 6.2]_@xmath1199 this family does not satisfy the condition _ _ of theorem _ [ main ] _ , 4 .
there is no family @xmath1129 of borel probability measures on @xmath54 that satisfies , 5 .
if @xmath42 , then @xmath466 generates stieltjes moment sequences . first , we note that if ( i ) holds , then by proposition [ hfi4 ] , @xmath825 is dense in @xmath467 .
we begin by proving that the condition ( iv ) follows from ( i ) , ( ii ) and ( iii ) .
suppose , on the contrary , that there exists a family @xmath1129 of borel probability measures on @xmath54 that satisfies .
it follows from theorem [ glowne ] that @xmath1200 is an s - representing measure of @xmath38 for every @xmath27 .
hence , by ( ii ) and ( iii ) , @xmath1201 for every @xmath27 .
this together with theorem [ main](i ) implies that the family @xmath1202 satisfies the condition of theorem [ main ] , which contradicts ( iii ) .
take any s - indeterminate stieltjes moment sequence @xmath1203 ( see e.g. , the classical example due to stieltjes @xcite ; see also section [ secasc ] ) .
let @xmath199 and @xmath200 be the krein and the friedrichs measures of @xmath116 ( see section [ secbc ] ) . then by lemma [ pron ] and , @xmath1204 .
hence , replacing @xmath1205 by @xmath1206 with @xmath1207 if necessary , we may assume that @xmath1208 it follows from lemma [ epslem](v ) that there exists @xmath402 such that holds .
this combined with and lemma [ transh ] shows that the measures @xmath1209 and @xmath1210 satisfy the conditions - of procedure [ proceed2 ] .
since the measure @xmath960 is n - extremal and @xmath47 , we infer from lemma [ pron ] that there exists a partition @xmath1211 of @xmath1099 such that @xmath1212 let @xmath63 , @xmath356 , @xmath1115 , @xmath1213 and @xmath466 be as in procedure [ proceed2 ] .
we will show that the operator @xmath466 has all the required properties .
indeed , set @xmath1214 for every @xmath40 .
we deduce from the equality and theorem [ b - d ] that each @xmath1215 is h - determinate and @xmath1216 fix @xmath1217 and @xmath417 .
it follows from that @xmath1218 .
hence , by and , we see that for every @xmath1219 , @xmath1220 this , , and proposition [ krytnd ] imply that the measure @xmath1221 is h - determinate and @xmath1222 . applying lemma [ pierwsze ]
, we conclude that @xmath466 satisfies the conditions ( i)-(v ) .
this completes the proof .
[ drugie1 ] let @xmath576 .
then there exists a discrete measure @xmath356 on @xmath999 such that the composition operator @xmath466 in @xmath467 with @xmath1001 has the property that @xmath986 and the following conditions hold _ : _ 1 .
@xmath1057 is an h - determinate stieltjes moment sequence with index of h - determinacy at @xmath456 equal to @xmath456 , 2 . for every @xmath827
, @xmath1223 is an s - indeterminate stieltjes moment sequence@xmath1199 in particular , the condition of theorem _ [ main ] _ holds , 3 .
there exists a borel probability measure @xmath1224 on @xmath54 that satisfies the conditions , and of theorems _ [ main ] _ and _ [ 6.2 ] _ , and does not satisfy the condition _ _ of theorem _ [ main ] _ , 4 . if @xmath42 , then there exists a family @xmath1225 of borel probability measures on @xmath54 that satisfies and , consequently , @xmath466 is subnormal . as in the proof of theorem [ drugie ] , we see that there exist measures @xmath70 and @xmath960 satisfying the conditions - of procedure [ proceed2 ] .
the only possible partition @xmath1211 of @xmath1099 with @xmath46 is @xmath1226 .
let @xmath63 , @xmath356 , @xmath1115 , @xmath1227 and @xmath466 be as in procedure [ proceed2 ] .
in view of lemma [ pierwsze ] , it remains to show that for every @xmath827 , the stieltjes moment sequence @xmath1223 is s - indeterminate .
suppose , on the contrary , that there exists @xmath410 such that @xmath1228 is s - determinate .
we show that then @xmath990 is s - determinate . for this
we may assume that @xmath709 .
it follows from that @xmath1229 hence , by ( * ? ? ?
* proposition 5.12 ) , @xmath1230 is s - determinate . applying backward induction , we deduce that @xmath990 is s - determinate .
since , by lemma [ pierwsze](ii ) , @xmath991 is an s - representing measure of @xmath990 and @xmath1231 ( see and ) , we infer from ( * ? ? ?
* corollary , p. 481 ) ( see also ( * ? ? ?
* lemma 2.2.5 ) ) that @xmath991 is h - determinate . by , there exists @xmath1232 such that @xmath1233 this combined with and proposition [ krytnd ] ( applied to @xmath960 and @xmath1234 ) implies that @xmath960 is h - determinate , which contradicts .
the main purpose of this section is to construct non - hyponormal composition operators @xmath466 in @xmath1235 that generate stieltjes moment sequences with @xmath1236 and @xmath1237 .
recall that @xmath847 is always injective ( cf . ) .
we begin with the case of @xmath49 . 1 .
@xmath466 is injective and generates stieltjes moment sequences , 2 .
@xmath466 is not hyponormal , 3 .
@xmath1240 is an h - determinate stieltjes moment sequence with index of h - determinacy at @xmath456 equal to @xmath456 , 4 .
@xmath34 is an h - determinate stieltjes moment sequence with infinite index of h - determinacy for all @xmath1241 .
arguing as in the proof of theorem [ drugie ] , we see that there exist measures @xmath199 and @xmath200 satisfying .
let @xmath1156 be an injective sequence such that @xmath1242 ( cf .
lemma [ pron ] ) .
it follows from lemma [ epslem](iv ) that there exists @xmath1243 such that holds .
this and lemma [ transh ] show that the measures @xmath1209 and @xmath1210 satisfy the conditions - of procedure [ proceed2 ] for @xmath42 and @xmath49 .
note also that in view of lemma [ epslem](i ) , @xmath1244 .
set @xmath1245 for @xmath933 .
clearly , the sequence @xmath1246 satisfies the condition for @xmath49 .
let @xmath63 , @xmath356 , @xmath985 and @xmath466 be as in procedure [ proceed2 ] for @xmath42 and @xmath49 .
it follows from , and that @xmath1247 for all @xmath1248 .
this is easily seen to imply ( iv ) .
since yields @xmath1249 an application of lemma [ pierwsze ] completes the proof .
now we concern ourselves with the case when @xmath12 is an arbitrary integer greater than or equal to @xmath15 .
the most striking case is that of @xmath1250 .
below , given @xmath331 , we write @xmath1251 for the largest integer not greater than @xmath35 , and @xmath1252 for the smallest integer not less than @xmath35 . 1 .
@xmath466 is injective and generates stieltjes moment sequences , 2 .
@xmath466 is not hyponormal , 3 .
@xmath34 is a stieltjes moment sequence for every @xmath462 , 4 .
@xmath1240 is h - determinate with index of h - determinacy at @xmath456 equal to @xmath456 , 5 . for all @xmath1254 and @xmath827 ,
the sequence @xmath1127 is h - determinate with infinite index of h - determinacy , 6 . for every @xmath1255 ,
the sequence @xmath1256 is h - determinate and its unique h - representing measure @xmath1257 satisfies the following condition @xmath1258 indeed , by lemma [ pron ] , we can extend the sequence @xmath1262 to an injective sequence @xmath1156 such that @xmath1155 .
it follows from and lemma [ epslem](iii ) that there exists @xmath402 such that the inequality holds for @xmath1263 .
this together with and lemma [ transh ] imply that the measures @xmath1209 and @xmath1210 satisfy the conditions - of procedure [ proceed2 ] for @xmath42 , and @xmath1264 .
now we define a partition @xmath1211 of @xmath1099 by @xmath1265 \{\theta_{\eta}^{(a ) } , \theta_{\eta+1}^{(a ) } , \dots\ } & \text { if } i = \eta .
\end{cases } \end{aligned}\ ] ] let @xmath63 , @xmath356 , @xmath985 and @xmath466 be as in procedure [ proceed2 ] for @xmath42 . since @xmath1253 , it follows from that @xmath1266 now applying lemma [ pierwsze ] we see that for every @xmath459 , @xmath34 is a stieltjes moment sequence with the s - representing measure @xmath983 ( in particular , ( iii ) holds ) , and @xmath466 satisfies ( i ) , ( ii ) and ( iv ) ( note that if @xmath46 , then , by lemma [ pierwsze](vii ) , @xmath466 is hyponormal ) .
arguing as in the proof of theorem [ trzecie ] , we deduce that ( v ) holds as well .
our next aim is to prove ( vi ) .
assume that @xmath1267 .
set @xmath1268 . clearly @xmath1269 . by (
* theorem 3.6 ) , @xmath1270 is h - determinate and @xmath1271 put @xmath1272 for @xmath827 .
it follows from and that @xmath1273 since , by , @xmath1274 , we infer from that @xmath1275 fix @xmath1255 . substituting @xmath1276 into and using and proposition [ krytnd ] , we deduce that @xmath1277 is h - determinate .
hence , applying , with @xmath1278 , and proposition [ krytnd ] , we see that @xmath1279 take now @xmath1219 such that the measure @xmath1280 is h - determinate .
since @xmath1281 , there exists @xmath1282 such that for every @xmath1283 , @xmath1284 applying and proposition [ krytnd ] again , we deduce that @xmath1285 .
this implies that @xmath1286 combining the above inequality with , we obtain .
indeed , let @xmath1187 .
by lemma [ euler ] , there exists @xmath1288 such that @xmath1289 as in ( * ? ? ? * proposition 4.5.1 ) ( see also ( * ? ? ?
* eq . ( 4.4 ) ) ) , we define the borel measures @xmath1100 and @xmath141 on @xmath53 ( with different notation ) by @xmath1290 ( another possible choice of measures @xmath1100 and @xmath141 is discussed in remark [ piate ] . ) in view of the proof of ( * ? ? ?
* eq . ( 5.10 ) ) , @xmath1100 and @xmath141 are probability measures .
it follows from ( * ? ? ?
* proposition 4.5.1 ) that @xmath1100 and @xmath141 are n - extremal measures of the same stieltjes moment sequence ( see also section [ secasc ] for a detailed discussion of this matter ) .
note that @xmath1291 .
following the proof of theorem [ drugie ] , we set @xmath1292 and @xmath1293 with @xmath1294 .
it is easily seen that the measures @xmath199 and @xmath200 satisfy .
now , combining with , we get @xmath1295 to finish the proof of the theorem , take measures @xmath199 and @xmath200 as in step 2
. then @xmath1155 , where @xmath1156 is a strictly increasing sequence ( cf .
lemma [ pron ] ) .
this implies that @xmath1296 , and thus the inequality holds for every @xmath1253 .
applying step 1 completes the proof . to the best of our knowledge
, there are few explicit examples ( or rather classes of examples ) of n - extremal measures .
the ones used in the proof of theorem [ czwarte ] are related to the al - salam - carlitz ( see section [ secasc ] ) . in turn , the measures discussed below come from an s - indeterminate stieltjes moment problem associated with a quartic birth and death process ( cf.@xcite ) .
[ piate ] define the borel measures @xmath1100 and @xmath141 on @xmath53 by @xmath1297 where @xmath1298 ( @xmath1299 stands for the euler gamma function . )
it was proved in ( * ? ? ?
* proposition 3.4.1 ) that @xmath1100 and @xmath141 are n - extremal measures of the same stieltjes moment sequence .
a careful inspection of @xcite reveals that these measures are probabilistic . as a consequence , @xmath1300 .
note that @xmath1296 , where , as in the proof of step 2 of theorem [ czwarte ] , @xmath1293 with @xmath1294 and @xmath1301 .
indeed , since @xmath1302 , @xmath1303 and @xmath1304 , we get @xmath1305 ( in fact , @xmath1306 . )
putting all this together , we see that the proof of theorem [ czwarte ] goes through without change if the measures @xmath1100 and @xmath141 are defined by instead of .
[ fri ] let @xmath1100 and @xmath141 be measures as in .
recall that they are representing measures of the same stieltjes moment sequence , say @xmath116 .
since @xmath1307 , we infer from that @xmath1100 is the krein measure of @xmath116 . in turn
, the measure @xmath141 is the friedrichs measure of @xmath116 .
indeed , this can be deduced by combining$ ] is replaced by @xmath1308 , where @xmath199 is as in ( * ? ? ?
* proposition 3.2 ) . ] and ( * ? ? ?
* proposition 3.1 ) with ( * ? ? ?
* proposition 3.2 ) and ( * ? ? ?
* lemma 4.4.2 ) ( that this particular @xmath141 is the friedrichs measure of @xmath116 follows also from theorem [ taso](i ) and the discussion in the paragraph containing ) .
hence , by corollary [ transh+ ] ( see also theorem [ taso ] ) , the measures @xmath200 and @xmath960 are the friedrichs measures of appropriate s - indeterminate stieltjes moment sequences ( see step 1 of the proof of theorem [ czwarte ] ) ; however , @xmath70 is not the krein measure ( of the s - indeterminate stieltjes moment sequence @xmath1309 ) .
it is easily seen that the above discussion applies to the measures @xmath1100 and @xmath141 defined by and the corresponding measures @xmath70 and @xmath960 ( but now we have to use ( * ? ? ?
* corollary 3.3.3 ) in place of ( * ? ? ?
* lemma 4.4.2 ) ) .
the construction of the composition operator @xmath466 that appears in the proofs of theorems of section [ sec5 ] depends on the choice of a partition @xmath1098 of @xmath1099 ( cf .
procedure [ proceed2 ] ) .
in particular , the hyponormality of @xmath466 requires finding a partition @xmath1098 satisfying the inequality .
hence , it seems to be of some independent interest to calculate the infimum and the supremum of the quantity appearing on the left - hand side of over all partitions @xmath1098 of @xmath1099 .
the following general proposition sheds more light on this issue .
[ supinf ] let @xmath960 be an n - extremal measure such that @xmath1311 .
then for every @xmath738 , the following two conditions hold _ : _ @xmath1312 where @xmath1313 if @xmath49 , then the inequality in becomes an equality . moreover , we have @xmath1314 we begin with two observations . first , if @xmath1211 is a partition of @xmath1099 , then by lemma [ pron ] , @xmath1315 for every @xmath652 .
second , due to the cauchy - schwarz inequality , the following two inequalities hold @xmath1316 therefore , if @xmath1098 is a partition of @xmath1099 , then @xmath1317 this proves and the inequality `` @xmath1318 '' in . to show the reverse inequality
, we may assume that @xmath47 .
let us define for every @xmath410 a partition @xmath1319 of @xmath1099 by @xmath1320 where @xmath1321 is a strictly increasing sequence such that @xmath1155 .
by lemma [ pron ] such a sequence exists and @xmath1322 .
then @xmath1323 hence , by applying the lebesgue monotone and dominated convergence theorems , we get the inequality `` @xmath1324 '' in .
this completes the proof of . to prove the remaining part of the conclusion , we define for every @xmath1325 a partition @xmath1326 of @xmath1099 by ( compare with ) @xmath1327 \{\theta_{\eta } , \theta_{\eta+1 } , \ldots\ } & \text{if } \eta <
\infty \text { and } i=\eta .
\end{cases } \end{aligned}\ ] ] then @xmath1328 applying the lebesgue monotone convergence theorem and using , we get . in the case of @xmath49
, we have @xmath1329 this completes the proof of the proposition .
it is worth pointing out that under the assumptions of proposition [ supinf ] , the following two inequalities hold @xmath1330 whenever @xmath1331 is a finite or infinite partition of a nonempty subset @xmath67 of @xmath1099 ( argue as in the proof of ) .
we will show that if , in addition , @xmath1332 for at least one @xmath1333 , then the second inequality ( counting from the left ) in is strict . indeed , otherwise taking a close look at
shows that the cauchy - schwarz inequality becomes an equality for every @xmath1334 . as a consequence , for every @xmath1335 ,
there exists @xmath1336 such that @xmath1337 for every @xmath1338 ( recall that @xmath1339 ) .
this implies that @xmath1340 for every @xmath1335 , which contradicts our assumption .
it follows from the previous paragraph that for every @xmath571 and for every partition @xmath1211 of @xmath1099 such that @xmath1341 we have @xmath1342 however , if @xmath1343 ( and consequently @xmath49 ) , then the strict inequality in becomes an equality ( cf . ) . applying and
, we deduce that the sequence @xmath1344 is monotonically increasing to @xmath1345 .
hence , there arises the question whether or not there exists @xmath1346 at which the above sequence stabilizes .
a substantial part of this paper was written while the second and the fourth author visited kyungpook national university during the spring and the autumn of 2015 .
they wish to thank the faculty and the administration of this unit for their warm hospitality .
99 w. a. al - salam , l. carlitz , some orthogonal , _ math .
* 30 * ( 1965 ) , 47 - 61 .
t. m. apostol , _ introduction to analytic number theory _ , undergraduate texts in mathematics , springer - verlag , new york - heidelberg , 1976 .
r. b. ash , _ probability and measure theory _ , harcourt / academic press , burlington , 2000 . c. berg , j. p. r. christensen , p. ressel , _ harmonic analysis on semigroups _ , springer , berlin , 1984 . c. berg , a. j. durn , the index of determinacy for measures and the @xmath1347-norm of orthonormal polynomials , _ trans .
soc . _ * 347 * ( 1995 ) , 2795 - 2811 . c. berg , m. thill , rotation invariant moment problems , _
acta math_. * 167 * ( 1991 ) , 207 - 227 .
c. berg , m. thill , a density index for the stieltjes moment problem , _ orthogonal polynomials and their applications ( erice , 1990 ) _ , 185 - 188 , _ imacs ann .
_ , * 9 * , baltzer , basel , 1991 .
c. berg , g. valent , the nevanlinna parametrization for some indeterminate stieltjes moment problems associated with birth and death processes , _ methods appl .
* 1 * ( 1994 ) , 169 - 209 .
e. bishop , spectral theory for operators on a banach space , _ trans .
* 86 * ( 1957 ) , 414 - 445 .
j. bram , subnormal operators , _ duke math
. j. _ * 22 * ( 1955 ) , 75 - 94 .
p. budzyski , a note on unbounded hyponormal composition operators in @xmath0-spaces , _
_ volume 2012 , doi:10.1155/2012/902853 .
p. budzyski , p. dymek , z. j. jaboski , j. stochel , subnormal weighted shifts on directed trees and composition operators in @xmath0 spaces with non - densely defined powers , _ abs .
* 2014 * ( 2014 ) , article i d 791817 , 6 pp .
p. budzyski , z. j. jaboski , i. b. jung , j. stochel , unbounded subnormal weighted shifts on directed trees , _ j. math .
* 394 * ( 2012 ) , 819 - 834 .
p. budzyski , z. j. jaboski , i. b. jung , j. stochel , unbounded subnormal weighted shifts on directed trees .
ii , _ j. math . anal .
* 398 * ( 2013 ) , 600 - 608 .
p. budzyski , z. j. jaboski , i. b. jung , j. stochel , on unbounded composition operators in @xmath0-spaces , _ ann .
* 193 * ( 2014 ) , 663 - 688 .
p. budzyski , z. j. jaboski , i. b. jung , j. stochel , unbounded subnormal composition operators in @xmath0-spaces , _ j. funct
. anal . _ * 269 * ( 2015 ) , 2110 - 2164 .
p. budzyski , z. j. jaboski , i. b.jung , j. stochel , a subnormal weighted shift on a directed tree whose @xmath33th power has trivial domain , _ j. math
* 435 * ( 2016 ) , 302 - 314 .
p. budzyski , z. j. jaboski , i. b. jung , j. stochel , unbounded weighted composition operators in @xmath0-spaces , http://arxiv.org/abs/1310.3542 . j. t. campbell , w. e. hornor , seminormal composition operators .
_ j. operator theory _ * 29 * ( 1993 ) , 323 - 343 . t. carleman , _ les fonctions quasi analytiques _ , gauthier - villars , paris , 1926 .
t. s. chihara , on determinate hamburger moment problems , _ pacific j. math .
_ * 27 * ( 1968 ) , 475 - 484 .
t. s. chihara , _ an introduction to orthogonal polynomials _ ,
mathematics and its applications , vol .
13 , gordon and breach science publishers , new york , 1978 .
j. b. conway , _ the theory of subnormal operators _ , mathematical surveys and monographs , * 36 * , american mathematical society , providence , ri , 1991 . j. b. conway , _ a course in operator theory _ , graduate studies in mathematics , * 21 * , american mathematical society , providence , ri , 2000 .
j. b. conway , n. s. feldman , the state of subnormal operators , _ a glimpse at hilbert space operators _ , 177 - 194 , _ oper .
theory adv .
_ , * 207 * , birkhuser verlag , basel , 2010
. j. b. conway , k. h. jin , s. kouchekian , on unbounded bergman operators , _ j. math
* 279 * ( 2003 ) , 418 - 429 .
r. e. curto , l. a. fialkow , recursively generated weighted shifts and the subnormal completion problem , ii , _ integr .
* 18 * ( 1994 ) , 369 - 426 .
m. r. embry , a generalization of the halmos - bram criterion for subnormality , _ acta sci .
_ ( _ szeged _ ) _ _ * 35 * ( 1973 ) , 61 - 64 .
c. foia , dcompositions en oprateurs et vecteurs propres .
i. , tudes de ces dcompositions et leurs rapports avec les prolongements des oprateurs , _ rev .
roumaine math .
pures appl .
_ * 7 * ( 1962 ) , 241 - 282 .
b. fuglede , the multidimensional moment problem , _ expo .
_ * 1 * ( 1983 ) , 47 - 65 .
r. gellar , l. j. wallen , subnormal weighted shifts and the halmos - bram criterion , _ proc .
japan acad .
_ * 46 * ( 1970 ) , 375 - 378 .
p. r. halmos , normal dilations and extensions of operators , _ summa brasil .
_ * 2 * ( 1950 ) , 125 - 134 .
r. halmos , ten problems in hilbert space , _ bull .
soc . _ * 76 * ( 1970 ) , 887 - 933 .
d. harrington , r. whitley , seminormal composition operators , _ j. operator theory _ * 11 * ( 1984 ) , 125 - 135 .
e. k. haviland , on the momentum problem for distribution functions in more than one dimension .
_ * 58 * ( 1936 ) , 164 - 168 .
d. herrero , _ subnormal bilateral weighted shifts _ , notas mimeografiadas , 1971 .
m. e. h. ismail , a queueing model and a set of orthogonal polynomials , _ j. math .
* 108 * ( 1985 ) , 575 - 594 .
m. e. h. ismail , _ classical and quantum orthogonal polynomials in one variable _ , encyclopedia of mathematics and its applications , vol .
98 , cambridge university press , cambridge , 2005 .
m. e. h. ismail , d. r. masson , @xmath323-hermite polynomials , biorthogonal rational functions , and @xmath323-beta integrals , _ trans .
* 346 * ( 1994 ) , 63 - 116 .
z. jabonski , hyperexpansive composition operators , _ math .
cambridge philos .
* 135 * ( 2003 ) , 513 - 526 . z. j. jaboski , i. b. jung , j. stochel , weighted shifts on directed trees , _ mem .
* 216 * ( 2012 ) , no .
1017 , viii+107pp . z. j. jaboski , i. b. jung , j. stochel , a non - hyponormal operator generating stieltjes moment sequences , _ j. funct .
* 262 * ( 2012 ) , 3946 - 3980 .
p. e. t. jorgensen , commutative algebras of unbounded operators , _ j. math
* 123 * ( 1987 ) , 508 - 527 .
i. b. jung , j. stochel , subnormal operators whose adjoints have rich point spectrum , _ j. funct
. anal . _ * 255 * ( 2008 ) , 1797 - 1816 .
s. kouchekian , the density problem for unbounded bergman operators , _ integr .
oper . theory _
* 45 * ( 2003 ) , 319 - 342 .
s. kouchekian , j. e. thomson , the density problem for self - commutators of unbounded bergman operators , _ integr .
oper . theory _
* 52 * ( 2005 ) , 135 - 147 .
s. kouchekian , j. e. thomson , on self - commutators of toeplitz operators with rational symbols , _ studia math . _
* 179 * ( 2007 ) , 41 - 47 .
a. lambert , subnormality and weighted shifts , _ j. london math . soc .
_ * 14 * ( 1976 ) , 476 - 480 .
a. lambert , subnormal composition operators , _ proc .
soc . _ * 103 * ( 1988 ) , 750 - 754 .
g. mcdonald , c. sundberg , on the spectra of unbounded subnormal operators , _ canad . j. math .
_ * 38 * ( 1986 ) , 1135 - 1148 .
e. nordgren , composition operators on hilbert spaces , lecture notes in math .
* 693 * , springer - verlag , berlin 1978 , 37 - 63 .
h. l. pedersen , stieltjes moment problems and the friedrichs extension of a positive definite operator , _
j. approx . theory _
* 83 * ( 1995 ) , 289 - 307 .
g. plya , g. szeg , _ problems and theorems in analysis , vol .
ii , theory of functions , zeros , polynomials , determinants , number theory , geometry _ , revised and enlarged translation by c. e. billigheimer of the fourth german edition , die grundlehren der mathematischen wissenschaften , band 216 .
springer - verlag , new york - heidelberg , 1976 .
m. riesz , sur le problme des moments et le thorme de parseval correspondant , _ acta litt .
. szeged _ * 1 * ( 1923 ) , 209 - 225 .
w. rudin , _ real and complex analysis _
, mcgraw - hill , new york 1987 .
s. saks , a. zygmund , _ analytic functions _ , monografie matematyczne , tom 28 , pastwowe wydawnietwo naukowe , warsaw 1965 .
b. simon , the classical moment problem as a self - adjoint finite difference operator , _ adv .
_ * 137 * ( 1998 ) , 82 - 203 .
t. stieltjes , recherches sur les fractions continues , _ anns .
* 8 * ( 1894 - 1895 ) , j1-j122 ; * 9 * , a5-a47
. j. stochel , moment functions on real algebraic sets , _ ark . mat . _ * 30 * ( 1992 ) , 133 - 148 .
j. stochel , seminormality of operators from their tensor product , _ proc .
* 124 * ( 1996 ) , 135 - 140 .
j. stochel , j. b. stochel , on the @xmath1348th root of a stieltjes moment sequence , _ j. math
_ , * 396 * ( 2012 ) , 786 - 800 . j. stochel , f. h. szafraniec , on normal extensions of unbounded operators . i , _ j. operator theory _ * 14 * ( 1985 ) , 31 - 55 . j. stochel and f. h. szafraniec , on normal extensions of unbounded operators , ii , _ acta sci . math .
@xmath135szeged@xmath136 _ * 53 * ( 1989 ) , 153 - 177 .
j. stochel , f. h. szafraniec , on normal extensions of unbounded operators .
spectral properties , _ publ .
rims , kyoto univ . _
* 25 * ( 1989 ) , 105 - 139 .
j. stochel , f. h. szafraniec , the complex moment problem and subnormality : a polar decomposition approach , _ j. funct .
* 159 * ( 1998 ) , 432 - 491 .
j. b. stochel , weighted quasishifts , generalized commutation relation and subnormality , _ ann .
_ * 3 * ( 1990 ) , 109 - 128 .
f. h. szafraniec , sesquilinear selection of elementary spectral measures and subnormality , _ elementary operators and applications _ ( blaubeuren , 1991 ) , 243 - 248 , _ world sci . publ . ,
river edge , nj _ , 1992 .
f. h. szafraniec , on normal extensions of unbounded operators .
iv . a matrix construction , _ operator theory and indefinite inner product spaces _ , 337 - 350 , _ oper . theory adv .
_ , * 163 * , birkhuser , basel , 2006 . j. weidmann , _ linear operators in hilbert spaces _ , springer - verlag , berlin , heidelberg , new york , 1980 . | a recent example of a non - hyponormal injective composition operator in an @xmath0-space generating stieltjes moment sequences , invented by three of the present authors , was built over a non - locally finite directed tree .
the main goal of this paper is to solve the problem of whether there exists such an operator over a locally finite directed graph and , in the affirmative case , to find the simplest possible graph with these properties ( simplicity refers to local valency ) .
the problem is solved affirmatively for the locally finite directed graph @xmath1 , which consists of two branches and one loop .
the only simpler directed graph for which the problem remains unsolved consists of one branch and one loop .
the consistency condition , the only efficient tool for verifying subnormality of unbounded composition operators , is intensively studied in the context of @xmath1 , which leads to a constructive method of solving the problem .
the method itself is partly based on transforming the krein and the friedrichs measures coming either from shifted al - salam - carlitz or from a quartic birth and death process . |
in 2005 akiyama _ et al . _
@xcite introduced so - called _ shift radix systems _ ( cf . also @xcite ) . for a natural number @xmath13 and a real vector @xmath14 the mapping @xmath15 defined by @xmath16
is called the _ d - dimensional shift radix system associated with @xmath6 _ ( _ srs _ ) and @xmath6 its _ parameter_. in @xcite the notion has been generalized to the complex setting . for a complex vector @xmath1 the analogously defined mapping @xmath2}^d \to { \mathbb{z}\left[\mathrm{i}\right]}^d$ ]
is called the _ d - dimensional gaussian shift radix system associated with @xmath6 _ ( _ gsrs _ ) ( note that @xmath17 and @xmath18 for @xmath19 ) . let ] @xmath20}^d : { \exists\:}(m , n ) \in { \mathbb{n}}^2 : m \neq n \land \gamma_{{\mathbf{r}}}^m({{\mathbf{a } } } ) = \gamma_{{\mathbf{r}}}^n({{\mathbf{a } } } ) \right\}}\\ { \mathcal{g}_{d}^{(0 ) } } & { \:\colonequals\:}{\left\ { { { \mathbf{r } } } \in { \mathbb{c}}^d \mid { \forall\:}{{\mathbf{a } } } \in { \mathbb{z}\left[\mathrm{i}\right]}^d : { \exists\:}n \in { \mathbb{n } } : \gamma_{{\mathbf{r}}}^n({{\mathbf{a } } } ) = { { \mathbf{0 } } } \right\}}\end{aligned}\ ] ] where for any @xmath21 , @xmath22 means the @xmath23-fold iterative application of @xmath7 to @xmath24 .
the gsrs @xmath7 is said from now on a real vector @xmath6 and its associated gsrs @xmath7 shall be identified in terms of properties . ] to have the _
finiteness property _
iff @xmath25 .
srs are closely related to two important notions of numeration systems .
indeed , as pointed out in @xcite , srs form a generalization of @xmath26-expansions ( see @xcite ) and canonical number systems ( cns ) ( see @xcite and ( * ? ? ?
* section 4.1 ) ) .
the finiteness properties in the contexts of @xmath26-expansions and cns are in one - to - one correspondence with the finiteness property for srs .
gsrs on the other hand are a generalization of gaussian numeration systems @xcite . for a @xmath27}\setminus{\left\ { 0 \right\}}$ ] and @xmath28\mid{\left \lfloor
c/\beta \right \rfloor}=0 \right\}}$ ] the pair @xmath29 is called _ gaussian numeration system _ iff every @xmath30}$ ] can be written uniquely in the form @xmath31 where @xmath32 , @xmath33 for @xmath34 and @xmath35 iff @xmath36 .
it is shown in @xcite that @xmath37 is a gaussian numeration system iff @xmath38 .
furthermore the digit representation of @xmath39 with respect to @xmath37 is given by @xmath40 where @xmath41 and the fractional part of some @xmath19 is defined as @xmath42 .
in the present paper a conjecture on the characterization of @xmath11 is given .
because of its shape and in honor of attila peth @xmath11 is known as peth s loudspeaker @xcite . in section [ sconjecture ]
the set @xmath12 is defined and it is conjectured that @xmath43 . the main result is stated in section [ sresult ] and proven in section [ sinclusion1 ] and section [ sinclusion2 ] where it is shown that @xmath44 and the other inclusion is settled in large parts leaving only a very small area of uncertainty . in section [ sconsequences ] consequences like the loudspeaker s circumference and area are derived under the assumption that the main conjecture holds .
finally all weakly critical and critical points of the loudspeaker are identified in section [ scritical ] .
let = @xmath45 = @xmath46 , = @xmath47let @xmath48 @xmath49 , @xmath50 , @xmath51 + @xmath52 @xmath53 , @xmath47 + @xmath54 @xmath55 , @xmath47 + @xmath56 @xmath57 , @xmath58 + @xmath59 @xmath60 , @xmath61 + @xmath62 @xmath63 , @xmath47 + @xmath64 @xmath65 , @xmath47 + @xmath66 @xmath67 , @xmath47 + @xmath68 @xmath69 , @xmath47 + @xmath70 @xmath71 , @xmath47 + @xmath45 @xmath72 , @xmath47 and let @xmath12 denote the union of the region bounded by the following infinite polygonal chain and the same region reflected at the real axis . the boundary of @xmath12 shall also be as given below where a solid line between two points indicates belonging of the corresponding line segment and an overline over a point indicates belonging of the corresponding vertex to @xmath12 .
@xmath73 @xmath74@xmath75 @xmath74=@xmath76 @xmath77=@xmath78 @xmath77=@xmath79 @xmath74@xmath80 @xmath77@xmath81 @xmath77@xmath82 @xmath74@xmath83@xmath48 @xmath74@xmath84 @xmath74@xmath85 @xmath77 + @xmath86 @xmath74@xmath87 @xmath74@xmath88 @xmath77@xmath89 @xmath77 + @xmath90 @xmath74@xmath91 @xmath77@xmath92 @xmath77@xmath93 @xmath77 + @xmath94 @xmath77@xmath95 @xmath74@xmath96 @xmath77@xmath97 @xmath77@xmath98 @xmath74 + @xmath99 @xmath77@xmath100 @xmath77@xmath101 @xmath74@xmath102 @xmath77@xmath103 @xmath77@xmath104 @xmath74 + @xmath105 @xmath77@xmath106 @xmath77@xmath107 @xmath74@xmath108 @xmath77@xmath109 @xmath77@xmath110 @xmath74@xmath111 @xmath74 + @xmath112 @xmath74@xmath113 @xmath74@xmath114 @xmath77@xmath115 @xmath77@xmath116 @xmath74@xmath117 @xmath77@xmath118 @xmath77@xmath119 @xmath74@xmath120 @xmath74 + @xmath73 @xmath74@xmath75 @xmath74@xmath76 @xmath77@xmath78 @xmath77@xmath79 @xmath74@xmath80 @xmath77@xmath81 @xmath77@xmath82 @xmath74@xmath83 @xmath74 + @xmath121 @xmath74@xmath122 @xmath74@xmath123 @xmath77@xmath124 @xmath77@xmath125 @xmath74@xmath126 @xmath77@xmath127 @xmath77@xmath128 @xmath74@xmath129 @xmath74@xmath130 @xmath77 + @xmath131 + @xmath132 @xmath74@xmath133 @xmath74@xmath134 @xmath77@xmath135 @xmath77@xmath136 @xmath74@xmath137 @xmath77@xmath138 @xmath77@xmath139 @xmath74@xmath140 @xmath74@xmath141 @xmath77 + @xmath131 [ cmain ] if @xmath12 is as defined above then @xmath43
. note that for all @xmath142 .
the following figure shows the part of @xmath12 which lies in the first quadrant and a magnification of the part where it gets regular .
it can be seen that ultimately the boundary of @xmath12 consists of a sequence of pikes which have ten vertices each . for @xmath143 pike
@xmath23 shall refer to the pike which contains the vertex @xmath62 .
[ tmain ] let @xmath12 be as in section [ sconjecture ] and @xmath144 . then * @xmath145 * @xmath146 the first part is proven in section [ sinclusion1 ] and the second part in section [ sinclusion2 ] .
for @xmath143 , @xmath147}^d)^n$ ] is called a _ cycle of @xmath148 _ ( or @xmath7 , see footnote [ fn_identify ] ) iff for all @xmath149 it holds that @xmath150 ( note that @xmath151 has precedence over @xmath152 and @xmath153 ) , a _ cycle _
iff there is a vector @xmath1 for which @xmath154 is a cycle of @xmath6 , and _ nontrivial _ iff @xmath155 , the _
trivial cycle_. let @xmath156 , the _ associated polyhedron of @xmath154 _ or - if @xmath154 is a nontrivial cycle - the _ cutout polyhedron of @xmath154_. @xmath157 is either empty or the intersection of finitely many half spaces and therefore it does in fact always form a - possibly degenerate - convex polyhedron @xcite .
it is clear that @xmath158 which provides a method to `` cut out '' regions ( the cutout polyhedra ) from @xmath159 @xcite .
cutout polyhedra can be used to prove that a given parameter @xmath148 does not belong to @xmath9 .
however they are insufficient to prove that it does belong . for that brunotte
s algorithm can be used ( ( * ? ? ?
* theorem 5.1 ) ) which is based on sets of witnesses .
a set @xmath160}^d$ ] is called a _ set of witnesses for @xmath6 _ iff it is stable under @xmath161 , @xmath162 , @xmath163 , and @xmath164 ( where @xmath165 is the identity on @xmath166}$ ] and @xmath167 is the function on @xmath168 which replaces every entry of the input vector by its complex conjugate ) and contains a generating set of the group @xmath169}^d , + ) $ ] which is closed under taking inverses . every such set of witnesses has the decisive property @xmath170 in the case of a finite set of witnesses this provides a method to decide whether or not a given parameter @xmath6 belongs to @xmath9 .
this is what brunotte s algorithm does for any given parameter @xmath6 in the interior of @xmath159 - it finds a finite set of witnesses .
it shall be denoted by @xmath171 - the _ set of witnesses associated with @xmath6 _ - and can be computed using the following iteration : @xmath172 if @xmath6 is an element of the interior of @xmath159 the mappings @xmath173 , @xmath174 are contractive apart from a finite subset of @xmath166}$ ] . therefore the above iteration becomes stationary eventually @xcite .
let @xmath175 - the _ graph of witnesses associated with @xmath6 _ - denote the edge - colored multidigraph with vertex set @xmath171 having an edge of color @xmath176 from @xmath24 to @xmath177 iff @xmath178 .
if @xmath179 is the set of all edges ( ordered pairs ) of color @xmath176 then the graph @xmath175 is completely characterized by @xmath180}^d)^2)^4 $ ] denotes the power set of a set @xmath181 . ]
( as there are no isolated vertices ) and thus the graph and the @xmath182-tuple can be identified . for any such graph
@xmath183}^d)^2)^4 $ ] let - just as for cycles - @xmath184 and @xmath185 .
if @xmath186 then @xmath175 is finite and @xmath187 is a convex polyhedron .
furthermore @xmath9 is the disjoint union of those @xmath187 the corresponding parameters @xmath6 of which belong to @xmath9 @xcite .
the algorithms introduced in @xcite are based on this fact and can easily be adapted to the complex case .
the results presented in section [ sinclusion2 ] have been achieved in this way . for further considerations
it should be noted that the loudspeaker is symmetric with respect to the real axis @xcite .
in the following we define 19 infinite families of cycles the corresponding cutout + polygons of which will cut out the whole region outside @xmath12 . let @xmath188=@xmath189@xmath190=@xmath188=@xmath188@xmath191@xmath192@xmath193 + @xmath194@xmath195@xmath196@xmath197 + @xmath198@xmath199@xmath200@xmath201 + @xmath202@xmath203@xmath204@xmath205 + @xmath189@xmath190@xmath206@xmath207 + @xmath208@xmath209@xmath210@xmath211 + @xmath212@xmath213@xmath214@xmath215 + @xmath216@xmath217@xmath218@xmath219 + @xmath220@xmath221 + + @xmath222 + = @xmath223 = @xmath224=@xmath225 @xmath226 + @xmath227 @xmath228 + @xmath229 @xmath230 + @xmath231 @xmath232 + @xmath233 @xmath234 + @xmath235 + @xmath236 @xmath237 + @xmath238 @xmath239 + @xmath240 @xmath241 + @xmath242 + @xmath243 @xmath244 + @xmath245 + @xmath223 @xmath246 + @xmath247 + @xmath248 @xmath249 + @xmath250 @xmath251 + @xmath252 + @xmath253 + @xmath254 @xmath255 + @xmath256 + @xmath257 @xmath258 + @xmath259 + @xmath260 + @xmath261 @xmath262 = @xmath263 + @xmath264 + for @xmath265 + @xmath266 + @xmath267 + for @xmath268 + @xmath269 + @xmath270 + @xmath271 + @xmath272 + for @xmath273 + @xmath274 + @xmath275 + @xmath276 + for @xmath277 + @xmath278 + @xmath279 + for @xmath280 + @xmath281 + @xmath282 + for @xmath283 + @xmath284 + @xmath285 + for @xmath286 + @xmath287 + @xmath288 + for @xmath289 + @xmath290 + @xmath291 + @xmath292 + for @xmath293 + @xmath294 + @xmath295 + @xmath296 + for @xmath297 + @xmath298 + @xmath299 + for @xmath300 + @xmath301 + @xmath302 + @xmath303 + for @xmath304 + @xmath305 + @xmath306 + @xmath307 + for @xmath308 + @xmath309 + @xmath310 + @xmath276 + for @xmath311 + @xmath312 + @xmath313 + @xmath276 + for @xmath314 + @xmath315 + @xmath316 + @xmath272 + for @xmath317 + @xmath318 + @xmath319 + @xmath296 + for @xmath320 + @xmath321 + @xmath322 + @xmath307 + for @xmath323 + @xmath324 + @xmath325 + for @xmath326 + @xmath327 + @xmath328 + for @xmath329 + @xmath330 + @xmath331 + for @xmath332 by lemma 5.1 of @xcite one can compute the corresponding cutout polygons .
the following selection of cycles cuts out everything outside @xmath12 which proves theorem [ tmain ] ( i ) : + @xmath333 = @xmath334 = @xmath335 = @xmath336 , @xmath337 , + @xmath338 @xmath339 , @xmath333 @xmath334 , + @xmath340 @xmath341 , @xmath335 @xmath342 , + @xmath343 @xmath344 , @xmath345 @xmath346 , + @xmath347 @xmath348 , @xmath349 @xmath350 , + @xmath351 @xmath352 , @xmath353 @xmath346 , + @xmath354 @xmath355 , @xmath356 @xmath346 , + @xmath357 @xmath358 , @xmath359 @xmath360 , + @xmath361 @xmath362 , @xmath363 @xmath364 , + @xmath365 @xmath366 , @xmath367 @xmath368 , + @xmath369 @xmath370 + the figures below show a regular sector ( where the polygons of the infinite families are sufficient to cut out the respective part ) . in the first figure
it can be seen for @xmath371 that the whole region outside @xmath12 in the sector @xmath372 of the unit disk is being cut out
. it can be shown by comparing the coordinates of the vertices of the polygons that this is the case for every @xmath373 .
the subsequent figure shows the polygons moved apart in groups to illustrate how they fit together .
it can be seen that the dotted lines ( indicating parts of the boundary which do not belong to the corresponding polygon ) of one group hit solid ones ( indicating parts which to belong ) of the other group and vice versa , and that single missing points are also complemented ( indicated by prominent dots at the respective position ) . note that the polygons from the family 19 are needed to cut out a small region remaining in the respective sector if only the families one to 18 are considered .
in fact a single ( but not arbitrary ) polygon of family 19 would be sufficient .
the families were found by manual search .
first a list of cycles was computed the corresponding cutout polygons of which would cover the necessary parts of several successive sectors .
these cycles were then grouped to families by hand .
by applying analogues of the two algorithms introduced in @xcite , section 3 , the proof of theorem [ tmain ] ( ii ) could be achieved ( being much more efficient it was mostly algorithm 2 that was applied ) .
the disk @xmath374 , which covers the settled region , contains all pikes up to and including the 30th .
since the 8th pike is already regular and the general regular structure of the loudspeaker is therefore verified for quite many pikes , it appears reasonable to assume that conjecture [ cmain ] is in fact true . despite best efforts a general proof could not be given by now .
translating the two algorithms of @xcite to the complex setting is straight forward . as pointed out in section [ spreliminaries ] , @xmath9 is the disjoint union of those @xmath187 the corresponding parameters @xmath6 of which belong to @xmath9 .
it is easy to see that , just as in the real case , any given convex hull @xmath375 of finitely many interior points of @xmath159 intersects with only finitely many of the @xmath187 ( cf .
@xcite , theorem 3.2 ) .
thus the analogue of algorithm 1 of @xcite , which essentially computes exactly those @xmath187 which intersect with @xmath376 , also holds for all inputs @xmath376 .
if @xmath377 then @xmath171 is finite and @xmath378 is given by a system of @xmath379 linear inequalities the solution of which is a convex polyhedron . for @xmath380 , @xmath381 , and @xmath382 ( we identify @xmath383 and @xmath166}^1\simeq{\mathbb{z}}^2 $ ] ) the @xmath384 inequalities induced by @xmath385 are given by @xmath386 + @xmath387=@xmath153=@xmath388=@xmath389=@xmath153=@xmath390=@xmath391@xmath392@xmath389@xmath390@xmath393 + @xmath394@xmath395@xmath396@xmath397 + @xmath153@xmath388@xmath389@xmath153@xmath398@xmath393 + @xmath153@xmath399@xmath395@xmath153@xmath400@xmath397 + @xmath401@xmath402@xmath403@xmath404 + @xmath405@xmath406@xmath407@xmath408 + @xmath153@xmath409@xmath402@xmath153@xmath410@xmath404 + @xmath153@xmath411@xmath406@xmath153@xmath412@xmath408 . if @xmath413 then the solution set of the system of inequalities above is the intersection of @xmath182 half - open squares with side lengths @xmath414 ( if @xmath415 then the solution set is of course equal to @xmath416 ) .
the @xmath182 squares are arranged in a way such that the intersection of them is either a singleton , an open line segment , or a nondegenerate , open , convex polygon . both algorithms from @xcite
can now be applied in the same way as for real srs with the only difference being the systems of inequalities one has to consider .
if the loudspeaker coincides with @xmath12 , its perimeter is two times the sum of all distances of successive vertices of the boundary of the intersection of @xmath12 and the first quadrant . if @xmath43 then the perimeter of the loudspeaker is + @xmath417 + @xmath418 + which is approximately @xmath419 .
the area can easily be calculated using the fact that @xmath12 is star - shaped with respect to the origin .
the total area is just two times the sum of the areas of all triangles where two vertices are successive vertices of the boundary of the intersection of @xmath12 and the first quadrant and the third one is @xmath420 .
if @xmath43 and @xmath421 denotes the digamma function then the area of the loudspeaker is + @xmath422 + which is approximately @xmath423 .
one consequence of theorem [ tmain ] ( i ) is the following proposition on weakly critical and critical points of @xmath11 .
a weakly critical point is a point @xmath148 any open neighborhood of which intersects with infinitely many cutout polyhedra .
a critical point is a point @xmath148 any open neighborhood @xmath424 of which satisfies that @xmath425 can not be covered by finitely many cutout polyhedra .
both notions were first introduced in @xcite , section 7 . for @xmath143 the line through @xmath62 and @xmath64 hits the origin and has a gradient of @xmath428 .
let @xmath429 ( we identify @xmath416 and @xmath430 ) such that @xmath431 and @xmath432 , and @xmath433}$ ] such that @xmath434 and @xmath435 .
then @xmath436 lies on the unit circle in the sector between the real axis and the line through @xmath437 and @xmath438 .
then one can deduce the following cases for the product @xmath439 : + @xmath440 + @xmath441 + @xmath442 + @xmath443 + so the product , which is just @xmath444 rotated by the argument of @xmath436 , is contained in the unit square lying next to @xmath444 in rotational direction .
this implies a specific behavior of @xmath445 if @xmath446 : + @xmath447 + @xmath448 + @xmath449 + @xmath450 + @xmath451 + @xmath452 + therefore the orbits of the gaussian integers @xmath453 and @xmath454 both end up in @xmath420 and cover the set @xmath455}\mid { \left \lvert a \right \rvert } + { \left \lvert b \right \rvert } \leq n \land ( ( a \leq 0 \lor b \leq 0 ) \rightarrow { \left \lvert a \right \rvert } + { \left \lvert b \right \rvert } < n)\}$ ] . for geometric reasons it follows that the orbit of any element of @xmath456 ends up in @xmath420 even if @xmath457 . in conclusion : + @xmath458 ( @xmath459 is a possible choice ) + therefore all cycles of any @xmath436 having the properties above have empty intersection with @xmath456 , which forces them to grow beyond all bounds as @xmath23 increases .
thus infinitely many cycles are needed to cut out , say , the set @xmath460 from the loudspeaker which actually is being cut out entirely ( theorem [ tmain ] ( i ) ) .
it follows that @xmath426 is a critical point .
+ since for all @xmath461}$ ] , @xmath462 is a cycle of @xmath463 and @xmath464 is a cycle of @xmath465 , it follows that @xmath466 and @xmath467 are weakly critical points .
theorem [ tmain ] ( i ) implies that the intersection of the closure of @xmath11 and the boundary of @xmath468 ( unit circle ) consists of these three points which implies that there are no other critical or weakly critical points . | for @xmath0 and @xmath1 let @xmath2}^d \to { \mathbb{z}\left[\mathrm{i}\right]}^d$ ] , where @xmath3 @xmath4 for @xmath5 , denote the _ ( d - dimensional ) gaussian shift radix system associated with @xmath6_. @xmath7 is said to have the _
finiteness property _
iff all orbits of @xmath7 end up in @xmath8 ; the set of all corresponding @xmath1 is denoted by @xmath9 .
it has a very complicated structure even for @xmath10 . in the present paper a conjecture on the full characterization of @xmath11 - which is known as peth s loudspeaker - is formulated and proven in substantial parts .
it is shown that @xmath11 is contained in a conjectured characterizing set @xmath12 .
the other inclusion is settled algorithmically for large regions leaving only small areas of uncertainty .
furthermore the circumference and area of the loudspeaker are computed under the assumption that the conjecture holds .
the proven parts of the conjecture also allow to fully identify all so - called critical and weakly critical points of @xmath11 . |
Subsets and Splits
Data Mining Abstracts
This query searches for and retrieves 2000 documents containing the phrase 'data mining' in the abstract, providing limited but specific document samples.